Extracts of the racine of Devil's claw (Harpagophytum procumbens) have become popular for degenerative conditions of the musculoskeletal system . Different manufacturers use different procedures to produce different extracts in which the quantity of the marker substance, harpagoside, does not need to be declared (in contrast with the requirements of the German Pharmacopoeia for the raw material, for which the harpagoside content must exceed 1% [2,3]. In the extract WS 1531 from Schwabe (Karlsruhe, Germany), the raw material is concentrated 6-9 times, so that a 200-mg tablet contains 17 mg of harpagoside: this is released quickly, is stable in gastric and intestinal fluids  and can be detected in the plasma . Harpagophytum extract WS 1531 produces a dose-dependent inhibition of both the prostaglandin and leukotriene pathways of arachidonic acid metabolism. But harpagoside is only one of a number of compounds in the extract, some of which are analgesic, whereas some may be anti-analgesic .
If Harpagophytum extract is to continue to contribute respectably to advances in therapeutics, it must do so by embracing the formalities, technicalities and niceties of the randomized, controlled trial. The analgesic effectiveness of a daily dose of another, less concentrated Harpagophytum extract (equivalent to about 50 mg of harpagoside) has been investigated previously in a randomized, double-blind, placebocontrolled pilot study of 118 patients suffering from low back pain with or without radiation to the legs . There have also been two post-marketing surveillance studies using a lower dose of harpagoside for less severe, non-radiating pain, one on 102 patients  and the other on 100 patients .
Chrubasik et al. used a modification of the Arhus index  as one of their outcome measures. The original index was designed for quick and easy use in routine clinical practice and is one of a number of standardized scoring systems that attempt to record the many facets of the clinical entity of 'low back pain'. The original score has a maximum of 130 possible points - 60 for pain, 30 for disability and 40 for physical impairment. The 60 pain points are divided equally between back and leg pain, with up to 10 points at each location (1-10 cm visual analogue scale) - 10 for pain at the time of examination (current pain), 10 for the average and 10 for the worst pain over the preceding 2 weeks. Disability is assessed by a structured series of 15 questions (including one on analgesic use) about how much the pain affects everyday activities, emotional and social interactions and employment prospects (never = zero, at times = 1 and yes = 2). Physical impairment comprises up to 10 points each for endurance of the back muscles, mobility of the back, mobility of the patient as a whole and requirements for analgesics. The Arhus score appears to be unidimensional according to the tests that have been applied, and the weightings of its constituent items have been validated , but little has been reported as to how it might change with treatment. The change in Arhus index in the study by Chrubasik et al. just failed to reach statistical significance at the 5% level, but the proportion of pain-free patients was significantly more by the end of 4 weeks of treatment with Harpagophytum (20%) than with placebo (5%). These proportions are similar to those obtained after 4-6 weeks of more established treatments such as oral non-steroidal anti-inflammatory drugs (NSAIDs), physical exercises and paravertebral injections of hyperosmolar glucose solutions (.
Now that some effectiveness of the extracts of Harpagophytum has been demonstrated in studies each using a single dose level, it is important to investigate the form of any possible dose dependency (particularly in view of the possible presence of anti-analgesic substances in some extracts). This study used the general investigational format of Chrubasik et al. in this next stage of the evaluation, but studies Harpagophytum WS 1532.
General plan of the study
The study was approved by the Human Ethics Committee of the University of Heidelberg and was a threegroup, randomized, double-blind comparison of the effectiveness and possible adverse effects of two dose rates of Harpagophytum WS 1532 extract (600 and 1200 daily containing 50 and 100 mg of harpagoside respectively) and a placebo over 4 weeks of treatment in patients suffering from low back pain with or without radiation to the legs. The dose range was chosen to reflect the upper end of the amount of hapagoside that is known to be used (ESCOP monograph), in order to maximize the estimates of effectiveness and the chances of detecting important adverse effects. The prospectively chosen principal outcome measure was the proportion of patients free of pain without any other analgesics for at least 5 days in the fourth week. Subsidiary outcome measures were to be the change in the Arhus Low Back Pain Index  relative to baseline and the consumption of the only allowable rescue medication, tramadol.
Numbers and selection of patients
The estimated minimum number of patients required in each group was 60, based on conventional calculations of 90% power for detecting differences in proportions of pain-free patients between 5% in patients receiving placebo and 15% and 25%, respectively, in those receiving daily doses of 600 and 1200 mg of the Harpagophytum extract, when 95% confidence was required to reject the null hypothesis that there would be no trend detectable by a one-sided, Cochrane-Armitage test.
Newspaper and radio advertisements were used to recruit suitable patients between October 1996 and May 1997 from the Regierungsbezirk Karlsruhe within the county of Baden-Wuertemberg. The selection criteria were: (i) age 18-75 years; (ii) a background of at least 6 months susceptibility to low back pain that was not readily attributable to identifiable causes, such as disc prolapse, hip disease, spondylolisthesis, osteomalacia or inflammatory arthritis; and (iii) a current exacerbation of their complaint that was affecting both rest and movement, was giving rise to pain greater than 5 on a 1-10 visual analogue scale and was expected to require at least 4 weeks of symptomatic treatment. Exclusion criteria were: (i) current or recent (within 30 days) participation in any other clinical study; (ii) serious organic illness affecting any organ system; (iii) a history of drug or alcohol abuse or requirement for psychotherapeutic agents; (iv) pregnancy, actual or possible (unreliable contraceptive practice), or lactation; (v) known allergy to any of the proposed trial medications; and (vi) difficulties with language or anticipated co-operation. An account of the plan and an invitation to participate (conditional on the results of an assessment in the clinic) was given to patients who met these criteria. Almost all of the suitable patients accepted the definitive invitation, and the first 197 who gave written consent to participate were entered into the study.
The patients were assigned to three groups by a stratified random allocation (ETV program) based on the sequence in which they gave informed consent: the stratification was according to the duration of the exacerbation (less than a week, between 1 and 6 weeks and more than 6 weeks), the intensity of their pain (less or more than 7 on the VAS scale of 1-10) and whether or not it radiated to the legs.
Conduct of the trial
To record their baseline state, the patients completed a standardized questionnaire about their general health, daily activity and the particular characteristics of their pain. They then underwent a clinical examination with recording of vital signs and gave a venous blood sample for baseline measurements of the conventional biochemical and haematological indices of organ system function. They also underwent questioning and examination for the completion of the Arhus score. This was modified for this study by the exclusion of the items relating to analgesic medications, because the patients were to be allowed only tramadol as rescue medication, so that tramadol consumption could be examined as a subsidiary outcome measure or confounder in the study. The modifications gave the score a maximum of 118 - 60 for pain, 28 for disability and 30 for physical impairment.
Trial and rescue medication The patients then received their coded trial medication, one tablet to be taken three times a day for 4 weeks. The tablets were identical in appearance, but the contents differed for the three treatment groups. In the placebo group (P), they contained lactose; in the group to receive a daily dose of 600 mg of Harpagophytum (H600), each tablet contained 200 mg of Harpagophytum WS 1531 extract (17 mg harpagoside); and in the group to receive a daily dose of 1200 mg (H1200), each contained 400 mg of the extract. The coding of the tablets concealed their identity from both the patients and the investigators. Patients were allowed to supplement their trial medication with tramadol liquid (2.5 mg mL−1) in doses from 50 to 400 mg per day.
Monitoring and completion Each day for the next 4 weeks, the patients were contacted by telephone. This enabled the clinical investigator to obtain daily verbal ratings of their pain (0 = none, 1 = mild, 2 = moderate, 3 = severe and 4 = unbearable) and requirement for tramadol as rescue medication and to encourage them to keep taking the study medication. At the end of the 4-week study period, the Arhus scoring, the vital signs and laboratory measurements were repeated in the same format as at baseline.
The principal outcome measure was the proportion of patients who responded to treatment by being pain free without tramadol for at least 5 days during the last week. The principal analysis of effectiveness was carried out on the principle of intention to treat, with patients who did not complete the study after randomization (drop-outs) being included as non-responders'. A one-sided, Cochrane-Armitage test was used to test the null hypothesis that there was no monotonic increase in the proportion of responders with increasing daily dose of Harpagophytum extract. Sensitivity analyses were carried out with alternative principal outcome variables in which the requirement for a responder was to remain pain free for only 3 or 4 days. Further sensitivity analyses were carried out with drop-outs excluded, on the subsets of patients who had or had not suffered more than 6 weeks of pain, who did or did not have initial pain greater than 7 on the VAS, who did or did not have pain radiating into the legs and who did or did not have a neurological deficit at the start of the trial.
The change in Arhus index relative to baseline (initial minus final value divided by initial value) was examined as a subsidiary analysis. The first analyses were by intention to treat with drop-outs awarded final scores equal to the initial scores. The relative changes in the component parts of the Arhus index were also examined in this way. Further sensitivity analyses were also conducted on the Arhus-based indices with drop-outs awarded maximum final pain and disability scores and with the same subgroups as for the principal analysis. Inferential testing was by Kruskal-Wallis test to investigate differences in relative change between groups, and by the one-sided, Jonkheere-Terpstra test for monotonic trend of effect with dose. To examine for possible confounding effects of age, gender, severity, duration and radiation of pain and tramadol consumption, these explanations were offered along with dummy variables for H600 and H1200 to multiple regressions in which the dependent variables were either the combined pain component of the Arhus index (maximum 60 points as above) or the current low back pain component (maximum 10 points as above).
These analyses and supporting tabulations and investigations were undertaken using the procedures available in the Statistical Analysis System software package (SAS Institute, Cary, NC, USA). Categorical data were examined in contingency tables, with inferential testing by the chi-squared test. Ordinal or continuously distributed data were summarized as median and quartiles, with inferential testing by Kruskal-Wallis or Mann-Whitney-Wilcoxon tests (paired as appropriate for differences between baseline and the end of the study). Spearman rank correlation was used to relate verbal rating scores with verbal rating of pain. The confidence level for rejecting the null hypothesis was taken as 95% (P< 0.05).
Characteristics on entry to the study
Table 1 lists some of the characteristics of the 197 patients who entered trial by treatment group, 66 in group P, 65 in group H600 and 66 in group H1200. The groups were clearly matched for age, height, weight and gender as well as for the major features and associations of their presenting pain. An additional set of about 100 features was recorded including indicators of general health, professional and employment status, sporting activity, further details of the low back pain, results of attempted treatments and the common screening laboratory measurements. The tabulation of these by treatment group has been placed on a website (http//www.rzuser.uni-heidelberg.de/∼cn6/harpago/). There were only four features around 120 in which the observed intergroup differences would have reached significance in single isolated comparisons.
In over 80% of the patients, susceptibility to low back pain had persisted for more than 6 years, and the current exacerbation had lasted for more than 3 months. About 80% of patients had suffered physical impairment for more than 14 days in the previous 6 months, and their condition had affected their work in up to 50% of cases. There was pain in at least one other site (shoulder, arm, neck, groin, buttock, leg) in most of the patients, and it was exacerbated by physical activity in up to 86% of them. Specific causes were generally difficult to identify, although a variety of possible causes had been advanced by the attending physicians, the most common of which were arthrosis, spondylosis and scoliosis. Pain tended to be greatest in the morning and evening. Almost all patients had previously resorted to some sort of treatment for the chronic condition, either regularly or irregularly. These had achieved variable success for an unspecified period but had been discontinued for a variety of reasons, some of which may well have been financial. About half of the patients in all groups had a demonstrable neurological deficit.
The trial was completed by 183 patients: of the 14 drop-outs, three were in group P (two with unbearable pain, one with cardiovascular complaints), eight in group H600 (virus infections and psychosomatic symptoms in two, and one case each of mucorrhoea, skin reaction, failure to comply with the medication and unbearable pain) and three in group H1200 (one with an allergic reaction, one with unbearable pain and one who did not attend the final examination).
Principal analysis of effectiveness
The number of patients who were pain free without tramadol for at least 5 days in the fourth week of treatment was three in group P, six in group H600 and 10 in group H1200 (P = 0.027, Cochrane-Armitage). In 13 of these 19 patients, pain had been present for less than 6 weeks before the start of the trial (two in group P, three in H600 and eight in H1200). Five of these responders (all in group H1200) had a neurological deficit at the start of the trial. The sensitivity analyses with the alternative principal outcome variables tended to confirm the results of the principal analysis (Table 2). The subgroup sensitivity analyses confirmed a significant monotonic increase in effect with dose only in the subgroups with more severe and radiating pain and neurological deficit.
The relative changes in the Arhus index and its components
The median relative change in the overall Arhus index was around 20% in all groups. The change in each group was very significantly different from zero but not significantly different from the changes in the other groups (Table 3). Of the Arhus components, the current pain at the time of examination shows the greatest differences between treatment groups: the changes are larger in both active treatment groups than in the placebo group, but the median change in the H600 group is greater than in the H1200 group. Table 3 also summarizes the median relative changes in current pain for sensitivity analyses in which drop-outs were omitted or assigned maximum final pain, and for a number of subgroup analyses. The relative changes in the H600 group are consistently larger than those in the H1200 group, and the P-values for the Kruskal-Wallis test (which simply examines for differences between groups) are consistently closer to significance than the P-values for the Jonkheere-Terpstra test (which examines for a monotonic increase in effect with dose).
The multiple regression analyses on the relative change in the pain components of the Arhus index did not identify prior duration of exacerbation, initial intensity or radiation of pain as confounders, only tramadol consumption, i.e. patients who consumed more tramadol also had less relief of pain. No significant effect could be identified for either dose of harpagophytum extract or interactions with other explanations of interest, but the coefficient for H600 was consistently larger than that for H1200.
Verbal ratings of pain and tramadol consumption
Of the 5516 patient-days on which patients could have supplied verbal ratings of pain if there had been no drop-outs, ratings were obtained for 5315. Table 4 shows how these patient-days were distributed among the five ratings of pain and associated numbers relating to tramadol use. The proportions of patient-days on which tramadol was used increased with the severity of pain, but was still less than 20% for patient-days on which pain was severe and less than 45% for patient-days on which it was unbearable. For the patient-days in which tramadol was used, the doses were far less than the maximum daily allowance, the median being about one-third on the patient-days with severe pain and two-thirds on the patient-days with unbearable pain. Thus, the pain-related variation in dose of tramadol per patient-day was produced more by the pain-related variation in proportion of patient-days with tramadol rather than the pain-related variation in dose per patient-day on which tramadol was used.
Table 5 shows the distribution of patients reporting different ratings of pain as their maximum for each of the 4 weeks. The proportion reporting severe or unbearable pain fell progressively from around 61% overall in the first week to around 33% in the third week: the fall was by 24% in the placebo group, 30% in the H600 group and 32% in the H1200 group. Table 5 also shows the number of patients who took tramadol in each week. The verbal ratings of pain averaged over the last week correlated reasonably well with the current pain component of the final Arhus assessment (Spearman rank correlation coefficient 0.768). Tramadol consumption showed some associations with severity of pain, but the associations were weak.
Clinical and laboratory indications of adverse effects
The details of the exploration for adverse effects are tabulated on the website (http//www.rzuser.uni-heidelberg.de/∼cn6/harpago/). The reports of adverse reactions were scrutinized by an independent investigator. The suspicions of Harpagophytum-related adverse reactions were generally weak (the strongest being in eight reports of mild gastrointestinal upset, four in each dose group, whereas only one patient in the placebo group reported any such effect).
Musculoskeletal pain and especially low back pain are the most frequent pain syndromes world-wide [11-16]. The health-related budgets of the western industrial countries carry a substantial burden from the consequences of low back pain - absenteeism, hospital care, diagnosis and therapy . The main causes of low back pain are non-specific and degenerative changes . In patients with a susceptibility to low back pain, the duration of an exacerbation determines both the treatment and its success. Whereas bed rest for less than 2 days, paracetamol and/or NSAIDs and treatment in back schools tend to be satisfactory for exacerbations lasting no more than 6 weeks [18, 19], longer lasting episodes of pain do not respond to oral analgesics. Of the wide range of treatments that have been tried (including oral analgesics, manipulation, back school, heat or cold, supervised exercises, facet injections, stretching and surgery), none is convincingly effective for patients who have been suffering for longer than 3 months .
A meta-analysis that included 26 randomized trials suggested that NSAIDs might only be effective for short-term symptomatic relief in patients with uncomplicated low back pain, but are less effective or ineffective in patients with low back pain, sciata and nerve root symptoms . Two weeks of treatment with 75 mg day−1 indomethacin produced no increase in the number of pain-free patients compared with placebo treatment , but no studies have yet been reported of the effects of 4 weeks of treatment. Piroxicam (20 mg day−1) for 4 weeks has been compared with placebo, without producing any convincing difference in visual analogue scores for pain .
This bleak outlook has not been changed by the results of our study in which over 80% of the patients had suffered from such prolonged exacerbations. Overall, 10% of our patients responded according to our primary outcome measure, most of the responders having had exacerbations that lasted less than 6 weeks. One could therefore argue that at least half of the patients who would have responded to anything responded to the treatment (including placebo) that they received in our trial. It should be remembered that the placebo group in this trial did not receive no treatment. The daily contact and interest of the investigators probably provided its own distinct psychotherapeutic benefits: admittedly, these benefits seem more appreciable in terms of the subsidiary measures than in the primary outcome - the median relative change of about 25% in the current pain component of the Arhus index and the reduction from about 60% to about 35% in the proportions of patients whose maximum pain for the week was severe and unbearable. It is likely that the placebo without the daily follow-up would have been less effective , but it is equally likely that this would also have been true for the Harpagophytum extract. Suffering from low back pain is heavily influenced by psychological and social factors , which is why a holistic, individualized and multidisciplinary treatment plan is the most appropriate approach to the constellation of problems that is faced by each patient [25-28].
The principal analysis of the primary outcome measure suggested a monotonic increase in effectiveness with dose. Although the Cochrane-Armitage test does not necessarily distinguish between the effects of the two doses, the estimate of effect size in the principal analysis did show a graded effect with dose. The effect is certainly not large on the whole cross-section of patients who entered the trial (as opposed to the number with exacerbations of shorter duration). Under the conditions of our study, one would, for instance, have to treat 21 patients with 600 mg of Harpagophytum extract per day to produce one extra responder compared with placebo, but fewer than 10 patients would be needed for the same benefit with 1200 mg of the extract.
In studies that incorporate alternative analyses to the primary analysis on the principal outcome variable, the objective is to offer some tentative confirmation or qualification of the results of the primary analysis. In this study, the conclusions of the primary analysis tended to be confirmed by the sensitivity analyses in which the responders were designated as having fewer than 5 days of freedom from pain in the last week. In the subgroup analyses, the estimates of greater effectiveness of the larger dose of Harpagophytum were seen only in the subgroups with more severe pain, with pain that radiated to the leg and was associated with neurological deficit. The 600 mg dose of Harpagophytum WS 1531 seemed to be ineffective in these patients, as was an equivalent dose of harpagoside in a different extract in the study of Chrubasik et al. (which did not examine for neurological deficit).
The analyses based on the current pain of the Arhus index showed only partial consistency with the above picture. In none of these analysis did duration of exacerbation seem to affect degree of response, as was so strongly suggested in the primary analysis of the principal outcome. In the subgroup sensitivity analyses (Table 3), the Kruskal-Wallis P-values (which mostly pointed to differences between the Harpagophytum groups and the placebo group without examining for monotonic increase in effect with dose) seemed to suggest a significant effect of Harpagophytum only in patients with non-radiating and less severe pain. However, the degree of response was consistently less with the higher dose of Harpagophytum, the decrement being least (giving a result least different from the trend in the primary analysis and a Jonkheere-Terpstra P-value closest to significance) in the patients without radiating pain.
The suggestion of a lesser response with the higher dose of Harpagophytum is weakly supported by the observation in Table 5 that fewer patients in group H600 were taking tramadol from the second to the fourth week than in group H1200. This could be attributed to bias from the greater number of drop-outs in group H600, but the difference started to appear by the end of the second week, by which time there had been only two drop-outs, at least one of whom had taken no tramadol. As in the study of Chrubasik et al., severe and unbearable pain by no means guaranteed that patients would take tramadol or make use of the maximum allowable dose, but the increased likelihood of tramadol consumption on days on which there was more pain accords with expectations that tramadol consumption should reflect the degree of pain to some extent. A possible reduction in effect with increase in dose of the extract may be noteworthy in view of the possible presence of anti-analgesic substances in Harpagophytum WS 1531.
The numbers of responders' in the primary analysis was small, in keeping with the large proportion of patients who had been suffering from prolonged exacerbations. There were also worrying suggestions of inconsistency between the primary and secondary analyses in the direction of any dose-related effect of Harpagophytum extract WS 1531 and in the subgroups of patients to whom they apply. These must raise concerns that unrecognized bias may have by-passed the randomization and produced the apparent increase in effect with dose that was seen in the primary analysis. The alternative is to believe all analyses and to make the interesting, but presently unsupportable, postulation that the different types of outcome variable are reflecting qualitatively different facets of the low back pain syndrome.
Irrespective of whether there is a difference in effect between the two doses studied, there seems to be some consistency in the suggestion that Harpagophytum WS1531 at one dose or another did have some effect, at least in the expected subgroups (shorter exacerbations, less pain, no radiation). This effect was additional to that resulting from the extra attention received by patients in all groups including the placebo group. The overall effect was certainly not large, but was achieved with no strong suspicion of dose-related adverse effects.
Thus, a reasonable conclusion might be that Harpagophytum can probably help many of those who might also be helped by bed rest, paracetamol or NSAIDs, or manipulation and back school. It may well do so with fewer side-effects than NSAID treatment (which are especially troublesome in the elderly [29-32]), at a cost that is certainly not excessive [33-35] and may even be as little as a third of the cost of conventional treatment .
The authors sincerely thank Dr Andy Black, University of Bristol, for help with the manuscript.
1 Chrubasik S, Wink M. Zur pharmakologischen Wirkung der Teufelskralle (Harpagophytum procumbens
). Forsch Komplementärmed
2 Chrubasik S, Sporer F, Wink M. Zum Harpagosidgehalt verschiedener Trocken extraktpulver aus Harpagophytum procumbens. Forsch Komplementärmed
3 Chrubasik S, Sporer F, Wink M. Zum Wirkstoffgehalt in Teezubereitungen aus Harpagophytum procumbens. Forsch Komplementärmed
4 Chrubasik S, Sporer F, Dillmann-Marschner R, Friedmann A, Wink M. Physicochemical properties of harpagoside and its in vitro
release from Harpagophytum procumbens
extract tablets. Phytomedicine
1999; (in press).
5 Loew D, Schuster O, Möllerfeld J. Stabilität und biopharmazeutische Qualität, Voraussetzung für Bioverfügbarkeit und Wirksamkeit von Harpagophytum procumbens.
In: Loew, D, Rietbrock, N, eds. Phytopharmaka II.
Darmstadt: Steinkopf-Verlag, 1996; 88-93.
6 Tippler B, Syrovets T, Loew D, Simmet Th. Harpagophytum procumbens:
Wirkung von Extrakten auf die Eicosanoidbiosynthese in lonophor A23187-stimuliertem menschlichem Vollblut. In: Loew, D, Rietbrock, N, eds. Phytopharmaka II. Forschung und klinische Anwendung.
Darmstadt: Steinkopf-Verlag, 1996: 95-100.
7 Chrubasik S, Zimpfer Ch, Schutt U, Ziegler R. Effectiveness of Harpagophytum procumbens
in treatment of acute low back pain
8 Chrubasik S, Schmidt A, Junck H, Pfisterer M. Wirksamkeit und Wirtschaftlichkeit von Teufelskrallenwurzelextrakt bei Rückenschmerzen: Erste Ergebnisse einer Kohortenstudie. Forsch Komplementärmed
9 Schmelz H, Hämmerle D, Springorum HW. Analgetische Wirkung eines Teufelskrallenwurzel-Extraktes bei verschiedenen chronisch-degenerativen Gelenkerkrankungen. In: Chrubasik, S, Wink, M, eds. Rheumatherapie mit Phytopharmaka.
Stuttgart: Hippokrates-Verlag, 1997: 86-89.
10 Manniche C, Asmussen K, Lauritsen B, Vinterberg H, Kreiner S, Jordan A. Low back rating scale: validation of a tool for assessment of low back pain
11 Bowsher D, Rigge M, Sopp L. Prevalence of chronic pain in the British population: a telephone survey of 1037 households. Pain Clinic
12 Brattberg G, Thorslund M, Wikman A. The prevalence of pain in a general population. The results of a postal survey in a county of Sweden. Pain
13 Chrubasik S, Junck H, Zappe HA, Stutzke O. A survey on pain complements and health care utilisation in a German population sample. Eur J Anaesthesiol
14 Crook J, Tunks E, Rideout E, Browne G. Epidemiologic comparison of persistent pain sufferers in a specialty pain clinic and in the community. Arch Phys Med Rehabil
15 Rekola KE, Keinanen-Kiukaanniemi S, Takala J. Use of primary health services in sparsely populated country districts by patients with musculoskeletal symptoms: consultations with a physician. J Epidemiol Commun Health
16 Von Korff M, Dworkin SF, Le Resche L, Kruger A. An epidemiologic comparison of pain complaints. Pain
17 Van Tulder MW, Koes BW, Bouter LM. A cost-of-illness study of back pain in the Netherlands. Pain
18 Deyo RA. Conservative therapy of low back pain
- distinguishing useful from useless therapy. JAMA
19 Nachemson AL. Newest knowledge of low back pain
. Clin Orthop
20 Koes BW, Scholten RJPM, Mens JMA, Bouter LM. Efficacy of non-steroidal anti-inflammatory drugs for low back pain
: a systematic review of randomised clinical trials. Ann Rheum Dis
21 Goldie I. A clinical trial with indomethacin (Indomee) in low back pain
and sciata. Acta Orthop Scand
22 Weber H, Holme I, Amlie E. The natural course of acute sciata with nerve root symptoms in a double-blind placebo-controlled trial evaluating the effect of piroxicam. Spine
23 Rene J, Weinberger M, Mazzuca SA, Brandt KD, Katz BP. Reduction of joint pain in patients with knee osteoarthritis who have received monthly telephone calls from lay personnel and whose medical treatment regimens have remained stable. Arth Rheum
24 Waddel G. Biopsychosocial analysis of low back pain
. Baillières Clin Rheumatol
25 Flor H, Fydrich T, Turk DC. Efficacy of multidisciplinary pain treatment centers: a meta-analytic review. Pain
26 Kleinke CL, Spangler AS. Predicting treatment outcome of chronic back pain patients in a multidisciplinary pain clinic: methodological issues and treatment implications. Pain
27 Lanes TC, Gauron EF, Spratt KF, Wernimont TJ, Found EM, Weinstein JN. Long-term follow-up of patients with chronic back pain treated in a multidisciplinary rehabilitation program. Spine
28 Love WA, Peck CL. The MMPI and psychological factors in chronic low back pain
: a review. Pain
29 Allison MC, Howatson AG, Torrance CJ, Lee FD, Russel RI. Gastrointestinal damage associated with the use of nonsteroidal antiinflammatory drugs. N Engl J Med
30 Fries JF. Assessing and understanding patient risk. Scand J Rheumatol
31 Griffin MR, Piper J, Daugherty JR, Snowdon M, Ray W. Nonsteroidal anti-inflammatory drug use and increased risk for peptic ulcer disease in elderly persons. Ann Int Med
32 Langman MJS, Weil J, Wainwright P, Lawson DH, Rawlins MD, Logan RFA, Murphy M, Vessey MP, Colin-Jones DG. Risks of bleeding peptic ulcer associated with individual non-steroidal anti-inflammatory drugs. Lancet
33 Bloom BS. Direct medical costs of disease and gastrointestinal side-effects during treatment for arthritis. Am J Med
34 Greene M, Winickhoff RN. Cost-conscious prescribing of nonsteroidal anti-inflammatory drugs for adults with arthritis. Arch Int Med
35 Levine JS. Mistoprostol and nonsteroidal anti-inflammatory drugs: a tale of effects, outcomes, and costs. Ann Int Med