Publications that met the inclusion criteria were predominantly classified as case-control or cross-sectional studies (Figure 4). Although all of the studies classified as case-control included a control group, as per our definition (see Section 3.1), not all studies provided appropriate matching of the control group to the amputee group. The most common problem encountered in this regard was comparison of healthy, able-bodied persons to diabetic amputees.5,43,46,51,52,55
4.2 REVIEW AND EVALUATION OF LITERATURE
Given that no systematic reviews were identified and the number of experimental trials was small, these works were reported and evaluated in conjunction with observational studies. The results are presented in discrete sections describing the biomechanics of gait following PFA in terms of temporospatial features, GRF and CoP, kinematics, kinetics, plantar pressure, EMG, and energy expenditure.
4.3 TEMPORAL AND SPATIAL ASPECTS OF GAIT
4.3.1 DESCRIPTION OF STUDIES
A small body of literature consisting of 10 articles9,38,42–48,55, three abstracts,51,52,54 and one thesis 6 reported temporospatial characteristics of PFA gait. One article36 was not included because the methods were inadequately described, and the results were implausible. The vast majority of these publications reported only walking velocity,9,38,43–47,51,55 usually in the context of a barrage of functional measures,44,47 or as part of studies focusing on other aspects of gait,45,55 such as plantar pressure measurement. Several articles reported step or stride length data42,48,52,54 or cadence.6,52 One investigation reported a more complete range of parameters with additional information on gait cycle times, proportions of stance and swing, as well as support phase data.6 These studies were predominantly observational in design,6,9,38,42,43,47,51,52,54,55 with a few experimental investigations44,45,48 comparing the effects of orthotic and footwear interventions in a controlled manner.
A variety of measurement equipment was used in these investigations, ranging from a stop watch to determine walking velocity43,44,47,55 to foot switches51,52 and gait laboratories for more complete analysis.6,42,48,54 Some studies did not adequately report the equipment used.9,46
Studies reporting temporospatial parameters focused, almost exclusively, on populations with transmetatarsal (TMT) amputation.6,9,38,43–45,47,48,52,54,55 The number of investigations reporting on toe,38,42,51,52 metatarsophalangeal (MTP),6,38,51 and ray resections9,38,51 and Lisfranc6,9 and Chopart6,9 amputees were more limited. The number of amputee participants in a single investigation was typically between 5 and 30, with the larger sample sizes consisting of smaller groups of subjects with different levels of amputation. The majority of these works report on populations with diabetes and vascular disease,38,43–47,51,52,55 with few investigating individuals with amputation because of trauma or other nondiabetic-related causes.6,9,48,54
4.3.2 METHODOLOGICAL QUALITY
There was a range of methodological issues affecting these investigations. Randomization of experimental conditions or the inadequate reporting of these procedures was an issue in several investigations.44,45,48 Amputee cohorts tended to be quite heterogeneous in terms of the time since amputation,5,43–45,47,55 amputation level including the number of toes amputated,9,38,51,52 age,9,43 and involvement of the contralateral limb.43–45,47 Aside from Mueller et al.,43 who described testing for inclusion of bilateral amputees in their discussion, most investigators failed to demonstrate the reasonableness of pooling data from subjects with disparate amputation levels or bilateral involvement. There are several instances in which the cohort was inadequately described in terms of level or cause of amputation.42,46,48 A couple of investigations drew comparisons between the amputee group and a suitably matched control group38,46 that accounted for the effects of systemic diseases, such as diabetes and peripheral neuropathy.
220.127.116.11 Walking Velocity
There exists strong evidence that persons with PFA, with a history of diabetes and vascular disease, walk relatively slowly at around 0.85 m/second38,43–45,47,55 or two-thirds that of a healthy, able-bodied population,51,52 with little difference evident based on amputation level. The consistency of this finding among investigations is striking. Although it includes authors who published multiple works on various aspects of gait from the same data collection,44,45,47 there are enough independent investigations38,43,51,52,55 to be confident of this observation.
In populations without diabetes and vascular disease, there is a moderate level of evidence describing that walking velocity is similar to that of healthy controls. The time series investigation by Tang et al.48 reported on a “mostly traumatic population” (p. 125), but the walking velocity observed in this amputee cohort (0.84 m/second) was comparable to studies of vascular amputees. By contrast, the observational studies by Greene and Cary,9 Wilson and Dillon,54 and Dillon6 reported faster walking velocities, which were comparable to those of healthy people. Greene and Cary9 reported walking velocity in groups with TMT/ray amputation (1.3 m/second) or Lisfranc/midtarsal/Chopart amputation without equinus (1.1 m/second) that were similar to the results of the case series presented by Dillon6 and a case study by Wilson and Dillon54 in which individuals with quite varied amputation levels walked between 1.18 and 1.46 m/second.
A number of these investigations compared an amputee population with systemic diseases, including diabetes and vascular disease, with a healthy, nonamputee, control group,43,51,52,55 making it difficult to separate the effects of amputation from those of the systemic disease. Two observational studies controlled for the influence of systemic disease by ensuring the control groups had the same underlying pathology38,46 or used an appropriately matched control sample for studies of nondysvascular amputees.6 The work of Kanade et al.,38 although one of the methodologically strongest papers in terms of controlling for the influence of systemic pathology and other factors known to influence gait such as vestibular problems and deficits in visual acuity, did not present post-hoc analyses to statistically test for differences in walking velocity between the control and amputee groups. Given that the mean data of Kanade et al.38 differ by only one standard deviation, it is difficult to confidently suggest a difference exists between these groups in the absence of statistical analysis. Pinzur et al.46 observed that midfoot amputees (exact level of amputation was not clearly stated) with a history of diabetes and vascular disease walked at the same velocity as did control subjects with the same systemic pathology. In a study of nondiabetic amputees, with amputation levels ranging from MTP to Chopart, Dillon6 reported that five of seven amputee subjects walked at a velocity comparable to that of the healthy, nonamputee control group. On this basis, there is limited support for the contention that PFA, per se, may not result in a slower gait but that the underlying systemic pathology may dictate the individual’s walking speed.
18.104.22.168 Cadence, Stride Length, and Step Length
There is a low level of evidence to support an understanding of how changes in cadence, step, and/or stride length reduce walking velocity following PFA. Reductions in walking velocity were consistent with reductions in both cadence and stride length in diabetic toe and TMT amputees compared with a healthy control group.52 The only other study reporting cadence and stride length data6 found idiosyncratic reductions in walking velocity secondary to reductions in stride length, not cadence. Reductions in walking velocity reported by Tang et al.,48 in a mainly traumatic population, were inconsistent with the step length data reported, but without measures of cadence and stride length no further interpretation is possible. Wilson and Dillon54 found no differences in contralateral step length or stride length in a single traumatic amputee when comparisons were drawn between gait using a slipper or slipper and toe-off AFO condition.
22.214.171.124 Duration and Phasing of the Gait Cycle
There is insufficient evidence to be confident of the duration and phasing of the gait cycle in persons with PFA given that these results were reported by a single author as part of a case series investigation.6 As such, many of the observations are likely to be idiosyncratic. For individuals with amputation because of nondysvascular causes, the duration of the gait cycle (in seconds) was comparable to the 95% confidence interval of the normal cohort. In about half the amputee cohort, the proportion of stance was increased on the intact limb, with comparative reductions in the duration of swing phase. In two of the seven amputee subjects, the duration of single limb support (as a percentage of the gait cycle [% GC]) was reduced on the amputated limb, with comparable increases in the duration of double limb support following intact limb initial contact. In all subjects, contralateral heel contact was around 50% GC.
126.96.36.199 Effect of Amputation on Level on Temporospatial Parameters
There is insufficient evidence to evaluate the effect of amputation level on temporospatial characteristics of gait. A few investigations included relatively homogenous groups of subjects with the same amputation level43,48,51,52,55 or reported individual data that could be matched to explicit amputation levels,6,54 which was necessary to relate outcomes to amputation level. Methodological shortcomings limited the use of other works. Some investigations pooled data of subjects with disparate amputation levels9,38 or included data from individuals with contralateral lower limb involvement at either the partial foot or transtibial level into a larger sample of unilateral amputees without demonstrating the reasonableness of this approach.44,45,47 Other works did not adequately describe the amputation level46 or reported results using descriptions such as “no consistent differences” (p. 192).42
Of the investigations that clearly linked temporospatial data to amputation level, most reported exclusively on cohorts with TMT amputation,43,48,54,55 and as such, comparison to other amputation levels was limited. Several observational studies reported no difference in temporospatial parameters such as walking velocity, stride length, or cadence between toe and TMT52 or toe, MTP, and ray resection groups.51 This was similar to findings reported by Dillon,6 in which changes in temporospatial parameters were thought to be idiosyncratic given that at each amputation level some individuals were comparable to normal and others were not. In a study38 that pooled individuals with quite disparate amputation levels (toe, MTP, TMT, and ray resections) the variability in walking velocity was comparable to that in investigations with homogenous samples.43,48,52,55
188.8.131.52 Effectiveness of Interventions
There is inadequate evidence demonstrating that any one intervention is better at improving temporospatial aspects of gait. To better understand the effect of different prosthetic/orthotic interventions or footwear, several investigators have used various experiments.44,45,48,54 Other forms of investigation such as observational studies do not allow the effects of different interventions to be elucidated. Thus, only these experimental studies were considered to describe the influence of prosthetic/orthotic and footwear intervention on temporospatial parameters. Collectively, these experimental investigations evaluate populations of TMT amputees only. Two articles44,45 report identical subject demographics and results for walking velocity despite the focus of each paper being unique, suggesting that the same people participated in both studies.
Tang et al.48 evaluated three conditions of walking: bare feet, with a full-length standard shoe, and with a full-length shoe (same shoe as previous condition) and Plastazote toe filler with full-length, medium stiffness Springlite carbon fiber foot plate. Results indicated no significant difference in walking velocity or step length (for either the intact or amputated limbs) between conditions. These results are comparable to those of Wilson and Dillon,54 in which comparisons were drawn between gait with a slipper and a slipper plus toe-off orthosis condition.
The two studies by Mueller et al.44,45 experimentally evaluated six different footwear/prosthetic/orthotic combinations, including: (1) full-length standard shoe; (2) full-length shoe, total contact insert and AFO; (3) full-length shoe, total contact insert, and rigid rocker bottom sole (RRB); (4) full-length shoe, total contact insert, RRB, and AFO; (5) short shoe, total contact insert, and RRB; and (6) short shoe, total contact insert, AFO, and RRB. Walking velocity was found to be significantly faster in conditions 3, 5 and 6 than in condition 1 (full-length standard shoe). No significant differences were reported between conditions 2 through 6.
There were a range of methodological issues affecting these experimental investigations. Randomization of different experimental conditions was either not reported48 or conditions were pseudo-randomized to reduce costs,44,45 and as such, the likelihood that a series effect has influenced the results is strong. The experiment designed by Mueller et al.44,45 involved testing several interventions at once (i.e., AFO with insole with RRB), making it impossible to distinguish the influence of each intervention. Moreover, different interventions were concurrently applied to the intact limb, further limiting the conclusions drawn from this work. The experiment, overall, took some 6 months, allowing sufficient time for some new amputees (2 months after surgery in some cases) to improve their walking velocity irrespective of intervention, and likewise others may have experienced deterioration in walking velocity with time. Conditions 2 to 6 were evaluated a couple of days apart but more than 5 weeks after the first condition. Authors reported modifying the AFO condition at times during the experiment to improve the available ankle range at the subject-s request or dealing with subjects refusing to wear the device altogether.
Given the methodological limitations of these works, particularly that of Mueller et al.,44,45 it was difficult to draw any conclusions about the effects of prosthetic/orthotic or footwear interventions on the temporospatial aspects of gait.
There is insufficient evidence describing the underlying mechanical causes of changes in temporospatial parameters, including walking velocity, stride length, and/or cadence following PFA. Most of these investigations do not provide any interpretation of the temporospatial aspects of PFA gait.9,38,42–46,52,55 This may be because most investigations were focused on using temporospatial data, such as walking velocity, to aid interpretation of other data, such as energy expenditure, pressure measurement, or functional outcomes where velocity is a likely covariant.38,43–46,51,52,55 Other works suggest that reductions in walking velocity, for example, are the end result of reductions in other gait parameters without explaining the underlying mechanical cause6,51 or make inferences not supported by the results of the investigation.48 Hypotheses that reductions in lower limb muscle strength47 and an inability to progress over the forefoot54 have been well rationalized and may underlie reductions in step length, stride length, and walking velocity.
4.4 EXTERNAL FORCE AND CENTER OF PRESSURE EXCURSION
4.4.1 DESCRIPTION OF STUDIES
Eight publications6,31,37,42,51,52,54,55 were reviewed that included GRF and CoP data from PFAs. Two of these publications were from the same author and included the same subjects and were therefore considered as a single study.6,31 Three publications were abstracts from conference proceedings,51,52,54 and one was a thesis chapter.6 The article by Pinzur et al.8 was excluded from consideration because, by the author’s own admission, the analysis was qualitative, not quantitative, and therefore did not meet our inclusion criteria.
Most of the studies were descriptive in nature.6,31,42,51,52,55 One study was experimental, evaluating an intervention (Chopart prosthesis versus semirigid foot orthosis) in a somewhat systematic manner.37 Kelly et al.55 tested their subjects walking with regular shoes and a toe filler but did not provide an appropriate control or comparison, so the study was considered descriptive. Subjects ambulated barefoot in three of the studies,42,51,52 with an AFO in one study,54 and wore the devices used routinely in their day-to-day activities in the studies by Dillon.6,31
Sample size ranged from 7 to 26 PFAs, with the larger sample sizes consisting of smaller groups of subjects with different levels of amputation. The sample size of most studies was too small for parametric statistical analysis, and where statistical analyses were undertaken, none of the authors reported that a power analysis was conducted. Kelly et al.55 were the only authors to note that their findings may have lacked statistical power because of their small sample size.
Three of the studies included toe amputees,42,51,52 whereas one study included Lisfranc and Chopart amputees.6,31 Most of the studies included TMT amputees6,31,52 or were exclusively on TMT amputees.37,54,55
The cause of amputation was diabetes in three studies,51,52,55 surgical in one study of toe-to-thumb transfers,42 and traumatic in three studies.6,31,37,54
Control subjects were included in five of the studies.6,31,37,51,52,55 All controls consisted of healthy, able-bodied adults, which were considered appropriate for comparison with nondysvascular and nondiabetic amputees6,31,37 but not for comparison with diabetic amputees.51,55 The remaining studies compared data from the amputated to the intact limb42 or between conditions.54
Force data were collected from force plates (Kistler and AMTI) in all studies except the one by Kelly et al.,55 who reported force data from the F-Scan pedobarograph the only report to not normalize force by body weight. Vertical GRF data were reported in four studies,6,31,51,52,55 anterior- posterior GRF data in two studies of traumatic amputees,6,37 and CoP data in three studies of nondysvascular, nondiabetic amputees.6,31,42,54
184.108.40.206 Ground Reaction Force
There is moderate to low evidence describing the effect of PFA on the GRF.
In the studies reviewed, magnitude and timing of the vertical GRF peaks were the variables most often reported. Burnfield et al.52 reported that the magnitude of the first peak of the vertical GRF in barefoot, diabetic, unilateral TMT amputees was greater on the intact side than on the amputated side but no different in the toe amputees. Burnfield et al.52 also reported that it was greater in the intact limb of TMT amputees than in the intact limb of toe amputees. Comparison to healthy control data was inappropriate, despite attempts to account for substantial differences in walking speed. Burnfield et al.52 stated that the second peak of the vertical GRF was inversely correlated to length of the residuum and that this finding was significant, but they did not report residuum length.
Dillon6,31 reported both the magnitude and timing of the vertical GRF peaks and trough. With regard to the magnitude and timing of the first peak of the vertical GRF on the intact side, Dillon6 reported that it was greater when compared to that of controls only in the three unilateral, traumatic Lisfranc subjects wearing their usual devices (slipper socket in two subjects and shoe with toe filler in one). It has been suggested that increased intact limb loading is related to a lack of sufficient support by the forefoot of the trailing amputated limb, as evidenced by a rapid drop-off in the second peak of the vertical GRF on the amputated side.56 Although Dillon6 does not comment on this, his data for the amputated side of all but the Chopart and MTP amputees demonstrate this drop-off (premature conclusion of the second vertical GRF peak). This suggests that where the effective foot length/lever was present (either because the metatarsal heads remain intact or the device is able to restore it), the drop-off does not occur; however, the impact of drop-off on intact limb loading appears to be inconsistent in the PFAs evaluated by Dillon.6
With regard to the trough in the vertical GRF, Dillon6 reported that it was delayed bilaterally in the bilateral Lisfranc and Chopart amputees, as well as the unilateral Chopart amputee and on the intact limb of the TMT amputee.
Along with Dillon,6 Boyd et al.51 and Kelly et al.55 reported the timing and magnitude of the second peak of the vertical GRF. Boyd et al.51 reported no differences among groups (barefoot, diabetic, unilateral toe, metatarsal, and ray amputees) in the peak and rise-rate during mid- to terminal stance of the second peak of the vertical GRF on the amputated side. Dillon6 reported that the second peak was lower than that for controls only for two of the traumatic Lisfranc amputees wearing their usual devices (slipper socket for one and shoe with toe filler for the other) but appears to finish earlier in all amputees except those with Chopart and MTP amputation (perhaps because the effective foot length/foot lever is adequate). In contrast, Kelly et al.55 did not find any difference between limbs in the timing of the second peak in their diabetic, unilateral TMT subjects wearing shoes with a toe filler. In the study by Kelly et al.,55 comparisons with healthy controls did not account for the effects of the systemic disease or substantial differences in walking speed.
With regard to anterior-posterior GRF data, Dillon6 reported that there were few changes, with the exception of a premature second (posterior horizontal force) peak on the amputated side in the traumatic, unilateral TMT and Lisfranc subjects walking with their own devices (slipper socket or shoe with toe filler).
220.127.116.11 Center Of Pressure Excursion
The level of evidence supporting our understanding of CoP excursion and the influence of prosthetic/orthotic intervention is moderate-low.
It is somewhat difficult to compare results of these investigations because of differences in population between studies: hallux amputees,42 midfoot amputees6,31 or a single TMT amputee.54 However, all studies suggest that velocity of the CoP is altered in PFAs.6,31,42,54 Mann et al.42 indicated that progression of the CoP slowed once the CoP was located beneath the metatarsal heads in the hallux amputees, whereas Dillon6,31 suggested that forward progression of the CoP slowed substantially during single limb support in that heterogenous group of midfoot amputees. Similar results were reported by Wilson and Dillon54 in a single subject with TMT caused by trauma: progression of the CoP slowed during midstance when a slipper prosthesis was worn but not when a Blue Rocker toe-off AFO (made of a carbon fiber reinforced thermoset and incorporating an anterior shell) was worn in conjunction with the slipper prosthesis.
Wilson and Dillon54 reported that advancement of the CoP beyond the distal end of the residuum did not occur until after the second GRF peak had occurred when the slipper socket alone was worn but that when the Blue Rocker AFO was added, the CoP was able to progress beyond the distal end of the residuum prior to the second vertical GRF peak. Dillon6 also reported that although no significant reductions in the total excursion of the CoP were observed, substantial progression of the CoP past the distal end of the residuum did not occur until contralateral heel contact in the TMT and Lisfranc amputees. These studies suggest that progression of the CoP is coupled to whether the limb is in single- or double-limb support: in the absence of a suitable device, the CoP can advance beyond the distal end of the residuum only when the amputated side is able to share the load with the contralateral limb (i.e., double limb support). It does not advance beyond the distal end of the residuum during single limb support. The CoP was able to progress beyond the distal end of the residuum in single-limb support when the clamshell prosthesis or Blue Rocker AFO were worn. Both of these devices have force systems suited to controlling forward progression of the tibia and are sufficiently stiff to support body weight.54
Changes in CoP excursion and timing of the second peak of the vertical GRF appear to contribute to altered moments and power profiles in PFAs.6,31 It may also play a role in increasing contralateral step length and walking speed, although this was not achieved with the Blue Rocker AFO, with the authors attributing this to decreased ankle dorsiflexion and the fact that walking speed was comparable to normal regardless of device worn.54 With regard to future research, any analysis of CoP and GRF should take into account the timing of contralateral initial contact.
The implication of force and CoP data for prosthetic and orthotic management is difficult to elucidate given the limited data described above. Based on their investigation of CoP excursion and timing, Dillon and Barker31 suggested that the assumption that a device can restore foot length or the toe lever to the extent that an amputee can actually load it is not supported by the force data and postulated that for this to occur, there must be adequate coupling between the device and residual limb, sufficient contact area for forces to be transferred to the residual limb comfortably, and a stiff enough toe lever to support the amputee’s body weight in late stance. It was their contention that the clamshell prosthesis and Blue Rocker AFO were the only devices that demonstrated potential in this regard and that the other devices evaluated in these studies were insufficiently stiff in the forefoot and may not have transmitted forces comfortably to the residuum.6,31,54 However, it remains unclear whether PFAs have sufficient strength to generate appropriate ankle moments.6
It was interesting to note that despite a “better gait” with the clamshell prosthesis over level ground walking, Hirsch et al.37 reported that the device was not used long term in their traumatic TMT amputees. Although this study did not contribute substantially to our understanding of the biomechanics of PFA gait, it raised some interesting issues regarding the need to consider the whole person and the use of devices in a broader context than level walking. Thus, future research should consider assessing function with devices more broadly. Hirsch et al. contended that function is strongly related to the number of stump problems. This is yet another confounding variable that along with etiology, level of amputation, and time since amputation should be considered when selecting subjects. These variables tend to be insufficiently described and controlled in the current PFA gait literature.
4.5.1 DESCRIPTION OF STUDIES
Nine publications5,6,27,36,40,42,48,51,54 were identified that referred to kinematic data from PFAs. Two of them did not present any results and were therefore excluded from further consideration regarding kinematics.36,42 An additional study was excluded from further consideration because the protocols, outcome measure, data analysis, and results were inadequately described and/or insufficiently reported.40 Thus, six publications were reviewed that included kinematic data from PFAs, two of which were abstracts from conference proceedings,51,54 and one a thesis chapter.6
The majority of the studies were descriptive in nature.5,6,27,51 Two studies were experimental, evaluating an intervention (standardized prostheses/orthoses) in a systematic manner54 and with a reasonable control or comparison.48 Mueller et al.5 also provided their subjects with standardized devices (a shoe with toe filler) but did not provide any adequate control or comparison, so the study was considered descriptive. Subjects ambulated barefoot in two of the studies,27,51 whereas subjects wore a variety of their own devices in the study by Dillon.6
Sample size ranged from 1 to 26 PFAs, with the larger sample sizes consisting of smaller groups of subjects with different levels of amputation. One study51 included toe amputees, and one study included some Lisfranc and Chopart amputees.6 The majority of studies included TMT amputees6 or were exclusively on TMT amputees.5,27,48,54
The cause of amputation was diabetes in three studies5,27,51 and exclusively nondysvascular in three studies.6,48,54
Control subjects were included in four of the studies.5,6,48,51 All controls were healthy, able-bodied adults, which was appropriate for comparison with nondysvascular and nondiabetic amputees6,48 but not for comparison with diabetic amputees5,51 because it did not allow the authors to account for the effect of the systemic illness on gait. The remaining studies compared data from the amputated to the intact limb27 or between conditions.54
Kinematic analyses were limited to the sagittal plane. All six studies assessed sagittal plane ankle kinematics,5,6,27,48,51,54 and two included sagittal plane knee and hip kinematics.5,6
4.5.2 METHODOLOGICAL QUALITY
It is worthwhile to note that protocol, instrumentation, and conditions tested varied widely across the studies. With one exception,27 most used commercially available motion analysis instrumentation.5,6,48,51,54 Marker placement is important in any discussion of kinematic data because it defines to a large degree what is being measured. Because most marker models are applied to normal anatomical landmarks, it is especially important to document where markers are placed when shoes and devices obscure those landmarks. Movement between the residual limb and shoe or device further complicates the measurement of joint motion and should be considered along with marker placement. Although two studies explicitly described marker placement,5,27 Mueller et al.5 did not account for the absence of anatomical landmarks and presence of the shoe. Because Tang et al.,48 Boyd et al.,51 and Wilson and Dillon54 used commercially available systems, it is assumed that they followed the standard marker placement procedures required of those systems. Unfortunately, this assumption does not completely inform our understanding of marker placement in the presence of devices such as in the studies of Tang et al. and Wilson and Dillon, in which placement of foot markers is likely to have changed between the conditions (barefoot and shoe only or shoe with prosthesis in the study by Tang et al. and slipper prosthesis and slipper prosthesis with Blue Rocker toe-off AFO in the study by Wilson and Dillon).
Mueller et al.5 describe the use of three markers on each of the thigh, leg and foot segments but only one marker on the pelvis superior to the greater trochanter and state that they undertook measurements of the thigh, leg, and foot in three dimensions (p. 201). It is difficult to understand how three-dimensional hip joint angles relative to the pelvis might have been calculated using this marker set because the pelvic segment appears inadequately defined for this task. If hip joint angles were calculated relative to the global coordinate system, or if the data were calculated in two dimensions rather than three, then the data probably would be different from that reported by other authors using commercially available systems in which hip motion is usually relative to the pelvis. Furthermore, for a two-dimensional analysis, the position of the pelvic marker would strongly influence hip angle calculations. Unfortunately, there is insufficient detail provided in the method section to be completely clear regarding the modelling used.
The assumptions inherent in the biomechanical models used by commercially available systems are not applicable to PFA, although this might not affect kinematic results substantially if marker placement were carefully considered. Only two studies used customized models: Garbalosa et al.27 assessed ankle motion using a model of the foot and ankle as two rigid segments with a 6° of freedom joint, while Dillon6 applied custom anthropometric models that accounted for the altered anatomical segments, devices, and shoes, although this is of greater relevance for kinetic measurements, rather than kinematic. Two studies assessed barefoot walking only,27,51 one assessed walking with a shoe with toe filler,5 and one study assessed walking in devices only,6 limiting interpretation of their results. Two studies48,54 compared walking across conditions: Wilson and Dillon54 assessed a slipper prosthesis with and without a Blue Rocker toe-off AFO, whereas Tang et al.48 assessed barefoot, shoe-only, and shoe with prosthesis walking, allowing the contribution to changes in ankle kinematics of the shoe alone as well as the device and shoe to be discerned, assuming marker placement was appropriate and comparable.
18.104.22.168 Passive Range of Movement
Two studies reported or commented on passive ankle range of motion (ROM).6,27 Garbalosa et al. ’s27 data from ten TMT amputees showed no difference in passive ankle dorsiflexion ROM between amputated and intact feet. Dillon6 reported on both passive dorsiflexion and plantarflexion ROM, commenting that, within eight amputees of various levels, dorsiflexion range was no different from that of able-bodied control subjects and that the bilateral Lisfranc, bilateral Chopart and unilateral Chopart amputees had limited plantarflexion. It would appear that none of the subjects included in these studies had limitations in dorsiflexion range due to equinus contracture, although this is often noted as an issue with regards to PFA, especially at the more proximal levels.
22.214.171.124 Kinematic Data
Two studies reported data for joints other than the ankle.5,6 A third group of authors48 indicated in their discussion that hip and knee kinematic data were analyzed as part of their study but were not reported because the changes observed were not as obvious as at the ankle.
126.96.36.199 Hip and Knee Joint Kinematics
There is an insufficient level of evidence describing the kinematics of the hip and knee joints for persons with PFA.
According to Mueller et al.,5 knee and hip ROM during stance were significantly less than that of able-bodied controls for 15 TMTs caused by diabetes walking with regular shoes and a toe filler. However, comparison to a healthy control group does not allow us to discern whether these differences were due to the amputation or the systemic disease; to walking speed, which was slower in the diabetic amputees; or to the device used because there were no other comparison conditions. Furthermore, overall hip ROM, and in particular peak hip flexion, for the control subjects appears rather small compared with what might be expected for able-bodied adults walking at the same speed, raising questions about the marker placement and modelling used. In Dillon’s PhD thesis6 the control data are comparable to that expected of able-bodied people walking at similar speeds. The bilateral sagittal plane kinematic data presented for the hip and knee appear to support the author’s assertion that there were few differences from the matched normals when the traumatic amputees walked with their customary devices.6 Unfortunately, despite these two reports,5,6 it remains unknown whether hip and knee kinematics vary across levels of amputation or what the effects of various assistive devices might be because these conditions have not been investigated.
188.8.131.52 Ankle Joint Kinematics
There is strong evidence that PFA affects ankle kinematics; however, only moderate to low evidence supports a more detailed understanding. Synthesis of the ankle kinematic results is difficult given that variables are inconsistently reported. Despite these inconsistencies, the timing and magnitude of peak ankle dorsiflexion and plantarflexion appear to be of predominant interest with regard to ankle kinematics.
Garbalosa et al.27 compared ankle ROM during gait between amputated and intact limbs and reported that there was significantly greater ankle dorsiflexion ROM in the intact limbs. Although Boyd et al.51 also evaluated diabetic PFAs during barefoot walking, it is difficult to compare the findings to those of Garbalosa et al. because neither reported walking speed or step lengths, factors that influence joint ROM. Boyd et al. did not provide data for the intact side, and it is unclear from the description if the levels of amputation are comparable. Although Boyd et al. allude to the influence of speed in their description of the statistical analyses used, it was unclear which variables they considered to be affected by speed and therefore analyzed using an ANCOVA (analysis of covariance), rather than ANOVA (analysis of variance).
Tang et al.48 reported ankle ROM during stance phase for both the intact and amputated limbs of TMTs and control subjects across three walking conditions: barefoot, shoe only, and shoe with prosthesis (final condition evaluated on TMT group only). They reported that both the shoe-only and shoe-with-prosthesis conditions allowed significantly more ankle dorsiflexion during stance than walking barefoot, with no difference between the shoe and shoe with prosthesis. Although overall ankle ROM was comparable to normal and more symmetrical in the two shod conditions, peak stance phase dorsiflexion on the amputated limb was greater than that of controls, whereas plantarflexion range remained substantially reduced in the shod conditions. Because marker placement was not explicitly described and it is obvious that markers were moved between testing of the different conditions (most likely between testing of barefoot and both shoe conditions but probably not between the two shod conditions), it is possible that alterations in marker placement contributed to this finding.
Motion of the ankle occurred through a similar range in the subjects in Dillon’s6 study who wore below-the-ankle devices: that is, total range between plantarflexion following initial contact and peak stance phase dorsiflexion is similar, as is the range between peak stance phase dorsiflexion and plantarflexion following toe-off. The distinction between the two studies is that the ankle kinematics reported by Tang et al.48 are biased toward dorsiflexion compared with what one would expect from able-bodied adults walking at a similar speed. This reinforces our concerns regarding marker placement.
With regard to ankle dorsiflexion, Boyd et al.51 reported that there was no significant difference in the magnitude of dorsiflexion but that time of peak dorsiflexion was significantly delayed relative to healthy, able-bodied controls for the amputated limb in all levels of amputation studied (toe, metatarsal, and ray amputations). This finding is consistent with that reported by Dillon6 for traumatic PFAs of various levels (MTP, TMT and Lisfranc) ambulating with their usual devices (custom orthosis, toe filler, slipper socket, stuffed shoe); Dillon reported that, regardless of the device worn, peak dorsiflexion was substantially delayed on the amputated limb of all subjects except the bilateral MTP amputee. It is difficult to assess whether the data reported by Tang et al.48 concur or not because the authors do not specifically comment on the magnitude and timing of peak ankle dorsiflexion, nor do they provide sufficient information, specifically variability within each condition, for the reader to interpret whether there were differences.
Although the study by Boyd et al.51 provided an inappropriate comparison in the younger, healthy controls, the consistency of their finding regarding delayed peak dorsiflexion with the finding of Dillon6 (who included a more appropriately matched control group) suggests that delayed peak dorsiflexion on the amputated side may be a characteristic of PFAs, regardless of etiology, and to the extent represented by the two study populations, regardless of level of amputation and walking condition (barefoot or device). With respect to the study by Boyd et al., it was difficult to assess the differences between levels of amputation because the mean (and it is assumed this is not an “adjusted” mean such as the authors report for moment and force data) is reported in the absence of any indication of variability. Because the studies of both Boyd et al. and Dillon were small and heterogenous, this finding needs to be additionally substantiated.
It was interesting that when walking barefoot, the diabetic TMT amputees in the study by Garbalosa et al.27 reportedly had less dorsiflexion on the amputated side than on their intact side. In this regard, Dillon’s6 data did not concur with the results of Garbalosa et al.
Peak ankle dorsiflexion was reduced in subjects who wore devices that extended above the ankle: Wilson and Dillon54 reported that peak ankle dorsiflexion decreased with the addition of a Blue Rocker AFO in their single subject with traumatic TMT. Dillon6 also reported a reduction in ankle motion in the Chopart subjects who wore clamshell prostheses. These results are not surprising because both devices are designed to control motion of the tibia.
The traumatic TMT amputees in the study by Tang et al.48 demonstrated limited plantarflexion in terminal stance compared with control subjects regardless of condition (barefoot, shoes only, or shoes with prosthesis). Similarly, Dillon6 reported that there were substantial reductions in the plantarflexion angle at toe-off for all amputated limbs in all subjects. The results of Mueller et al.5 with regard to plantarflexion also concur with those of Tang et al. and Dillon. Mueller et al. reported that there was a significant reduction in the peak plantarflexion achieved at toe-off in a group of 15 TMTs caused by diabetes evaluated while walking with a shoe and toe filler. However, similar to observations made with regard to hip data, the ankle plantarflexion ROM during stance for the control subjects in the study by Mueller et al. appears to be less than would be expected for able-bodied adults walking at the same speed.
Dillon6 was the only author to comment on swing phase kinematics, suggesting that there were two distinct kinematic profiles during swing phase on the amputated limb of the unilateral amputees that appeared to be related to the maximum plantarflexion angle and initial contact angle. Although Tang et al.48 did not report swing phase kinematics directly, the data shown graphically appears similar to that reported by Dillon.
Both Mueller et al.5 and Dillon6 pointed out that deviations in the gait of PFAs occur mostly in late stance and that this is the most demanding period of the gait cycle; a point supported by the results of Garbalosa et al.27 and Tang et al.48 These results support the notion that forefoot loading is compromised. Since his was a thesis, Dillon provided the most substantive discussion regarding interpretation of his results. In the other studies,5,27,48,51 interpretation of the results was limited.
Overall, the results from this handful of studies suggest that bias of the ankle into dorsiflexion appears to occur throughout stance and swing to some extent, irrespective of amputation level, etiology of amputation, or device, although none of these variables have been fully investigated. Dillon6 proposed that the observed dorsiflexion bias was indicative of substantial reductions in the resistance to ankle dorsiflexion from loading response through to heel-off by the posterior calf musculature. However, he was unsure if the corresponding decrease in CoP excursion was an attempt to protect the distal residuum or to control the position of the trunk. With regard to the effect of devices, Dillon commented that for the Chopart amputees in his study, the kinematic patterns at the ankle were dominated by the clamshell prostheses, which severely limited dynamic ankle range but also appeared to more adequately moderate tibial progression during midstance, suggesting it was likely a benefit of the counterforce generated by the anterior wall of the socket. Similarly, Wilson and Dillon54 commented that by controlling tibial advancement through the anterior shell, the Blue Rocker AFO allowed for substantial load to be applied distal to the residuum but also reduced ankle dorsiflexion in their single TMT amputee, which was perhaps responsible for the unchanged temporospatial variables. The data of Tang et al.48 suggest that we cannot ignore the contributions of the shoe alone to altering gait in TMT amputees, especially when compared with devices that do not extend outside the shoe. Unfortunately, this is the only study that allowed such an insight. Future studies should include a shoe-only condition for comparison of PFA gait between walking barefoot and walking with devices. Furthermore, any assessment of ankle motion requires that marker model and the effect of relative movement between the residual limb and device be carefully considered.
4.6.1 DESCRIPTION OF STUDIES
Five publications reported on kinetic aspects of PFA gait: three articles,5,32,48 one abstract,51 and one thesis.6 The article by Dillon32 reported on aspects of a larger investigation forming part of a PhD thesis.6 These studies were predominantly observational in design,5,6,32,51 with one experimental investigation.48
Only sagittal plane kinetic data was reported in the literature. Investigations have tended to focus on the amputated limb, reporting kinetic data for the ankle48,51 or entire lower limb.5,6,32 Only one investigation reported kinetic data for the intact limb.6
Two of these investigations5,48 reported solely on populations with TMT amputation. Limited kinetic data were available for other levels of PFA, including small cohorts of individuals with toe, MTP, or ray resection amputations51 and a case series of one MTP, one TMT, four Lisfranc, and two Chopart amputees.6,32 Individuals in two investigations5,51 had amputation because of diabetes and vascular insufficiency. Other cohorts sustained amputation as a result of trauma or nondiabetic related illness.6,32,48
4.6.2 METHODOLOGICAL QUALITY
There were a number of methodological issues affecting these investigations. The following synopsis is designed to overview these issues, and where these are believed to have influenced specific kinetic data, additional discussion has been provided.
Control subjects were inadequately matched in two of these investigations,5,51 which made it difficult to account for the influence of systemic diseases, such as diabetes, on gait. The other investigations6,32,48 reported kinetic data for individuals with amputation secondary to trauma or nonsystemic diseases with an appropriately matched control group.
The heterogenous population of PFA in the study by Mueller et al.5 was affected by several potentially confounding variables, including the time since amputation (which was very recent for some subjects), the presence of severe peripheral neuropathy, and the inclusion of bilateral amputees alongside unilateral amputees.
Of greater concern to the kinetic data, particularly at the hip and knee joints, were issues regarding how the joint centers were located and how joint angles were calculated in the study by Mueller et al.5 It appears that while the thigh, leg, and foot segments were defined in 3D using a suitable marker triad, no such marker set was used to define the pelvis. A single marker was located on the pelvis superior to the greater trochanter, and hip joint angles were probably calculated in 2D, but the method section provides inadequate detail to be confident of this. Estimates of the joint centers were established using a tape measure, rather than from segment anthropometric data and known anatomical landmarks defined by marker placement. The kinematic and kinetic data reported for the control group is unusual compared with other investigations of normal adult gait, as evidenced by the tabular data for the control sample and the graphical data of a single “representative” control subject. In general, the magnitude of the peak joint moments and powers was small, as were the ranges of motion observed at the hip. The representative normal data from a single control subject has no knee flexion moment during the middle portion of stance and a significant knee extension moment peak during terminal stance. There is no knee power absorption during loading response as the knee flexed during stance phase. The hip joint goes through a range of about 20° for the normal cohort, and power generation across the ankle was about half that of other studies on normal adults. Given these modelling concerns and the normative data presented, it is difficult to confidently interpret the amputee data.
The investigation by Dillon6 used a 2D modelling approach. Joint angles were given as the difference in adjacent segment angles using an arc-tan function, with each segment initially defined in 3D by a typical marker triad. Kinetic estimates of gait were calculated using custom-linked segment models that accounted for the unique anthropometry of the amputated foot, proximal limb segments, and any prosthesis and footwear.6 For devices that did not eliminate ankle motion, the physical characteristics of the residuum, prosthesis/orthoses, and footwear were combined and described by a single set of mass and inertial parameters with respect to the ankle.6 To allow the clamshell prosthesis to be incorporated into the linked-segment model, the body segment parameters for the residuum, leg, prosthesis, and shoe were lumped such that they could be described by a single set of inertial characteristics with respect to the knee joint.6 As such, there was no ankle joint within the model. To draw comparisons of ankle moment data to other investigations and better understand the influence of the CoP and GRF, ankle joint moments were separately calculated for individuals using clamshell prostheses using a conventional linked-segment model. The conventional linked-segment model used anthropometric data from the intact limb. Given that the ankle joint moment is dominated by the magnitude of the vertical GRF and its lever-arm from the ankle, it was sufficiently robust to errors in the anthropometric input data.6 These novel modelling approaches influenced swing phase kinetic data, particularly during terminal swing, where the peak moments and powers were significantly increased compared with more conventional linked-segment models.6 The kinematic and kinetic data obtained from these modelling approaches were otherwise comparable to those of other investigations of normal gait.57,58
184.108.40.206 Ankle Kinetics
Although there is strong evidence that ankle kinetics are altered following PFA, the level of evidence supporting a more detailed understanding is moderate to low, depending on the number and quality of investigations reporting on different levels of amputation and interventions.
The ankle joint moment data for the amputated limb was relatively consistent across investigations.5,6,32,48,51 Both the level of amputation and type of prosthetic or orthotic fitting had some bearing on the ankle joint moment. In cases in which amputation did not change the overall foot length, as was the case in the toe and ray resection groups studied by Boyd et al.,51 the ankle plantarflexor moment was comparable to normal. However, once amputation disarticulated the MTP joint, the ankle plantarflexor moment was reduced substantially compared to normal51 or was observed at the lower 95% confidence interval of the normal population.6 In the TMT5,6,48 or Lisfranc amputees6 the peak plantarflexor moment was about two-thirds that of normal, with considerable variability between individuals within each investigation.5,6,48 In a small cohort of Chopart amputees,6 the peak plantarflexor moment was comparable to normal, highlighting the influence of prosthetic fitting.
Although these investigations describe consistent reductions in the ankle plantarflexor moment for all levels of amputation except in the Chopart, the peak amplitudes varied dramatically among investigations as a result of differences in walking velocity and normalization to control for the influence of changes in speed.51 The peak plantarflexor moment reported for control subjects was relatively small (1.35 Nm/kg) in some investigations5,48 and larger (1.72 Nm/kg) in another.6 The faster walking velocity reported by Dillon6 compared with the results of other investigations5,48 likely explains the increased amplitude of these moment data given the influence of velocity on the gain of lower limb joint moments.58–61 Boyd et al.51 reported “adjusted means” for parameters that covaried with velocity, which made comparison to other investigations difficult, particularly since velocity was expressed as a proportion of a laboratory normal database.51 Moreover, kinetic data were reported in unspecified units such as “Nm/kgLL,” which presumably indicates that these data were normalized by limb length or leg length in addition to body mass.51
In addition to considering the plantarflexion moment peak, the ankle moment pattern is equally as important in terms of understanding the underlying mechanical deficit. The gradient of the plantarflexion moment curve was consistently reduced compared to normal in the TMT5,6,48 and Lisfranc6 amputees such that by midstance the ankle joint moment was about half that of the control group.5,6,48 In other words, the moment curve was considerably flatter. However, in the Chopart amputees the joint moment profiles were comparable to normal.6
These investigations reported ankle moment data for individuals using toe fillers,5,6 ethyl vinyl acetate (EVA) insoles with48 or without a carbon fiber foot plate,6 slipper sockets incorporating Plastazote toe fillers or polypropylene foot plates,6,31 or clamshell prostheses with a toe lever of thick carbon fiber or the distal portion of a prosthetic foot.6 Irrespective of the investigation, the below-the-ankle devices were unable to restore the normal pattern or amplitude of the ankle moment, whereas the clamshell devices were. This interpretation is contrary to that expressed in the one experimental investigation48 that compared kinetic data in barefoot walking, shoes, and prosthesis conditions. Use of emotive language by the authors in the discussion section suggests that the authors may have had a personal investment in the outcome, introducing the potential for bias. For example, authors stated that “After wearing the prosthesis (insole), ankle moment improved to 62% as compared to the normal group”(p. 127).48 Although this statement is, in the strictest sense, true, closer inspection of the results indicates that significant differences in the peak ankle plantarflexion moment existed between the barefoot and shoe-only conditions. No significant differences were observed between the shoe-only and insole-with-shoe conditions. Thus, an alternative interpretation may be that the insole did not improve the peak ankle plantarflexor moment beyond that achieved by provision of a shoe only.
The majority of these investigations5,48,51 provided limited interpretation of their results with few insights into the underlying cause of the abnormalities observed5,51 or presented discussion that appears unrelated to the results.48 Dillon6 suggested that reductions in the ankle plantarflexion moment observed across the TMT and Lisfranc amputees reflect the limited distal progression of the CoP when the largest vertical GRFs occurred. In the TMT and Lisfranc amputees, the CoP remained well proximal to the distal end of the residuum when the peak vertical GRFs were observed. The CoP did not progress distal to the end of the residuum until double support, when the superincumbent mass could be transferred to the contralateral limb. In this way, the longest lever arm(s) of the GRF were commensurate with the rapidly diminishing magnitude of the vertical GRF force—thus the reduced plantarflexion moment(s).
It has been suggested6,32 that the TMT and Lisfranc amputees may have adopted this gait pattern to their advantage for any one of a number of reasons, including
- to spare the sensitive distal residuum from the extreme forces typically observed during terminal stance;
- to reduce the requirement of the triceps surae musculature or prosthesis by moderating the external moment;
- to minimize the heel slipping out of the device and sheer on the distal end; or
- the inadequacy of prosthetic/orthotic design.
To expand on the last point dealing with the inadequacy of prosthetic/orthotic design, the toe filler, foot orthosis, or slipper socket fitted to the TMT and Lisfranc amputees in Dillon’s study6 were unable to restore the effective foot length. Amputees were unable to use the prosthetic forefoot for substantial weight bearing.6,31 There may be a number of possible explanations. The toe fillers and foot orthoses did not incorporate a socket that could rigidly couple the residuum and prosthesis. The slipper sockets did incorporate a true socket but may not have been able to comfortably transmit the forces caused by loading the toe lever to the small surface area of the Lisfranc residuum.6 The toe levers in all these devices were likely to have been too flexible, given that they were manufactured from either Plastazote or polypropylene of little more than a few millimeters in thickness.6
The clamshell devices fitted to the Chopart amputees were able to effectively restore the lost foot length, thus enabling the amputees to adopt a gait pattern whereby the peak vertical GRF occurred commensurate with the CoP being located well beyond the distal residuum.6 The clamshell prostheses seemed well suited to the task because they incorporated a rigid toe lever, of either thick carbon fiber or the distal half of a prosthetic foot bonded to the socket and a substantial socket capable of comfortably managing the external torques caused by loading the prosthetic forefoot.6
Significant reductions in the peak ankle power generation were consistently reported across investigations.5,6,32,48 In a single MTP amputee, peak power generation was at the lower 95% confidence interval of the normal sample.6,32 Once amputation compromised the metatarsal heads, power generation was virtually negligible (0.5–0.9 W/kg) irrespective of amputation level.5,6,32,48 Reductions in power generation in the TMT and Lisfranc amputees were attributable to the diminished ankle moment5,6,32 coupled with reduction in the joint angular velocity.6,32 In the Chopart amputees, reductions in power generation reflect the relative elimination of ankle motion using a clamshell prosthesis (noting the influence of the force-deflection characteristics of the foot creating pseudoankle motion) and not a reduction in the ankle joint moment.6,32
Several investigators reported on comparatively minor abnormalities with the timing51 or magnitude48 of the ankle power absorption peak. The timing of the ankle power absorption peak was significantly delayed in cohorts of toe and MTP amputees51 compared with normal. Similar, but not statistically significant, observations were noted in a case series of TMT, Lisfranc and Chopart amputees,6 and in figures of an isolated TMT amputee.5 The magnitude of the power absorption peak was significantly larger in an experimental investigation48 following provision of a prosthesis and shoe compared with a barefoot condition. These observations were thought to be reflective of the delayed and rapidly increasing ankle dorsiflexion angle.6 The rapid and exaggerated ankle dorsiflexion angle common with shod walking is thought to be reflective of a measurement error associated with the marker triads used. As such, it is unlikely that this is a true reflection of the mechanical work adaptation to PFA.
Of some concern were the varied amplitudes of the ankle power data observed in the control subjects among investigations,5,6,48 which makes it difficult to feel confident about the absolute magnitude of the peak power generation in the amputee cohorts. Differences in the peak power generation for control groups varied between averages of 1.75 W/kg5 and 3.66 W/kg.6,32 It is difficult to appreciate from where these differences stem, given that all investigations used healthy control subjects who were similar in virtually all respects and gait analysis equipment that seemed comparable.
Interpretation of the ankle power data by authors was limited, with few insights into the underlying cause of the movement patterns observed.5,6,48 Tang et al.48 suggested that the provision of a prosthesis (insole with carbon fiber foot plate) significantly improved power generation despite that the descriptive statistics (particularly the variability between experimental conditions) suggested otherwise. Mueller et al.5 offered the observation that reductions in peak power generation were most likely caused by the shortened foot and plantarflexor lever arm. Dillon6 stated that reductions in power generation across the ankle reflected the limited work by the ankle plantarflexors to accelerate the limb segment into swing phase, contribute to the forward kinetic energy of the trunk, and maintain the vertical height of the center of mass of the upper body.
In terms of prosthetic fitting, devices used by subjects with TMT and Lisfranc amputation (toe fillers and slipper sockets), which were designed to allow relatively unrestricted ankle motion, did not improve power generation compared with the clamshell prostheses used by subjects with Chopart amputation in whom ankle motion was eliminated.6,32 As a result of this observation, it has been suggested6,32 that the ankle contributes little to the work required to walk at the TMT, Lisfranc, and Chopart levels, and given the likelihood of complications such as ulceration and skin breakdown, it seems reasonable to advocate routine use of a clamshell device where the risks of these complications might be minimized.
220.127.116.11 Hip and Knee Kinetics
With respect to the hip and knee joint kinetics, the level of evidence is generally low to insufficient.
The MTP and Chopart amputees exhibited relatively normal knee joint moments on the amputated limb, both in terms of the basic moment pattern and the magnitude and timing of the moment peaks.6 By contrast, the TMT amputee and Lisfranc amputees typically exhibited a relatively neutral knee moment from about midstance until the end of the gait cycle.6 Mueller et al.5 reported peak knee extension moment data for a cohort of TMT amputees, but it is not clear from the graphical data of “a TMT subject” (p. 204) which peak was evaluated because an extension moment is maintained throughout stance phase following loading response.
Dillon6 provides little insight explaining why the knee flexion moment was negligible from midstance in individuals with TMT and Lisfranc amputation, which was surprising given the consistency of this observation.
Based on a single investigation,6 there were some notable abnormalities in the knee power data, particularly during loading response for Chopart amputees using clamshell prostheses and for the TMT and Lisfranc amputees following midstance. In the MTP amputee and on the intact limb of the amputee subjects, a normal pattern of work exchange was observed across the amputated limbs with the magnitude and timing of the power peaks comparable to normal.6
In the Chopart amputees,6 the magnitude of power absorption across the knee on the amputated limb following initial contact was reduced, consistent in some individuals with the reduced magnitude and delayed timing of stance phase knee flexion. This kinetic pattern is similar to that observed in other groups of lower limb amputees whereby the foot and leg segments move synchronously between initial contact and foot flat because of the elimination of ankle motion by the prosthesis.6
Consistent with the knee moment profiles previously described, the normal pattern of power absorption and generation was relatively negligible on the amputated limbs of the TMT and Lisfranc amputees following midstance.6 Given that the knee joint moment was virtually zero after midstance, there was little need for the typical muscle response usually associated with maintaining the position of the trunk as it progressed anteriorly relative to the base of support.6
Observational studies by Mueller et al.5 and Dillon6,32 reported on aspects of hip kinetics. Both graphical and numerical data were available to describe the hip joint kinetics,6,32 and isolated peaks were evaluated by Mueller et al.,5 with graphical representation of the moment and power patterns presented for only a “representative” TMT amputee. Unlike kinetic data for other lower limb joints, observations for the hip joint were more varied between investigations5,6 and, to some degree, were at odds with each other.
The hip joint moments observed on the amputated limbs of the amputees were relatively normal in terms of the moment profile, magnitude, and timing of the moment peaks.6 However, considerable variability was evident in this case series.6 In a small number of subjects, a large hip extension moment was observed during loading response on the amputated limbs of between 1 and 1.5 Nm/kg.6 In these same amputee subjects, an extension moment was maintained on the amputated side hip beyond midstance, after which the magnitude of the extension moment was comparable to the upper 95% confidence interval of the control group.6 On the intact limb, the hip extension moment during loading response was also unusually large with a peak around 1 Nm/kg in a number of subjects.
Mueller et al.5 observed no difference in the magnitude of the hip extension moment during loading response in a cohort of TMT amputees, but considerable individual variability was also present. If the mean and standard deviations were considered, some of the amputee subjects would have exhibited hip extension moments during loading response of similar magnitude to those reported in the case series investigation by Dillon.6 The onset of the hip flexion moment occurred prematurely in the TMT amputees (47% ± 18% stance) compared with the control group (61% ± 13% stance).5
Mueller et al.5 hypothesized that because of the shortened foot, amputees would have an earlier onset of the hip flexor moment. As an extension or justification for this hypothesis, the authors suffixed the hypothesis by stating that “patients with TMT would have to pull their leg forward using hip flexor muscles to compensate for the limited push-off from the ankle” (p. 201). The rationale for this hypothesis does not seem at all intuitive and was not presented as part of the introduction to the article.
If amputees were to use a hip flexor gait strategy, whereby the amputated limb was advanced forward using the hip flexor muscles rather than relying on the ankle plantarflexors,5 then evaluating the hip power generation data, rather than the timing of the joint moment, would seem more illuminating. To further explore this concept, it is necessary to consider the graphical data for a single “illustrative” TMT amputee because no moment figures were presented for the amputee group as a whole.5 Following the initiation of the hip flexor moment (25% stance), the hip joint began absorbing power and did so until late stance. Moreover, the amputated limb is not in a trailing posture, as evidenced by the kinematic data. As such, there is little mechanical evidence to support the interpretation that a hip flexor strategy is being used to advance the lower limb. This does not mean that the hip flexor gait strategy does not exist, it simply implies that evaluating the timing of the hip moment and identifying the early onset of the hip moment are not good indicators of a hip flexor gait.
Although the hip flexor moment began significantly earlier in the TMT cohort compared with normal,5 it is difficult to be confident that this is a clinically important finding, particularly given the absence of a well-articulated rationale for examining this aspect of gait and the previously described concerns over the modeling underpinning the kinematic and kinetic data.
These hip moment data are perhaps best interpreted in light of the hip power data. Mueller et al.5 found no significant difference in the peak hip joint power during loading response on the amputated limbs of the TMT amputee cohort. In the case series investigation by Dillon,6,32 the amputated limb hip power generation during the same period was comparable to normal in the majority of PFA, consistent with findings reported by Mueller et al. However, in a small number of individuals, power generation during loading on the amputated side was significantly increased compared with the control sample, with peak power generation between 0.75 and 1.5 W/kg. Similar power generation was observed on the intact limb in the majority of unilateral amputee subjects during loading response.6,32
These results describe two types of mechanical work adaptation to compensate for the limited power generated across the amputated side ankle during terminal stance. The first pattern describes an increase in work across the intact side hip joint during early stance to provide forward impulse for the pelvis commensurate with the limited power generation across the amputated side ankle.6,32 The alternative pattern relies on increased power generation across the amputated side hip during early stance to propel the body forward from the rear. These adaptations are consistent with those observed in other amputee groups, where ankle power generation is limited, including studies on transfemoral62 and transtibial amputees.63
In the MTP amputee in whom ankle power generation was comparable to normal, these mechanical work adaptations at the hip joint were not observed.6,32
In terms of the hip flexor gait, the period of power generation across the hip during terminal stance (H3) for the single TMT amputee whose data has been graphically presented5 suggests that the integral of the H3 period would be somewhat comparable to that of the control subject. Similarly, Dillon6,32 found no significant difference in the magnitude or timing of peak power generation during terminal stance (H3) on either the intact or amputated limb in any of the amputee subjects.
4.7 PLANTAR PRESSURE
Provision of plantar pressure relief is recognized as an important part of plantar ulcer management,64 particularly in people with compromised circulation or sensation. There is some evidence that pressure relief is effective in preventing or treating diabetic foot ulcers.65 Given that foot ulceration is often a precursor to a PFA, it is logical to assume that plantar pressure may be an important factor in management of a PFA, particularly in individuals with compromised vascular status. A number of researchers have measured plantar pressure variables in people with a PFA.
4.7.1 DESCRIPTION OF STUDIES
Eleven studies were identified as reporting pressure variables, including 10 papers and one abstract.53 After reading those papers, one was eliminated because the data were considered unreliable because of methodological flaws.36 Thus, this section reports on the 10 remaining studies.27,33,34,38,41,42,45,53,55,66
Nine papers were ranked as observational in nature. These consisted of five case-control studies (O2), three cross- sectional studies (O3), and one case series (O5). In addition, there was one experiment, an interrupted time series trial (E3). None of these studies involved randomization of group or intervention.
4.7.2 METHODOLOGICAL QUALITY
The body of literature regarding plantar pressure and PFA is fairly recent (1988–2006),38,42 and confidence in the results is moderate.
PFA cohorts were generally small, with fewer than 10 participants. Two studies34,45 involved 27 and 30 people with a PFA, respectively. Five studies made comparisons to another group. Control groups were of similar size to the relevant PFA cohort, except in study by Armstrong and Lavery,34 which presented results from 150 control participants. Matching of characteristics between the PFA and control group was excellent in one study,38 good in two studies,33,34 and low in two studies.53,55 Kanade et al.38 recruited participants with diabetic peripheral neuropathy into all groups and reported careful control of a number of physical and disease characteristics. Armstrong and Lavery34 and Ademoglu et al.33 matched co-morbidity in the PFA and control group in that both groups had diabetes or both groups were otherwise healthy. Unfortunately, Kelly et al.55 and Randolph et al.53 compared a diabetic PFA cohort to healthy controls, making it difficult to attribute changes in pressure to the PFA or to the underlying disease.
The amputated side was compared with the intact side in 9 of the 10 studies. The one exception was the study undertaken by Armstrong and Lavery.34 Such side-to-side comparison is common in plantar pressure research. It is thought that the intact limb provides a “matched control” for the amputated limb because, theoretically, the limbs differ only in the unilateral pathology present. A risk of this assumption is that function of each limb affects that of the other limb and may contribute to differences in the variables measured.
Ambulatory plantar pressure variables were measured barefoot in five studies, four of which used a Novel Emed platform27,33,34,66 and one a Harris mat.42 In-shoe measurements were taken in the remaining five studies: one used the Novel Pedar system,38 three used the F-Scan system,41,45,55 and one system was unidentified.53 Footwear is known to reduce peak plantar pressure values compared with barefoot measurement.67 Absolute pressure values differ among platforms and among in-shoe systems.68–71 Reliability of systems varies.72 Provided calibration processes are followed and appropriate procedures are enacted, data from the same system within the same study can be confidently compared to itself and can be compared generally to data collected in a similar fashion, but caution must be exercised when comparing absolute values of data between studies that use different apparatus.
Eight of the 10 studies measured peak pressure (PP) on the plantar surface,27,33,34,38,41,45,53,55 and two described the PP distribution.42,66 Although PP is a commonly examined variable that has been linked to ulcer formation, interpretation is facilitated when time and contact area variables are also provided because tissue overloading is a complex interaction of force magnitudes or peaks, the area loaded, PPs applied, and the duration of loading. Just one study examined time and surface area aspects of PP and of peak force.55 Randolph et al.53 presented a duration of pressure result that was insufficiently defined.
The foot was either divided into 3 to 8 regions, considered as a whole, or the “forefoot” was identified as the region of interest. When the foot was divided into multiple regions, authors named each region such that it was apparent that all studies used different regional definitions.27,33,38,41,53 Descriptions of regions were insufficiently detailed to allow replication or identification of regional boundaries. This variation in regional definitions limits combination of data from multiple studies because the location of pressure cannot be standardized between papers.
Kelly et al.55 and Mueller et al.45 both examined a forefoot region that was the default size window (4.13 cm2) for the F-Scan software, and neither author clearly described how that window was located for each individual data set. As well as introducing the possibility of investigator bias in selection of the region of interest, similar variation in regional definition as described is present when this technique is used. If, to address this problem, window location relative to the foot boundary is automated by the software, the small size of the window may exclude areas of interest if PP locations shift between conditions.
The remaining three papers considered the foot as a whole.34,42,66 This resolves the problem of region definition but lacks discrimination between problematic locations and others.
Walking velocity has an effect on measurement of plantar pressure and should be taken into account in pressure studies involving multiple groups or multiple data collection episodes.67 Two of the five case-control studies and the only interrupted time series trial measured and reported walking velocity.38,45,55 Of those, only one included it as a covariate in some, but not all, analyses.38 Control for walking velocity is not required if left-to-right comparisons are made using bilateral data from the same walking trials because, obviously, velocity is the same for both feet.
Five papers reported selection of statistical tests that were appropriate for the investigation concerned.27,34,41,45,55 Three studies undertook no statistical analysis of pressure data, basing interpretation of the results solely on description and observation.33,42,66 One study (an abstract) briefly described analyses that were appropriate for part of the work.53 The final paper38 reported precise statistical analyses for most of the work, from power analysis in the establishment of group sizes to testing of relationships between variables and analyses for main effects of independent variables. Unfortunately, the authors did not conduct any pair-wise (post hoc) comparisons to identify differences between groups, so conclusions could not be drawn about effects of amputation or amputation level on regional or whole foot PP.
All authors presented some pressure-related results, although some studies omitted reporting of some variables identified in procedures, and two studies provided brief statements about whole foot pressure patterns.42,66 Results were presented by authors in text, tabulated or graphed to varying degrees of clarity.
18.104.22.168 Peak Pressure Magnitude in Foot Regions
The reviewed studies provide moderate evidence that a PFA causes forefoot or metatarsal head PP to increase when compared with the intact side or appropriate controls. Findings from several papers support this generalization to some extent.27,33,34,41,42,45,53 It is impossible to refine this observation to describe effects of particular amputation levels because of limitations in the studies, as described.
Although mean PP magnitudes provided by Kanade et al.38 contradict the above observation, no statistical analyses were conducted to test for differences between conditions, and the PFA group includes multiple amputation levels. For example, the five TMT amputations in the group are likely to contribute to PP measured in the heel and midfoot regions but may not load the metatarsal head regions at all. Inclusion of zero values for regions not loaded would artificially reduce PP scores for those regions. The brief observational statement made about one case by Ramseier et al.66 that “both feet show low pressures under the first metatarsal head” (p. 57) contributes little to the contradictory argument. Stronger evidence against an increase in distal PP is provided by Kelly et al.,55 who reported no difference in forefoot PP when diabetic individuals with a TMT amputation were compared to the contralateral side and to age-matched healthy control subjects. It is possible that the significantly reduced walking velocity of the TMT group reduced forefoot PP such that it was not different from the control group or contralateral side.
A very low level of evidence exists demonstrating reduced heel PP following a PFA when compared with the contralateral limb. Two studies reported significant heel PP reductions,27,41 and two studies provided mean data that suggest no change in PP but provided no statistical test of differences.33,38
22.214.171.124 Peak Pressure Timing
There is insufficient evidence for any conclusion to be made about effects of PFA on timing aspects of PP. The time PP occurs is important because it identifies when in the gait cycle a reduction may be required. Kelly et al.55 were the only investigators to measure this variable and report it as a percentage of stance phase. They found no differences among the TMT amputated side, contralateral side and control group. Poorly matched control subjects and omission of walking velocity as a covariate in analyses reduce confidence in findings from this study. Randolph et al.53 reported differences in the “duration of pressures” at the heel and midfoot but did not state whether the comparison was between foot regions (heel versus mid-foot), limbs (amputated versus intact side), or group (amputee versus control).
126.96.36.199 Peak Force Magnitude and Timing
Insufficient evidence exists for any conclusion to be made about effects of PFA on peak force magnitude and timing. To date, Kelly et al.55 are the only investigators to measure these variables. They reported no differences in peak force magnitude among the TMT amputated side, contralateral side and control group. Kelly et al.55 noted that peak force occurred significantly earlier in the TMT group than in the control group. Unfortunately, inadequately matched control subjects and omission of walking velocity as a covariate in analyses raise questions about results from this study.
188.8.131.52 Area Loaded
Similar to the above two variables, there is insufficient evidence for any conclusion to be made about effects of PFA on areas loaded. Only Kelly et al.55 has reported the plantar area through which force is applied, producing pressure, and found no differences among the TMT amputated side, contralateral side and control group. Again, inadequately matched control subjects and omission of walking velocity as a covariate may affect interpretation of this result.
Logic tells us that amputation of parts of the foot must reduce the plantar surface available for loading. Quantification of the area actually loaded at key points in the gait cycle, for each level of amputation, should inform design of pressure reduction interventions. For example, if there is insufficient plantar surface remaining to distribute weight-bearing forces at important stages of gait such that pressure reduces, transfer of force to more proximal regions such as the tibia may be indicated.
184.108.40.206 Relationships Between Plantar Pressure Variables and Other Variables
There is insufficient evidence of the effect of PFA on relationships between pressure variables and other variables or, indeed, on relationships between pressure variables themselves. Although seven studies measured nonpressure variables,27,38,42,45,53,55,66 only Kanade et al.38 tested relationships between pressure variables and others. Kanade et al.38 demonstrated a significant positive relationship between whole foot PP and severity of foot pathology in diabetic participants with peripheral neuropathy when intact feet were compared with ulcerated feet and feet with a PFA. Kelly et al.55 and Mueller et al.45 measured slower walking velocity in the PFA group and used that result to assist in interpreting pressure data but did not include it in statistical analyses. Similarly, Garbalosa et al.27 measured passive and walking ankle ROM, demonstrated a significantly reduced dorsiflexion range in the TMT amputated side compared with the contralateral side, and used that information when interpreting pressure data. Mann et al.42 and to a much smaller extent Ramseier et al.66 and Randolph et al.53 measured or described various gait parameters and made clinical observations, then described how the parameters and observations fitted with pressure data. When considered together, these papers indirectly provide support for the assumption that ambulatory plantar pressure and gait variables are related.
220.127.116.11 Effectiveness of Pressure Reduction Techniques
This review provides no evidence to recommend any pressure reduction technique for the PFA. Only one paper compared various pressure reduction interventions.45 Five different combinations of pressure reduction techniques (full-length and short footwear, custom-molded shoe inserts, custom-molded AFO, and rigid rocker-bottom soles) were equally effective in reducing forefoot PP compared with a full-length shoe with toe filler. The combinations tested represented an eclectic mix of techniques, which were presented in such a way that it was impossible to distinguish the most effective technique(s) from the least or ineffective options. In addition, although six combinations were tested, they were not presented in random order, which may have affected the results.
It is possible that gait adjustments made by the individual, such as slowing walking velocity, are as or more effective than footwear or devices in reducing pressure to vulnerable foot regions.
Optimization of plantar pressure following a PFA to minimize risk of further breakdown requires knowledge of PP locations, magnitudes, and timing. To understand pressure, it is necessary to know about the force involved (location, magnitude, and timing) and the size and location of the area loaded. As force and area (and so pressure) change throughout the gait cycle, these variables need to be examined throughout the gait cycle and related to kinematic and kinetic changes during gait. Such examination should involve careful selection of similar amputation levels and of control subjects.
4.8 ENERGY EXPENDITURE
4.8.1 DESCRIPTION OF STUDIES
Review of abstracts identified two papers as reporting variables that reflect energy expenditure.38,46 Both studies were observational in nature and compared a PFA group to a disease-matched control group, and both studies included additional groups with more proximal amputations. PFA group sizes were 546 and 16,38 which were compared with control groups of similar size.
4.8.2 METHODOLOGICAL QUALITY
Kanade et al.38 used the Total Heartbeat Index (THBI) as an estimate of energy expenditure during a 2-minute walk test. The THBI is a ratio of total heartbeats per meter travelled.73 Although Kanade et al.38 presented units of beats per minute (beats/min) in results tables and graphs, the means presented indicate that beats per meter (beats/m) were probably calculated. A heart rate monitor recorded the total number of heartbeats; instrumentation for recording of total distance was not described.38 Pinzur et al.46 measured the amount of oxygen consumed per kilogram of body weight per minute to indicate energy expenditure. They also calculated energy cost (ratio of energy expenditure at comfortable to maximum velocities) and functional energy cost (ratio of oxygen consumption at a comfortable velocity to resting oxygen consumption). A Douglas Air Bag was used to record oxygen consumption at rest and while each participant walked on a treadmill at their comfortable velocity and at their maximum velocity.46
Unfortunately, neither study presented statistical analysis to compare energy expenditure between groups.38,46 Both studies provided mean data and illustrative graphs. Kanade et al.38 conducted appropriate statistical tests for relationships between energy expenditure and the severity of limb pathology.
There is insufficient evidence to indicate whether PFA causes an increase in energy expenditure during gait, compared with a nonamputated foot or a more proximal amputation. Although mean scores presented in both studies suggest an increase may be present compared to a whole foot, the lack of statistical comparison diminishes confidence in this observation. Kanade et al.38 reported a significant positive relationship between the severity of limb pathology and the THBI, demonstrating that energy expenditure increased as pathology became more extensive across four categories: a whole foot, ulcerated foot, PFA, and transtibial amputation. However, their analysis did not allow group-to-group comparisons to be made.
No studies have attempted to compare the effects of different levels of PFA on energy expenditure. Although Kanade et al.38 included participants with different levels of PFA, they were combined in a single group such that no level-by-level comparison could be made.
4.9.1 DESCRIPTION OF STUDIES
EMG data were reported in two investigations6,42 for a cohort of three individuals with hallux amputation42 or as part of a case series, including a single MTP, one TMT, and three Lisfranc amputees.6 Both of these investigations reported data on the amputated limb for tibialis anterior and gastrocnemius medial head,6,42 whereas other muscles examined by Dillon6 included gastrocnemius lateral head, soleus, biceps femoris, and rectus femoris. Intact limb EMG data were also available for the same muscle set.6
One investigation used in-dwelling electrodes,42 whereas the other used surface electrodes.6 Mann et al.42 did not report further details about the EMG equipment, data collection, or signal processing, but sufficient information was provided in the other investigation6 to consider the results in light of some methodological issues.
4.9.2 METHODOLOGICAL QUALITY
The EMG data reported by Dillon6 seems to have been affected by a number of signal processing issues. The data appear to be systematically affected by a noisy baseline, and the amplitude of the EMG signal is small, which would make distinguishing meaningful muscle activity from the underlying noise difficult.
The effective bandwidth of the filtered EMG signal was 6 to 500 Hz,6 which may, in part, explain the noisy baseline. Other investigations using similar EMG techniques have high-pass filtered their EMG data with a cutoff frequency between 10 and 40 Hz,74–76 resulting in a more stable baseline, particularly at the higher cutoff frequencies. Alternative techniques, such as measuring the baseline noise and subtracting this from the walking data, were not used by Dillon.6
The amplitude of the EMG data was about one-third that of other investigations that have used similar amplitude normalization techniques.77,78 The small amplitude of the EMG signal is a likely result of the method by which the voltage, reflective of 100% muscle activity, was determined. Although the manual muscle test (MMT) seems to have been appropriately conducted, the voltage characteristic of 100% muscle activity was determined by averaging the peak voltages of each 10 millisecond interval of the integrated signal over a stable 1-second sample of maximum contraction EMG data.6 Determining the voltage characteristics of 100% muscle activity in this way is somewhat unconventional for the MMT technique and would have reduced the overall magnitude of the gait EMG data.
Dillon6 distinguished periods of muscle activity using a threshold-based criteria in which EMG data in excess of 5% MMT with duration of more than 5% of the gait cycle was considered reflective of meaningful muscle activity. However, much of the data reported seem to have baseline noise in excess of the 5% MMT. As such, many of the periods of muscle activity reported seem uncharacteristically long and appear consistent with a noisy baseline signal. Alternative techniques using threshold criteria calculated as the mean plus a number of standard deviations above the baseline noise may have proven more reliable.
Publications in this area provide an insufficient contribution to understanding the intensity and periods of muscle activation typically available through reporting of EMG.
Mann et al.42 reported that the phasic activity of tibialis anterior was within normal limits. Tibialis anterior was active between 0% and 14% of the gait cycle (GC) and from 55% to 100% GC. A similar statement was made with regard to gastrocnemius medial head; however, the authors reported “normal phasic activity from 30% to 5% GC” (p. 202). No graphical results were presented, which made it difficult to interpret this finding.
Limitations in the signal processing technique significantly compromised the usefulness of Dillon’s6 contribution to understanding the muscle activity of PFA. As such, the results were considered inadequate to report here.
This review used a study design classification system that is under development by the AAOP79 and is described in Appendix 1. Using this system, each publication was ranked based on study design. Similar to ranking schemes applied by other systematic review processes (e.g., Cochrane Collaboration), the system values structured reviews the most, followed by experimental work with randomized control trials ranking most highly, then observational studies, and finally expert opinion.
The relevance of such rating systems to the PFA literature reviewed here requires clarification. First, study design in the context of the knowledge base should be considered, and second, confidence in the findings from the studies should be assessed. Thus, the outcomes of this review are summarized in tables that combine study rankings, confidence in the results, and reference to the relevant section of this review where the results are discussed in detail.
5.2 STUDY DESIGN IN THE CONTEXT OF THE KNOWLEDGE BASE
When interpreting the ranking assigned to each study, it is useful to remember that the research aim or question determines the most appropriate study design. Consideration of the relationship between the study aim and design is helpful when ranking papers systematically because it clarifies the value of the work being ranked in the context of the existing knowledge base. Not all knowledge requires a randomized control trial to demonstrate that it is factual or valuable. A model for contextualizing evidence rankings is provided here.
In 2004, Morris80 provided a framework for understanding current orthotics research and explained how the National Institutes of Health phases of clinical trials could be used to suggest the type of research needed. The underlying premise of the trial phases is that each new intervention should first be demonstrated to be safe and effective in a controlled environment on few participants before being compared with other options and tested in the wider community.81 Levels of study design generally parallel this process. Morris’ ideas may be expanded by including a preliminary phase that defines a problem and proposes a solution and by embedding the relevant AAOP study designs within each research phase. This relationship among the research purpose, phase of clinical trial, and study design is presented in Table 4.
Addressing a new problem or introducing a new intervention requires sequential progression through the levels of purpose, phases of trials, and study design (Table 4). To begin, people (often experts) define a problem and collect sufficient foundation knowledge to propose a solution, using observational studies if available. The solution is then tested in a safe setting and is compared to the best alternative solution, using experiments with small cohorts or before-and-after methods. (It is pertinent to note that prosthetics and orthotics practice often involves provision of devices before they have been independently tested, making it possible to conduct an observational study of an intervention already provided. This is different from many other fields, such as pharmacology, in which legislation demands new products are tested experimentally before being provided clinically.) Only after an intervention is proven to have an effect and be safe is it tested further in the wider community through implementation of randomized control trials. When the solution is applied more generally in the community, it should be monitored for side effects using an interrupted time series trial and mandatory reporting of adverse events.
The evidence within this review can be considered in the context of this model. Most researchers are working to establish foundation knowledge about the gait of PFA. This is evidenced by studies that seek to measure various gait parameters and establish differences from the normal population. Study designs that address these aims include expert opinion and observational studies, and most of the literature included in this review involved the latter.
The ability to answer questions about the efficacy of various prosthetic and orthotic interventions is limited by the fledgling body of evidence. This need not be a negative reflection on the literature but merely a statement marking our current position in a continuum of understanding from foundation knowledge to effectiveness and from observational studies to experiments to structured reviews. There is good reason for our current position within this continuum.
Before meaningful experimental investigations can be undertaken to understand the benefits of one intervention over another, observational studies are a necessary first step. Observational studies provide a platform of understanding upon which experiments may be built. Although it may seem fairly self-evident, until observational studies were carried out, it was not known, for example, whether PFAs walked more slowly; what effect systemic disease had on walking velocity; whether differences existed based on level of amputation or type of prosthetic and orthotic intervention; or what the underlying causes of reduced walking velocity were. Without the insights provided by observational studies, it is difficult to build worthwhile and methodologically sound experiments.
Many of the shortcomings of the experimental studies published to date could be the end result of researchers “jumping the gun” to describe the efficacy or effectiveness of an intervention without first understanding the insights provided by observational studies. To use walking velocity as a convenient example, experiments reporting this parameter consistently failed to control for the influence of systemic disease to a greater44,45 or lesser48 extent. Each of these studies failed to provide clear and well-rationalized hypotheses conveying how various interventions would affect walking velocity. Interventions were inadequately considered,44,45,48 and often multiple interventions were provided at once.44,45 Outcome measures were reported, usually without strong a priori hypotheses,44,45,48 probably because investigators were unsure about the underlying mechanics of the intervention and what they were looking for.
As a result of these inadequately designed experiments, little is known about the efficacy of one intervention over another for PFA. Thus far, insights have been provided mostly by observational studies that were not designed to elucidate the effect of different interventions on gait. As such, there are a number of insights into the influence of prosthetic and orthotic intervention on gait that remain to be pursued and proven with appropriately designed experiments.
5.3 QUALITY OF THE EVIDENCE
The draft AAOP rating system allowed reasonably clear definitions of study design but, like other rating systems,82,83 does not take into account methodological rigor or quality of the evidence.
Any study design may have problems in implementation of the method or in reporting of the study that reduce confidence in the outcomes or quality of the conclusions. Included in the results section of this review are many examples of methodological problems that affect the reliability or validity of outcome measures and bias results. Although these sorts of methodological problems are often highlighted within a journalistic review, as was attempted in this systematic review, these should be considered when determining the strength or level of evidence as part of the conclusions of a systematic review.
By jointly considering the “ranking” based on study design and the “level of confidence” the reviewers have in the outcomes reported (given due consideration of the impact of methodological problems), it is possible to provide a better indication of the strength of evidence in support of a particular outcome. In reporting the outcomes of this review, such an approach was adopted.
The “level of confidence” in a particular outcome was determined in the following way: Each paper that contributed to our understanding of a particular outcome was rated according to an adjective-driven scale describing the level of confidence (Table 5). As an illustrative example, Figure 5 portrays the level of confidence each paper contributed in support of the following outcome statement: “Partial foot amputees with a history of diabetes and vascular disease walk slowly—about two-thirds the velocity of healthy controls.” Most works instilled a high degree of confidence in support of this outcome statement (Figure 5). Three investigations offered only a moderate degree of confidence given modest methodological problems. In two of the three cases, investigators expressed results as a proportion of the laboratory normal database without describing the normal population or their walking velocity.51,52 The remaining investigation38 measured walking velocity using video footage of amputees walking past two rods. The rods were placed one meter apart, providing an almost instantaneous measure of walking velocity, but the results were comparable to other investigations using more robust techniques. One investigation42 instilled an insufficient level of confidence because the researchers did not present data aside from a subjective comment that walking velocity did not differ between the sound and affected limbs despite that walking velocity is not a side-to-side dependent measure.
Overall, the literature provided a “high” degree of confidence in support of this particular outcome statement (Figure 5), suggesting that “the reader has high confidence in the outcome based on findings from multiple independent investigations that consistently support the statement. The articles, on the whole, are methodologically strong; or where methodological issues occur, they are unlikely to impact the confidence with which the statement can be made.” Similar adjectival descriptors have been provided for each level of confidence when groups of papers were synthesized in support of a particular outcome statement (Table 6).
5.4 OUTCOMES FROM THIS REVIEW
The final section of this discussion presents outcome statements for the various aspects of gait incorporated in this review and, when possible, includes comments as to the relative efficacy of different orthoses and prostheses (Tables 7–13). Each statement is accompanied by a summary of the number of studies of each design used to assess the statement and the level of confidence in the results. When considered together, the ranking of study design and level of confidence provide an indication of the strength of evidence attached to each statement. For example, in Table 7 (Outcome Statements for Temporospatial Gait Variables), statement 2 reflects findings from 10 papers, including two experiments and eight observational studies, and inspired a high level of confidence, suggesting a strong level of evidence. In contrast, statement 4 reflects findings from 13 papers, including the same experiments and observational studies, but inspired an insufficient level of confidence, suggesting low or insufficient evidence. For further clarification and discussion regarding each statement and the assigned level of confidence, the reader is directed to the relevant section of the review document in the final column of Table 7.
5.5 LIMITATIONS OF THIS REVIEW
An obvious limitation of this review and potential source of bias was the contribution of one of the reviewers (M. Dillon) to several of the publications included in this review. Potential bias was hopefully attenuated to some extent by the contribution of two reviewers who have not published on this topic. All reviewers endeavored to critically evaluate and report insights into the available body of literature in an impartial way.
For the purposes of the review, PFA was defined as an amputation of a portion of either the hind or forefoot. As such, publications that focused solely on amputations disarticulating the ankle joint (Symes) were excluded consistent with the ISO definition50 of PFA. Although partial foot and Symes amputations have often been reported together within the literature, they are indeed unique amputation procedures with relatively unique interventions and complications. It is acknowledged that some readers may find this decision a limitation of the review.
The review does not report on footwear interventions, aside from results presented in a few experimental studies. This does not reflect the discarding of relevant literature but is a reflection of the limited research published on this issue in the journals accessed as part of this review. Given that footwear interventions seem to be common in this population, it was surprising that more biomechanics-related research regarding the effect of footwear on PFA gait was not identified.
Following the initial literature search, the reviewers made the decision to include other sources of work not published in the journals indexed by the databases used for this review. As such, several abstracts and a thesis were included in the review because they met all other inclusion criteria and had undergone independent peer review. The reviewers believed it would be an oversight to exclude meritorious publications simply because they were not suitably indexed. It is acknowledged that although the nonsystematic inclusion of these studies limits the reproducibility of the literature search and that similar studies may have been overlooked, the additional four publications contributed significantly to the body of evidence and our understanding of the topic.
It should be noted that the abstracts provided by RECAL are typically descriptions written by a third party and tend to provide less information than abstracts available through the other databases used in this review. This may have limited the reviewer’s ability to determine the suitability of the work for inclusion in the review, but it is hoped that this was attenuated to some degree by searching the RECAL database last.
Through the process of ranking and reviewing the literature, concern developed regarding the ability of the study design classification scheme to adequately describe the level of evidence and reviewer confidence in the findings. As such, an alternative method for describing the quality of the literature was developed that allowed reviewers to articulate their confidence in the findings based on the quality of the methodology in conjunction with the classification of study design. The scheme attempts to address, to some extent, the limitations of existing methods for reporting the level of evidence based solely on hierarchical ranking of study design. This limitation is particularly an issue in prosthetics and orthotics literature, where studies tend to be predominantly observational and therefore ranked lower, and there is no further ability to discriminate as to the value and quality of the evidence. Although the reviewers are confident that the proposed scheme served this review well and may be useful for other reviews in prosthetics and orthotics, its applicability to larger and more developed bodies of literature has not been evaluated.
5.6 FUTURE RESEARCH DIRECTIONS
Our understanding of the biomechanics of PFA gait and the influence of prosthetic and orthotic intervention is very much in its infancy. Until several years ago, our understanding of PFA gait was based on static force analyses. There has seen a significant increase in quantifiable biomechanics research in the last decade (Figure 2). The current literature consists predominantly of observational studies (Figure 4) consistent with the need to establish a foundation of knowledge (Table 4).
Based on the outcome of the systematic review, it is clear that our understanding of the biomechanics of gait following PFA lacks depth. There exists a high level of evidence only for broad observations in a few aspects of gait, such as “PFA has an affect on ankle kinematics,” with limited evidence to support a more detailed understanding of how PFA affects ankle kinematics. For example, there is low evidence that ankle dorsiflexion is affected by PFA. We are unable to answer questions such as “Is the peak dorsiflexion angle increased or decreased compared to normal?” with any confidence, let alone understand the influence of prosthetic and orthotic intervention. Understanding of the biomechanics of PFA gait is still in its infancy.
A greater emphasis needs to be placed on improving the depth of foundation knowledge so that we may attempt to answer some basic questions about PFA gait and establish the efficacy of prosthetic and orthotic intervention. In further developing this body of literature, we need to learn more about the underlying mechanics of PFA gait so that we can develop well-rationalized hypotheses regarding the effect of prosthetic and orthotic intervention. There are consistent methodological issues that need to be recognized and resolved, many of which have been described in this review. These methodological issues limit the quality of many investigations to date and compromise the confidence we may invest in the evidence. Without addressing these issues, our ability to progress the body of knowledge toward studies of effectiveness (phase III) is limited.
Many of the methodological issues identified in this review reflect the small and heterogenous population typical of PFA. This can make large-scale studies challenging. The PFA populations in the reviewed studies were heterogeneous in terms of time since amputation, number of limbs amputated, contralateral limb involvement, age, or presence of systemic disease. Although challenging, many of these issues of heterogeneity can be resolved with careful planning and appropriate study design. There exist many study designs and statistical techniques appropriate for small populations that seem not to have been used by researchers to date. Where appropriate, studies may also be improved substantially through randomization of interventions, matching of groups to control for the influence of systemic disease, establishing and testing a well-rationalized a priori hypothesis, or even presenting well-considered discussion that can inform others.
It is many of these methodological issues that have affected, in particular, the experimental investigations designed to elucidate the efficacy and effect of various prosthetic and orthotic interventions. It could be argued that there is insufficient foundation knowledge to be conducting phase III investigations in most areas of PFA gait. Most researchers, and indeed P&O clinicians, want to establish the effectiveness of one intervention over another. Third party payers seek this level of understanding to justify prescription or the cost of intervention. However, establishing the sort of quality research necessary to contribute meaningfully to the evidence base at a phase III level is a developmental process and one that is not well served without a strong foundation.
In line with the tabled evidence (Tables 7–13), future research should be aimed at developing well-controlled and rigorously designed observational studies to improve the level of evidence for items marked as “moderate,” “low,” or “insufficient” evidence. In some cases, items identified as having “low evidence” require additional well-developed research to support understanding that is currently based on a single small, well-designed investigation whose evidence can not, in isolation, be invested with greater confidence. In other cases, methodological issues have limited the confidence of various findings, and researchers should address these issues in designing future studies so as to enhance the value of their contribution.
Given ongoing issues with small sample sizes, researchers should consider the potential for collaboration whereby investigators using comparable instrumentation and methodologies might pool their findings. The value of small sample research or even well-designed case studies should not be ignored, particularly given that it is such small samples that are most appropriate to developing foundation knowledge and phase I research. In this review, some of the research that inspired the greatest confidence, and therefore the highest level of evidence, came from small, well-designed investigations.
The purpose of this systematic review was to establish what is known about gait and prosthetic/orthotic intervention in persons with PFA and to identify what needs to be known to optimize gait and prosthetic/orthotic intervention. A systematic search of the literature identified 437 citations, with 28 publications selected for review based on inclusion criteria. Overall, there was a high level of evidence that PFA generally affects temporospatial, GRF, ankle kinetics, and plantar pressures during gait, but there was less confidence in the evidence regarding more detailed statements about exactly how these aspects of gait are affected by PFA or prosthetic and orthotic intervention. This is reflective of the small and heterogenous populations included in the reviewed literature. Because the studies were largely observational, there is insufficient evidence regarding the efficacy of specific prosthetic and orthotic interventions, although generally there was low to moderate evidence that prosthetic and orthotic interventions affect ankle kinematics and moments and may moderate CoP progression. There is a need to further our understanding of the biomechanics of PFA gait and establish hypotheses regarding prosthetic and orthotic requirements for improved ambulation/function and protection of the residuum. These hypotheses should then be evaluated with well-controlled, rigorously designed experimental studies to ascertain the efficacy of prosthetic and orthotic devices in the treatment of PFA.
7.1 ARTICLES REVIEWED IN DETAIL
Table 14 is an alphabetized list of articles that met the inclusion criteria and were considered in detail during this review. The reference numbers are consistent with those used in text and in the reference list.
1. Mueller MJ, Allen BT, Sinacore DR. Incidence of skin breakdown and higher amputation
after transmetatarsal amputation
: implications for rehabilitation. Arch Phys Med Rehabil
2. Miller N, Dardik H, Wolodiger F, et al. Transmetatarsal amputation
: the role of adjunctive revascularization. J Vasc Surg
3. McKittrick LS, McKittrick JB, Risley TS. Transmetatarsal amputation
for infection or gangrene in patients with diabetes mellitus. J Am Podiatr Med Assoc
4. Sanders LJ, Dunlap G. Transmetatarsal amputation
. A successful approach to limb salvage. J Am Podiatr Med Assoc
5. Mueller MJ, Salsich GB, Bastian AJ. Differences in the gait
characteristics of people with diabetes and transmetatarsal amputation
compared with age-matched controls. Gait Posture
6. Dillon MP. Biomechanical models for the analysis of partial foot amputee gait
[dissertation]. Brisbane, Australia: Mechanical, Manufacturing and Medical Engineering, Queensland University of Technology; 2001.
7. Tang SFT, Chen CPC, Chen MJL, et al. Transmetatarsal amputation
prosthesis with carbon-fiber plate: enhanced gait
function. Am J Phys Med Rehabil
8. Pinzur MS, Wolf B, Havey RM. Walking pattern of midfoot and ankle disarticulation amputees. Foot Ankle Int
9. Greene WB, Cary JM. Partial foot
amputations in children. A comparison of several types with the Syme amputation
. J Bone Joint Surg Am
11. 3222.0 - population projections, Australia, 2004 to 2101. Australian Bureau of Statistics. Available at: http://www.abs.gov.au/
. Accessed December 12, 2006.
12. Collins JN. A partial foot
prosthesis for the transmetatarsal level. Clin Prosthet Orthot
13. Imler CD. Imler partial foot
prosthesis IPFP: “Chicago boot.” Clin Prosthet Orthot
14. Latorre R. The total contact partial foot
prosthesis. Clin Prosthet Orthot
15. Lange LR. The Lange silicone partial foot
prosthesis. J Prosthet Orthot
16. Moore JW. Prostheses, orthoses, and shoes for partial foot
amputees. Clin Podiatr Med Surg
17. Rubin G. Indications for variants of the partial foot
prosthesis. Orthop Rev
18. Rubin G. The partial foot amputation
. J Am Podiatr Assoc
19. Stills ML. Partial foot
prostheses/orthoses. Clin Prosthet Orthot
20. Condie DN. Biomechanics of the Partial-Foot Amputations.
1st ed. London: Edward Arnold Publishers Ltd.; 1970.
21. Mueller MJ, Sinacore DR. Rehabilitation factors following transmetatarsal amputation
. Phys Ther
22. Weber D. Partial foot amputation
. In: Weber D, Agro M, eds. Clinical Aspects of Lower Extremity Prosthetics.
Oakville, Ontario: Elgin Enterprises;1991:197–209.
23. Barry DC, Sabacinski KA, Habershaw GM, et al. Tendo Achillis procedures for chronic ulcerations in diabetic patients with transmetatarsal amputations. J Am Podiatr Med Assoc
24. Giurini JM, Basile P, Chrzan JS, et al. Panmetatarsal head resection. A viable alternative to the transmetatarsal amputation
. J Am Podiatr Med Assoc
25. Santi MD, Thoma BJ, Chambers RB. Survivorship of healed partial foot
amputations in dysvascular patients. Clin Orthop Relat Res
26. Stuck RM, Sage R, Pinzur M. Amputations in the diabetic foot. Clin Podiatr Med Surg
27. Garbalosa JC, Cavanagh PR, Wu G, et al. Foot function in diabetic patients after partial amputation
. Foot Ankle Int
28. Pullen JJ. A low profile paediatric partial foot
. Prosthet Orthot Int
29. Sanders LJ. Transmetatarsal and midfoot amputations. Clin Podiatr Med Surg
30. Rubin G, Danisi M. Functional partial-foot prosthesis. Bull Prosthet Res
31. Dillon MP, Barker TM. Can partial foot
prostheses effectively restore foot length? Prosthet Orthot Int
32. Dillon MP, Barker TM. Preservation of residual foot length in partial foot amputation
: a biomechanical analysis. Foot Ankle Int
33. Ademoglu Y, Ada S, Kaplan I. Should the amputations of the great toe be replanted? Foot Ankle Int
34. Armstrong DG, Lavery LA. Plantar pressures are higher in diabetic patients following partial foot amputation
. Ostomy Wound Manage
35. Erdemir A, Piazza SJ. Rotational foot placement specifies the lever arm of the ground reaction force during the push-off phase of walking initiation. Gait Posture
36. Friedmann LW, Padula PA, Weiss JM. Studies on the survival of transmetatarsal amputation
stumps. Vasc Surg
37. Hirsch G, McBride ME, Murray DD, et al. Chopart prosthesis and semirigid foot orthosis in traumatic forefoot amputation
: comparative gait
analysis. Am J Phys Med Rehabil
38. Kanade RV, van Deursen RWM, Harding K, Price P. Walking performance in people with diabetic neuropathy: benefits and threats. Diabetologia
39. Kelly JJ. Peripheral neuropathy: the importance of history and examination for correct diagnosis. Rev Neurol Dis
40. Kulkarni J, Curran B, Ebdon-Parry M. Total contact silicone partial foot
prostheses for partial foot
41. Lavery LA, Lavery DC, Quebedeax-Farnham TL. Increased foot pressures after great toe amputation
in diabetes. Diabetes Care
42. Mann RA, Poppen NK, O’Konski M. Amputation
of the great toe. A clinical and biomechanical study. Clin Orthop Relat Res
43. Mueller MJ, Salsich GB, Strube MJ. Functional limitations in patients with diabetes and transmetatarsal amputations. Phys Ther
44. Mueller MJ, Strube MJ. Therapeutic footwear: enhanced function in people with diabetes and transmetatarsal amputation
. Arch Phys Med Rehabil
45. Mueller MJ, Strube MJ, Allen BT. Therapeutic footwear can reduce plantar pressures in patients with diabetes and transmetatarsal amputation
. Diabetes Care
46. Pinzur MS, Gold J, Schwartz D, Gross N. Energy demands for walking in dysvascular amputees as related to the level of amputation
1992; 15:1033–1036; discussion 1036–1037.
47. Salsich GB, Mueller MJ. Relationships between measures of function, strength and walking speed in patients with diabetes and transmetatarsal amputation
. Clin Rehabil
48. Tang SFT, Chen CPC, Chen MJL, et al. Transmetatarsal amputation
prosthesis with carbon-fiber plate: enhanced gait
function. Am J Phys Med Rehabil
49. Andersen JA, Klaborg KE. Forefoot amputation
in rheumatoid arthritis. Acta Orthop Scand
50. Schuch C, Pritham CH. International forum: International Standards Organisation terminology: application to prosthetics and orthotics. J Prosthet Orthot
51. Boyd LA, Rao SS, Burnfield JM, et al. Forefoot rocker mechanisms in individuals with partial foot amputation
. Paper presented at the Gait
and Clinical Movement Analysis, Dallas, Texas, March 10–13, 1999. Gait Posture
52. Burnfield JM, Boyd LA, Rao SS, et al. The effect of partial foot amputation
on sound limb loading force during barefoot walking. Paper presented at the Gait
and Clinical Movement Analysis, San Diego, CA, April 15–18, 1998. Gait Posture
53. Randolph AL, Tin WW, Abayev B. Foot pressures after great toe amputation
in diabetic patients: a pilot study (abstract). Arch Phys Med Rehabil
54. Wilson EJ, Dillon MP. Restoring center of pressure excursion using a toeoff orthosis in a single partial foot
amputee. Paper presented at the International Society of Prosthetics and Orthotics, Australian National Member Society, Carlton Crest Hotel, Sydney, Australia, November 2–5, 2005. D.C. Conferences, NSW
55. Kelly VE, Mueller MJ, Sinacore DR. Timing of peak plantar pressure during the stance phase of walking: a study of patients with diabetes mellitus and transmetatarsal amputation
. J Am Podiatr Med Assoc
56. Hansen AH, Meier MR, Sessoms PH, Childress DS. The effect of prosthetic foot roll-over shape arc length on gait
of transtibial prosthesis users. Prosthet Orthot Int
57. Allard P, Lachance R, Aissaoui R, et al Able-bodied gait
in men and women. In: Allard P, Capozzo A, Lundberg A, Vaughan C, eds. Three-dimensional Analysis of Human Locomotion.
West Sussex, United Kingdom: John Wiley and Sons. Ltd.;1997:307–334.
58. Winter DA. Biomechanical motor patterns in normal walking. J Mot Behav
59. Winter DA. Kinematic and kinetic patterns in human gait
: variability and compensating effects. Hum Mov Sci
60. Winter DA. Biomechanics
of normal and pathological gait
: implications for understanding human locomotor control. J Mot Behav
61. White SC, Lage KJ. Changes in joint moments due to independent changes in cadence and stride length during gait
. Hum Mov Sci
62. Seroussi RE, Gritter A, Czernickie JM, Weaver K. The mechanical work adaptations of above-knee amputee ambulation. Arch Phys Med Rehabil
63. Winter DA, Sienko SE. Biomechanics
of below-knee amputee gait
. J Biomech
64. Brem H, Sheehan P, Rosenberg HJ, et al. Evidence-based protocol for diabetic foot ulcers. Plast Reconstr Surg
2006;117:193S–209S; discussion 210S–211S.
65. Spencer S. Pressure relieving interventions for preventing and treating diabetic foot ulcers. Cochrane Database Syst Rev
66. Ramseier LE, Jacob HAC, Exner GU. Foot function after ray resection for malignant tumors of the phalanges and metatarsals. Foot Ankle Int
67. Burnfield JM, Few CD, Mohamed OS, Perry J. The influence of walking speed and footwear on plantar pressures in older adults. Clin Biomech (Bristol, Avon)
68. Putti AB, Arnold GP, Cochrane L, Abboud RJ. The Pedar in-shoe system: repeatability and normal pressure values. Gait Posture
69. Murphy DF, Beynnon BD, Michelson JD, Vacek PM. Efficacy of plantar loading parameters during gait
in terms of reliability, variability, effect of gender and relationship between contact area and plantar pressure. Foot Ankle Int
70. Hsiao H, Guan J, Weatherly M. Accuracy and precision of two in-shoe pressure measurement systems. Ergonomics
71. Quesada P, Rash G, Jarboe N. Assessment of Pedar and F-scan revisited. Clin Biomech (Bristol, Avon)
72. Quaney B, Meyer K, Cornwall MW, McPoil TG. A comparison of the dynamic pedobarograph and Emed systems for measuring dynamic foot pressures. Foot Ankle Int
73. Hood VL, Granat MH, Maxwell DJ, Hasler JP. A new method of using heart rate to represent energy expenditure: the total heart beat index. Arch Phys Med Rehabil
74. Hodges PW, Bui BH. A comparison of computer-based methods for the determination of onset of muscle contraction using electromyography. Electroencephalogr Clin Neurophysiol
75. Basmajian JV, Deluca DJ. Muscles Alive: Their Functions Revealed by Electromyography. 5th ed. Baltimore, MD: Williams & Wilkins; 1985.
76. Kadaba MP, Ramakrishnan HK, Wooten ME, et al. Repeatability of kinematic, kinetic and electromyographic data in normal adult gait
. J Orthop Res
77. Perry J. Gait
Analysis: Normal and Pathological Function. 1st ed. Thorofare, NJ: Slack. Inc.; 1992.
78. Powers CM, Roa S, Perry J. Knee kinetics in transtibial amputee gait
. Gait Posture
79. American Academy of Orthotists and Prosthetists. Draft literature review and ranking in the P&O profession. 2007.
80. Morris C. Current research in orthotics. In: Condie E, eds. Report of a consensus conference on the orthotic management of stroke patients.
Copenhagen, Denmark: International Society for Prosthetics and Orthotics;2004:269–273.
81. National Institute of Health. What Are the Phases of Clinical Trials?
ed. Washington, DC: U.S. National Library of Medicine; 2005.
82. Fatone S. A review of the literature pertaining to KAFOs and HKAFOs for ambulation. J Prosthet Orthot
83. Condie E. Report of a consensus conference on the orthotic management of stroke patients. 1st ed. Copenhagen, Denmark: International Society for Prosthetics and Orthotics; 2004.
APPENDIX 1: LITERATURE REVIEW AND RANKING IN THE P&O PROFESSION
Identifying and classifying the design of a published study is a critical component in the systematic review of literature. To standardize this process for Academy-sponsored State-of-the-Science Conferences (SSCs), the following draft study design classification scale (Appendix Table 1) and associated descriptions are recommended to classify the study type when performing a literature review.
Structured reviews are the methodological collection, analysis, and presentation of information from multiple sources. The analysis of the collected information may be statistical or descriptive in nature, which will identify the structured review as a meta-analysis (R1) or a systematic review (R2).
A statistical analysis that combines the results from multiple studies. Meta-analyses adhere to a structured and appropriate procedure for identifying, including, and analyzing data found in a body of literature.
SYSTEMATIC REVIEW (S2)
A comprehensive methodological review and critical appraisal of literature obtained from multiple sources. Systematic reviews adhere to a structured and appropriate procedure for gathering, selecting, evaluating, and reporting the evidence found in a body of literature.
Experimental trials are prospective research studies that include one or more subjects, a control, one or more interventions, and data collected at known times. In an experimental trial, interventions are applied by the researchers. The number of subjects, assignment of the subject(s) and control(s), and the frequency of data collection will identify an experimental trial as a randomized control trial (E1), controlled trial (E2), interrupted time series trial (E3), single subject experimental trial (E4), or a controlled before-and-after trial (E5).
RANDOMIZED CONTROLLED TRIAL (E1)
A prospective experimental study in which subjects are randomly assigned to either a control or intervention group. Outcome measures are assessed after an appropriate follow-up time, and results are compared between the control and intervention groups.
CONTROLLED TRIAL (E2)
A prospective experimental study in which subjects are nonrandomly assigned to either a control or intervention group. Outcome measures are assessed after an appropriate follow-up time, and results are compared between the control and intervention groups.
INTERRUPTED TIME SERIES TRIAL (E3)
A prospective experimental study in which multiple subjects are assigned only to an intervention group. No control group is formed; instead, subjects serve as their own control. Subjects are evaluated multiple times before and after one or more interventions. Repeated outcome measures are assessed at known intervals, and results are compared between the studied conditions.
SINGLE-SUBJECT EXPERIMENTAL TRIAL (E4)
A prospective experimental study in which one subject is given one or more interventions. The subject serves as his/her own control. The subject is evaluated multiple times before and after each intervention. Repeated outcome measures are assessed at known intervals and results are compared between the studied conditions.
CONTROLLED BEFORE-AND-AFTER TRIAL (E5)
A prospective experimental study in which one or more subjects are assigned to an intervention group. No control group is formed; instead, subjects serve as their own control. Subjects are evaluated once before and once after each intervention. Outcome measures are assessed after an appropriate follow-up time, and results are compared between the studied conditions.
Observational studies include one or more subjects evaluated at a moment in or over a period of time. In observational studies, interventions are not applied by the researchers. Instead, outcomes and influencing factors are observed in order to draw correlations between them. The timing (prospective or retrospective), number of subjects, assignment of subject(s) and control(s), frequency of data collection, and type of data collected will identify an observational study as a cohort study (O1), case-control study (O2), cross-sectional study (O3), qualitative method study (O4), case series (O5), or case study (O6).
COHORT STUDY (O1)
A prospective, observational study of subjects that may develop a specific condition. Subjects may be subdivided into groups based on exposure to factors that may influence occurrence of the condition. Incidence of the condition is assessed after an appropriate follow-up time (typically long-term) and subjects are compared to predict influential factors.
CASE-CONTROL STUDY (O2)
A retrospective, observational study in which a subject group with an existing condition is compared to a similar, but different, subject group that does not have that condition. Outcome measures are obtained from subject histories and used to evaluate the relationships between the outcome measures and the developed condition.
CROSS-SECTIONAL STUDY (O3)
A descriptive, observational study in which a subject group is evaluated at one point in time to assess the relationship between outcome measures and a condition.
QUALITATIVE STUDY (O4)
A descriptive, observational study in which a subject group is evaluated through subjective, open-ended questions and interview techniques.
CASE SERIES (O5)
A descriptive, observational study of the diagnosis, prognosis, treatment, and/or outcome of a subject group with the same (or similar aspects of a) condition.
CASE STUDY (O6)
A descriptive, observational study of the diagnosis, prognosis, treatment, and/or outcome of a single subject.
Expert opinions are peer-reviewed descriptive documents by acknowledged experts. The extent of agreement and synthesis of results will identify an expert opinion as a group consensus (X1) or an individual opinion (X2).
GROUP CONSENSUS (X1)
A peer-reviewed, descriptive synthesis of the results from a conference with multiple experts in a particular topic area. This may also include unstructured literature reviews that were not conducted with a comprehensive methodology consistent with a systematic review (R2).
INDIVIDUAL OPINION (X2)
A peer-reviewed descriptive document by one or more recognized experts in a particular topic area.
Keywords:© 2007 American Academy of Orthotists & Prosthetists
amputation; biomechanics; gait; partial foot