Secondary Logo

Journal Logo

Pain assessment in advanced dementia. Validity of the German PAINAD—a prospective double-blind randomised placebo-controlled trial

Lukas, Alberta,b,*; Hagg-Grün, Ulricha; Mayer, Benjaminc; Fischer, Thomasd; Schuler, Matthiase

doi: 10.1097/j.pain.0000000000001430
Research Paper
Global Year 2019

Pain in combination with dementia is a common condition that makes pain recognition significantly more difficult. This results in undertreatment of pain in those suffering from dementia. The Pain Assessment in Advanced Dementia (PAINAD) scale currently represents one of the best approaches to pain detection in dementia. In a pilot study, strong inter-rater and retest reliability of the German version (PAINAD-G) was proven. However, the available data concerning the validity of this instrument were insufficient. The aim of the study was to validate the PAINAD-G scale by a double-blind randomised placebo-controlled trial in people with advanced dementia expected to be in pain. A second aim was to examine whether other observational tools (BISAD = Observation Instrument for Assessing Pain in the Elderly with Dementia) (German: Beobachtungsintrument für das Schmerzassessment bei alten Menschen mit Demenz, Checklist of Nonverbal Pain Indicators, Algoplus) were also able to demonstrate a significant difference between the study groups. Surprisingly, the study revealed no difference in “pain reduction” between those treated by oxycodone compared with those treated by placebo. Equally, none of the other 3 observational tools were able to demonstrate a significant difference between the study groups. However, correlations among the 4 observational tools were mostly moderate to high. A number of possible reasons for this observation, such as difficulties regarding sensitivity to change/responsiveness, consistence of the fundamental construct, influence of the early onset study, and efficacy of the analgesic in advanced dementia are discussed.

Pain treatment compared with placebo did not lead to significantly lower pain behaviour scores in people with advanced dementia, expected to be in pain.

aAGAPLESION Bethesda Clinic, Competence Centre of Geriatrics and Aging Research, University of Ulm, Ulm, Germany

bMalteser Hospital Bonn, Geriatric Centre, Academic Teaching Hospital, University of Bonn, Bonn, Germany

cInstitute of Epidemiology and Medical Biometry, University of Ulm, Ulm, Germany

dEvangelische Hochschule Dresden, University of Applied Sciences for Social Work, Education and Nursing, Dresden, Germany

eDiakonissenkrankenhaus Mannheim, Clinic for Geriatric Medicine, Academic Teaching Hospital, University of Mannheim and Heidelberg, Mannheim, Germany

Corresponding author. Address: Malteser Hospital Bonn/Rhein-Sieg Geriatric Department, Teaching Hospital of the University of Bonn, Von-Hompesch-Straße 1, D-53123 Bonn, Germany. Tel.: +49 228 6481-512; fax: +49 228 6481-891. E-mail address: (A. Lukas).

Sponsorships or competing interests that may be relevant to content are disclosed at the end of this article.

Received May 17, 2018

Received in revised form October 02, 2018

Accepted October 18, 2018

Back to Top | Article Outline

1. Introduction

Pain and dementia are very common in nursing home residents.28 It is suggested that up to 50% suffer from both conditions simultaneously.22 Pain is often under-reported and inadequately treated due to impaired verbal communication and insufficient instruments for pain assessment in patients with dementia.12 Numerous attempts have been made to develop instruments to detect pain in patients with advanced dementia. Up to now, however, validity and reliability have not been established to a satisfactory degree12,19 in most cases. Although initial comparative studies have already been conducted20,21 to guide best-practice recommendations, more comparative studies are required to help identify those instruments, which prove to be the most effective in the detection of pain as well as being the most sensitive in the detection of response to treatment.12,15

One of the instruments that has been recommended for further investigation13 is “Pain Assessment in Advanced Dementia” (PAINAD),34 which is also available in a German version (PAINAD-G or in German: BESD—Beurteilung von Schmerz bei Demenz).1,30 Construct and concurrent validity for the PAINAD have been demonstrated as well as high levels of inter-rater and retest reliability.27,30,34 Internal consistency for the PAINAD was reported to be moderate to high. For the German PAINAD-G construct, validity has been investigated in a prospective one-factorial (=measurement points) observation study, which involved repeated measurements that demonstrated reduced pain behaviours as assessed with the PAINAD-G after administration of analgesia to patients with dementia and suspected pain.1 However, this study had a high risk of bias because it was neither blind nor randomised, and the choice of analgesic regime was also not standardised.

Although the PAINAD was designed for use with patients with advanced dementia, validity has up to now been tested on patients with mild to moderate dementia who are still able to provide self-reports as the gold standard of pain assessment but has excluded patients with severe dementia.27 This leaves the question still unanswered of whether the PAINAD is a valid tool for pain assessment in nonverbal patients with advanced dementia.

Therefore, the primary objective of this study was to test the construct validity of the PAINAD-G for pain assessment in nonverbal patients with advanced dementia by a double-blind randomised placebo-controlled trial.

The secondary objective was to investigate how results of the PAINAD-G and 3 other observational pain tools for patients with dementia are correlated. A high correlation between assessments that are supposed to measure similar constructs, namely pain in people with dementia, should support their validity and can be seen as additional evidence.

Back to Top | Article Outline

2. Methods

2.1. Study design

This study was a 2-factorial, double-blind randomised placebo-controlled trial, stratified according to the participating study centres. Factor I consisted of the analgesic intervention and factor II included the measurement points. Participants were randomised to factor I receiving either oxycodone (10 mg) or a placebo. Oxycodone has been a well-established analgesic for many years, having a favourable effect/side effect profile and is recommended for use in older persons.11 Factor II was varied 3 times. The first measurement point was conducted before the intervention (t1); the second measurement followed 3 hours after the first intervention (t2), and the third 6 hours after the first intervention (t3). Immediately after t1, patients received—blind for the patient and study team—either oxycodone or placebo and vice versa (still blinded) after t2, to treat all patients' very likely suffering with pain at least at t2. Because of ethical reasons, every study participant should be treated by a verum as soon as possible (also those who received placebo in the first 3 hours). The selected design did not fully comply with a common cross-over scheme (=“quasi cross-over design”). There was no wash-out phase, why analgesic residual effects in the t2 to t3 placebo group had to be assumed and could not be counted as a real placebo phase (Fig. 1). The time interval of 3 hours was determined by the pharmacokinetics of the selected opioid according to its maximum analgesic effect, approximately 3 hours after administration.32 Because of the fact that 38% of the available oxycodone dose is rapidly absorbed with a mean half-life of 37 minutes, first analgesic effects can be already seen within 1 hour.23 Patients' blood pressure, respiratory rate, and alertness were monitored every 2 hours over an 8-hour period after the first intervention, as well as a particular attention being paid to possible side effects of the medications (such as hypotension, respiratory depression, sedation, dizziness, and nausea).

Figure 1

Figure 1

To eliminate observation bias, we used randomisation and blinding. Tablets were packed in consecutively numbered blisters, each containing one verum and one placebo tablet enclosed in identical looking capsules. The medication to be taken first was marked as such. Randomisation was performed by an institute of biometrics (computer-generated list), whereas packaging of the capsules was conducted by a university pharmacy. Envelopes containing emergency information were delivered to each study centre. In the event of an emergency, the individual envelope could have been opened, which would have led to the exclusion of this participant. All envelopes were returned unopened to the main centre at the end of the study. Blinding ended after the last patient was treated.

Pain was rated twice per measurement and with all behavioural pain assessment tools included in the study in parallel. The first rating was taken in a situation where the patient was at rest (eg, lying in bed) and the second during movement.12 For ethical reasons, the movement could not be standardised for all patients. However, to ensure consistency for the individual patient, a standardised movement protocol was implemented, which described a range of different movements the assessors could choose based on the patient's situation and needs. The initial type of movement at t1 and any support given was documented and was repeated at any subsequent measurement (t2 and t3).

Data were collected by study nurses. To ensure uniform practice, 1-day training was organised in the main study centre to train assessors in all relevant aspects of the trial. To ensure standardised rating, particularly concerning the observational tools, video examples of people with pain and dementia (own collection) were used for training. An additional emergency phone with direct contact to the principal investigator was set up, to clarify possible upcoming questions and uncertainties.

The study was approved by the Ethics Committees at each study centre, namely the University of Ulm (No. 13/09), University of Mannheim (No. 2009-318N-MA), University of Heidelberg (No. S-313/2009), and the State Chamber of Medicine in Brandenburg (Woltersdorf/Berlin), Germany. All procedures were conducted in accordance with ethical standards relating to human experimentation and in accordance with the Helsinki Declaration. All data were collected initially in strictly pseudonymous form. After completing the study, the reference list of complete names was deleted to transfer the data anonymously. Furthermore, the study was officially registered with DRKS (No. DRKS00000525) and submitted to the WHO (Universal Trial Number [UTN]: U1111-1116-6820).

Back to Top | Article Outline

2.2. Patients

Patients with dementia and probable pain were eligible for participation in this study. Inclusion criteria included a diagnosis of Alzheimer disease (ICD F00.-) or vascular dementia (ICD F01.-) or a combination of both (ICD F01.3). As pain assessment instruments used in the study were developed for patients who were no longer able to verbalize their pain, the patients' inability to verbalize whether or not they were in pain, and their inability to rate their pain intensity using a 5-item verbal rating scale was set as the second selection criterion. Current available evidence suggests that the use of observational scales should be limited to patients who demonstrably cannot reliably verbalise their pain.19 Complementarily, cognitive function was tested using the Mini-Mental State Examination (MMSE)8 to explain the cognitive level of the participants and to exclude patients with mild or no cognitive impairment (MMSE equal or above 20). Patient eligibility for the study was primarily defined by the inability to verbalize pain, not using an MMSE score. Patients also had to be diagnosed with a disease frequently associated with pain (eg, arthrosis or fractures) and reach an initial PAINAD-G score of 4 or higher during movement to suggest presence of pain, according to ROC curve-derived cut-off points by Lukas et al.21

Patients were excluded if they were diagnosed with other forms of dementia or other diseases causing communication impairments (such as stroke or Parkinson disease). To prevent behaviour syndromes based on delirium, participants with probable delirium measured by confusion assessment method14 (CAM) ≥3 were also excluded.

In addition, contraindications against the use of oxycodone led to the exclusion of the patient. Patients receiving analgesic drugs of WHO level II or higher were also not eligible for participation. Because of ethical reasons, patients who were on a standard medication of a WHO level I analgesic had this prescription continued throughout the trial, as was the case with existing long-term medication. However, influence of pre-existing (analgesic) medication was considered minimal because of the relative high PAINAD-G cut-off of ≥4 (according to Lukas et al.21), which indicates substantial expressions of pain.

To prevent pharmacological effects other than through the study drug, no additional central-acting drugs (such as benzodiazepines or antipsychotics) were allowed to be administered during the 6-hour study period.

Furthermore, strict guidelines regarding nonpharmacological interventions were made to harmonize the intervention, to prevent possible confounders, and to ease the final interpretation. During the study, no additional nonpharmacological treatments were allowed to be applied. After including a new patient in the study, all relevant nurses in the environment of the patient were informed about the patient's inclusion and the need to avoid additional new treatments. This requirement was repeated every 2 hours while monitoring possible side effects.

Additional assessments (see below) were conducted after determining the eligibility, to give a profound insight into the sample, but were not used as inclusion or exclusion criteria.

Patients were recruited by 3 study centres in Ulm, Mannheim, and Woltersdorf/Berlin. A fourth study centre (Heidelberg) withdrew participation because of recruitment difficulties. After 6 months without screening and recruiting any participants, recruitment was stopped in this study centre. However, the Ulm study team was able to replace the remaining number of patients required. Woltersdorf/Berlin contributed 6, Mannheim 5 and Ulm 34 participants in the study. All 3 study centres were geriatric acute hospitals, with patients usually older than 70 years showing characteristic multimorbidity. Most of the study participants had commenced their inpatient treatment as planned admissions.

Prospective participants and their legal guardians were provided with written information and consent forms, and all relevant aspects of the study were discussed in detail. Because of the severity of the cognitive impairment, consent was exclusively given by the legal guardian according to the presumed wishes of the participant.

Back to Top | Article Outline

2.3. Outcome

2.3.1. Pain assessment

The primary outcome for this study was pain behaviour as measured by 4 behavioural pain assessment tools. The main study instrument was the PAINAD-G,30,34 with the BISAD,7 the Checklist of Nonverbal Pain Indicator (CNPI),5 and the ALGOPLUS29 being used as additional instruments. Along with PAINAD-G, BISAD represents the second pain assessment instrument recommended by the German Pain Society. Checklist of Nonverbal Pain Indicator was already used as a comparative assessment tool in a pilot analysis regarding psychometric properties of the PAINAD-G.17 In anticipation of having a significant percentage of participants with acute pain, ALGOPLUS was selected as a tool especially developed for monitoring acute pain.

The PAINAD-G comprises the 5 categories of breathing, negative vocalisation, facial expression, body language, and consolability, each with 3 manifestations, rated by trained health care professionals. Depending on the symptoms observed, the total score ranges from 0 to 10.34

BISAD is the German version of the French ECPA (Echelle comportemental de la douleur pour personnes âgées non communicantes).7,25 The instrument's first section (4 behavioural items) is scored at rest, the second section (4 items) during movement. The intensity of each behaviour is scored on a 5-point ordinal scale (range 0-4), leading to a total score between 0 and 32. In contrast to the other observational tools, distinct changes in behaviour over time are usually assessed as well. BISAD showed some indication of construct validity, in addition to significantly higher scores during movement and in the presence of pain-inducing diseases. However, internal consistency was weak at rest and during movement.6 Unfortunately, in this study, we were not able to collect data regarding the 2 “change in behaviour” items, because of the short intervention time and lack of knowledge of patient behaviours before hospital admission. Therefore, we decided to forego these 2 items using an adjusted version of BISAD (=BISAD modified).

The CNPI5 consists of 6 items: vocal complaints (verbal and nonverbal), facial grimaces, bracing, restlessness, and rubbing. Each item can be scored as 0 (behaviour not observed) or 1 (behaviour observed). The psychometric quality of the instrument is limited, showing lower correlation with the verbal descriptor scale, moderate levels of internal consistency, but good inter-rater reliability.

The ALGOPLUS,29 a newer French observational tool for acute pain, showed good discriminant validity with adequate internal consistency, excellent inter-rater reliability and high sensitivity to change after pain treatment.29 Excellent correlation was observed between the ALGOPLUS and the self-report.29

The PAINAD-G, the CNPI, and the ALGOPLUS had to be rated separately at rest and during activity, whereas the BISAD already includes both measuring situations.

Acute pain was assumed by the study team in the case of an acute disease that usually involves pain (derived from a list, eg, bone fracture).

Back to Top | Article Outline

2.3.2. General assessment

The following constructs were also assessed, to provide additional data for inclusion (MMSE),8 to provide a more profound insight into the study sample (Functional Assessment Screening Tool [FAST], Goodglass, and Kaplan),10,31 to detect possible confounders (CAM),14 and to reveal potential side effects of the drug used (Richmond Agitation Sedation Scale [RASS]).4

The MMSE8 measures cognitive function in terms of orientation, memory, comprehension, and use of language. The score ranges from 0 to 30. The lower the score, the worse the cognition. An MMSE score <10 indicates severe cognitive impairment.

The FAST31 reflects functional deterioration throughout the course of dementia. The scale was applied to use an additional tool to describe the participants' severity of cognitive impairment (not to clarify the eligibility). The scale consists of 7 stages with stage 6 representing severe cognitive impairment and stage 7 representing very severe cognitive impairment and complete physical dependency.

The Goodglass and Kaplan assessment10 is a hierarchical communication scale consisting of 7 levels. The lower the score, the more impaired the communication. The scale was used to give a deeper insight into the ability to communicate and to describe the study sample (not to clarify the eligibility).

The CAM14 is a screening tool to detect delirium. This instrument showed high sensitivity and specificity, as well as interobserver reliability in a well-trained research team.14

The RASS4 was used to assess alertness and agitation of the patients and to monitor possible side effects of the administered medication during the trial. It consists of a numerical scale between −10 and +10, where the positive figures represent increasing levels of alertness and negative figures increasing levels of drowsiness. An inconspicuous calm and alert patient is rated 0.

Back to Top | Article Outline

2.3.3. Power calculation

It is widely believed that a placebo effect is a result of 2 predominant mechanisms, expectation and condition, whereby expectation of the therapeutic benefit seems to play the crucial role.2 However, only people with intact cognitive function seem to be able to expect an involuntary placebo effect. It has been demonstrated that patients with Alzheimer disease show reduced placebo effects from analgesic treatment. The effects of the placebo even disappear as the disease progresses, especially when the frontal lobe functioning of the brain is affected.3 Therefore, variations within our placebo group were not expected to be significant. Gimbel et al.9 saw a reduction in pain intensity (rated on a numeric scale of 0-10) of 2.8 units in cognitive nonimpaired patients receiving oxycodone compared with 1.6 units in those only given a placebo. In a preliminary study, a similar average improvement of approximately 2.8 units between t1 and t2 was seen in the verum group, measured by the PAINAD-G.1 The standard deviation was estimated at about 2.3. Assuming a slight placebo effect, we made a conservative estimate of 0.8 units (medium variation) within the placebo group. To show a statistically significant disparity of the t1 to t2 differences between the groups (significance level of 0.05 and a power of 80%), we required a minimum total sample size of 44 participants.

Back to Top | Article Outline

2.4. Statistical analysis

To test validity with respect to pain discrimination, we analysed the t1 to t2 difference in pain behaviour in the 2 treatment groups, measured by the behavioural pain scales, using a Mann–Whitney U Test because of nonsymmetrical data.

To examine the independence of the placebo effect before treatment, the differences (δ[INCREMENT]) in PAINAD-G after 3 hours of treatment were analysed using the t2 to t3 calculation in the placebo group vs the t1 to t2 results in the treatment group. This was tested by a t test for an equivalent hypothesis, whereas δ[INCREMENT] ≤0.3 was defined as an equivalence margin in advance. Spearman correlation coefficients were calculated to check correlations among the 4 pain assessment tools at each measurement point.

Absolute and relative frequencies, mean values and SDs, or median with min/max values, respectively, are provided to describe the population. Differences in prevalence as presented in the tables were tested using the χ2 test for categorical data and t test or Mann–Whitney U test for parallel groups in the case of continuous or discrete data. Apart from the main hypothesis, all analyses were handled in an explorative manner. The significance level was set to 5%.

All analyses were performed using Statistical Package for the Social Sciences (SPSS for Windows version 20.0 (SPSS, Inc, Chicago, IL).

Back to Top | Article Outline

3. Results

3.1. Sample

Between September 17, 2010, and June 26, 2013, a total of 45 in-patients of 6361 admitted to the geriatric hospitals were enrolled in the study and underwent randomisation. Twenty-one patients were assigned to receive oxycodone, and 24 to receive a placebo (Fig. 2). No drop outs were seen during the short intervention period. The patients mean age was 84.7 years, and 77.8% of the patients were women. A total of 82.2% (N = 37) of the participants suffered from severe dementia (MMSE ≤ 10). Three patients had an MMSE score of 11, 2 of 12, and 2 of 14 points. Only one participant had an MMSE score of 19. Baseline demographic and background characteristics are summarised in Table 1. Both study groups differed significantly regarding the analgesic WHO level I medication. In the placebo group, significantly more participants were treated by a level I analgesic (placebo: 17 [70.8%] vs verum: 6 [28.6%], P < 0.005). Primarily, metamizole in a dosage of 3 or 4 times 500 mg per day was administered (44.4%), followed by ibuprofen (4.4%), and paracetamol (2.2%). 48.9% did not receive any WHO level I analgesic.

Figure 2

Figure 2

Table 1

Table 1

Furthermore, no additional significant differences in these characteristics between the 2 study groups were found.

Back to Top | Article Outline

3.2. Baseline characteristics

There were no adverse or side effects (eg, on blood pressure, vigilance, or respiratory function) seen among the participants over the observational period of 8 hours. In addition, no substantial variations of the observed values were seen.

Regarding the 4 observational scales provided, only the CNPI total score during movement shows a marginal significant difference between the 2 study groups (P = 0.05). No significant differences in other assessments could be seen before intervention, regardless of treatment group (Table 2).

Table 2

Table 2

All participants were recorded to have various illnesses that are most likely associated with pain, particularly degenerative joint diseases (75.6%), osteoarthritis (66.7%), fractures with surgery (51.1%), falls with bruising (40%), and osteoporosis with fractures (37.8%). There were no significant differences between the 2 study groups regarding possible pain causes. Although the total number of possible causes differed in the 2 study groups (66 vs 93), there was a small although not significant difference seen between the study groups concerning “causes of pain per participant” in favour of the placebo group (Table 3). Almost 70% of the participants were estimated by the study team to have acute pain. Acute pain was determined by considering medical records and identifying diagnoses, typically associated with acute pain, such as hip fractures.

Table 3

Table 3

Back to Top | Article Outline

3.3. Validity

Regarding the primary objective of this study PAINAD-G total scores (t1-t2) did not differ significantly in both study groups, neither if the observation took place at rest or during movement (P = 0.85 and 0.65, respectively) (Table 4).

Table 4

Table 4

To exploit the power of the “quasi cross-over design,” whereby each patient serves as his own control, a secondary analysis was conducted by both verum phases (t1-t2 and t2-t3, n = 45 patients) comparing with the placebo group at t1 to t2 (n = 24 patients). The second placebo group at t2 to t3 could not be used for analysis because of residual analgesic effects in this study group (Fig. 1). Similarly, PAINAD-G total scores did not differ significantly in both study groups (P = 0.88 [rest] and 0.41 [movement], respectively).

To exclude possible influences of moderate cognitive decline, an additional analysis was conducted after excluding participants with MMSE scores >10. Likewise, there was no significant difference seen between the study groups (rest or during movement) regarding the primary outcome (P = 0.92 and 0.77, respectively).

Similarly, no differences were seen when considering single items of the PAINAD-G (t1-t2) (such as breathing and vocalization), regardless of observation at rest or during movement (Table 5).

Table 5

Table 5

There were no significant difference seen between the 3 study centres regarding PAINAD-G scores in the 2 study groups (verum and placebo) at rest and during movement (P = 0.76 and 0.24 [rest] or 0.84 and 0.43 [movement], respectively) (not shown).

Figure 3 depicts the course of the PAINAD-G scores and the associated box plots (observation during movement) during the 6-hour observational period.

Figure 3

Figure 3

Similar results were seen in the other observational tools (BISAD and ALGOPLUS) except CNPI. The latter was the only tool to show a marginal significant difference in the 2 study groups at rest (Table 6).

Table 6

Table 6

Back to Top | Article Outline

3.4. Correlation

Correlations between the observational tools differed at the 3 measurement points (t1, t2, and t3) (Table 7). For example, correlation between PAINAD-G and BISAD ranged from P = 0.609 at t1 to 0.805 at t3. Mostly, correlations increased over time, but not exclusively. Moderate to high correlations between the 4 pain assessment tools ranged from P = 0.414 to 0.805 (P = 0.01). The highest correlation was seen between PAINAD-G and BISAD, followed by PAINAD-G and CNPI.

Table 7

Table 7

Back to Top | Article Outline

3.5. Verification of the equivalent hypothesis

Examination of the equivalent hypothesis revealed ≤0.3 mean differences between the verum group (t1-t2) and placebo group (t2-t3), at rest and during movement (Table 8). Regardless of whether a placebo was given before the verum or not, no difference was seen in the 2 study groups. This can be interpreted as independence of the placebo effect before treatment (equivalence not statistical significance, P > 0.05).

Table 8

Table 8

Back to Top | Article Outline

4. Discussion

The purpose of the current study was to confirm the construct validity of the PAINAD-G scale, and to examine whether it actually measures pain effectively within the population for which this instrument was originally designed, namely noncommunicative patients with advanced dementia.

In contrast to a preliminary study,1 which demonstrated a marked effect after pain treatment, the current study could not detect any significant differences between the 2 study groups regarding pain behaviour in people with advanced dementia, expected to be in pain. Pain treatment with an analgesic WHO level III medication compared with placebo did not lead to significantly lower behaviour scores as measured by the PAINAD-G scale.

Equally, our findings suggest that none of the other 3 observational tools used in the study were able to demonstrate a significant difference between the study groups. The marginal significant difference in the 2 study groups at rest using the CNPI is attributable to the slightly differing data distribution, as indicated by equal medians, and, thus, is negligible.

However, correlations among the 4 observational tools were mostly moderate to high, which can be interpreted as additional proof of the underlying validation concept according to the concurrent validity (comparing the score of one tool to another). Higher correlations between assessments that seem to measure similar constructs should support their validity and can be seen as additional evidence. Correlations increased from t1 to t3. Nevertheless, there was no obvious reason for that arising from the study design or data recording. A possible association between rising correlations at a later time points due to increased validity of the measurement instruments or a more sensitive assessment of pain by the staff can neither be sufficiently excluded nor be confirmed.

Several reasons can be discussed, which may explain our results:

Without exception, none of the applied observational tools showed, compared with the placebo group, sufficient sensitivity to change/responsiveness after analgesic intervention. Responsiveness represents an important prerequisite for proper treatment. Up to now, only a few studies have investigated the responsiveness of pain assessment tools used in people suffering from dementia. Husebo et al.15 found 4 studies that reported responsiveness, whereby the Mobilization-Observation-Behaviour-Intensity-Dementia (MOBID)-2 Pain Scale was the only tool that has been tested in advanced dementia and according to the latest COSMIN recommendations. In only 2 of the 4 studies, the investigator was subject to a blind procedure (MOBID and Algoplus), both with promising effects regarding responsiveness after analgesic treatment. However, Husebo et al. ended with the conclusion that the evidence regarding responsiveness of pain assessment tools is still limited.15

Regarding sensitivity to change/responsiveness, our study could demonstrate significant reduction of pain behaviour after intervention too; however, we would only have expected significant changes in the verum group.

A possible effect that may explain our observation is called the Hawthorne effect.24 This effect is a type of reaction in which study participants modify or improve their behaviour in response to their awareness of being observed.26 In general, pain management is more than giving an analgesic. Every contact to the patient during the study can be interpreted as social intervention and might influence the results such as the Hawthorne effect. In addition, nonpharmacological pain treatment should be considered. Despite reducing these influences as much as possible while establishing strict guidelines during the study, we cannot absolutely exclude certain external influences. Hence, intervention in this study does also not consist only of administering medication, but in accurate monitoring and help during movement over at least 5 minutes.

Other placebo effects that may influence our results are the natural course of the disease and the Rosenthal effect.26 All diseases have a natural course that includes fluctuations in symptomatology that may explain, at least in some parts, our results. The Rosenthal effect describe the phenomenon that positive expectations and beliefs of the study lead have an influence on the person observed.

People with advanced dementia would not be expected to remember previous care, and therefore, this effect would be expected to be very low in people with advanced dementia. Similar, it is largely unclear if positive expectations can influence patients with advanced dementia. However, we did see a strong reduction in behaviour in the placebo group too. Basler et al.,1 as well as Jordan et al.,16 were able to show effectiveness of intervention to treat pain measured using the PAINAD. They demonstrated a significant reduction in pain behaviour after analgesic treatment. But although these first PAINAD studies were neither controlled nor blind, the study presented here was double-blind, randomised, and placebo-controlled. Blinding excludes bias such as expectation by the carers.3

Sedation, as another potential influence factor for the effectiveness of observational tools, seems not to have a substantial impact on our current results. The RASS assessment demonstrated no significant deviations regarding alertness.

One important issue that has to be discussed is whether the fundamental construct regarding validity was consistent. Validation requires a well-established gold standard against which a new assessment method needs to be tested. Self-report is mainly considered as the gold standard for identifying the presence and intensity of pain.12 However, this gold standard is no longer applicable in people with advanced dementia. This is why we used a different method to prove validity in a double-blind, randomised, placebo-controlled design, assuming that sufficient analgesic treatment should lead to substantial reduction in pain behaviour and could be interpreted as strong proof for the validity of the assessment tool being used.

Because of the need for a verbal response, previous studies have been conducted with people suffering from not more than moderate cognitive impairment.21 This study focused on patients suffering from very advanced dementia (average MMSE: 5.5) and probable pain unable to verbally communicate. Thus, observational assessment tools may reach their limits in advanced dementia, although they are designed for this group.

The fact that almost 70% of the participants were estimated as suffering from acute pain-associated conditions could be interpreted as strong indication that the patients likely were in pain—an important precondition of the fundamental construct. However, our results could not demonstrate any difference between the 2 study groups.

Another aspect that might support the fundamental construct is the correlation between the different assessments tools used in the study. Pain assessment instruments that are supposed to measure pain in people with advanced dementia should demonstrate a higher correlation (=concurrent validity). Hence, moderate to high correlations, as seen in the study, may to a certain extent support the fundamental construct.

Pain is a subjective experience; thus, the gold standard in pain assessment is self-report. However, where self-report is not any longer applicable due to cognitive impairment, alternative ways to assess patient's pain must be found. One aspect in this context—also regarding the fundamental construct—is to measure pain at rest and during movement.12,15 Pain avoidance may mask pain behaviour, whereas moving detects pain signs. Indeed, our results showed consistently lower levels of behaviour at rest. This so-called floor effect is a general problem of observational tools. Therefore, results at rest are often insufficient and can only provide limited information as is the case in the current study. By contrast, movement provokes observable pain signs usually to a considerable degree. The typical differences between rest and during movement regarding pain signs can be used as further proof of the fundamental construct.

Similar effects can be seen regarding correlations between the observational tools used in the study. They were considerably lower at rest than during movement. This also underlines the effect of “movement” in detecting pain and can be used as additional proof for the fundamental construct.

PAINAD-G seems to detect particularly emotional aspects in the context of pain. Might it be possible that these emotional aspects increasingly disappear in advanced dementia? In line with other authors,6 observable behaviour seems to be demonstrated differently in more advanced dementia. Various expressions of pain in different patients, disease stages, and aetiologies of the dementia can all play a part regarding observable pain behaviour. Therefore, findings in people with earlier stages of dementia cannot easily be transferred to advanced stages.

Another question that may arise is whether the onset of the study was introduced too soon after admission. In general, behaviour can be affected by multiple internal and external factors. Pain is one important influence factor—but not the only one. Jordan et al.16 also mentioned a high proportion of false-positive results using the PAINAD, indicating the importance of seeking other potential causes of particular behaviour than pain. Perhaps, in our study, we assessed agitation or anxiety because of environmental changes rather than real pain behaviour. Because of ethical considerations, we began treatment within the first 24 hours after admission, assuming that all our participants were suffering from substantial pain that had not yet been adequately treated. Our results might have been influenced so heavily by other factors besides pain that no differences between the study groups could be demonstrated at this early stage of hospital stay.

According to Snow et al.,33 pain assessment in people with dementia must probably go beyond a unidimensional model of pain assessment. It is a complex communication process and can be affected by several influence factors. Pain recognition in people with dementia starts with a nociceptive stimulus, leading to pain sensation, followed by pain perception, followed by the exhibition of external signs of pain, followed by an external rater's observation, and finally interpretation of those external signs. Every step can be influenced by patient's factors (eg, sex, physical status, behaviour, cognitive impairment, and delirium), method factors (eg, assessment instrument), and rater factors (eg, pain knowledge/beliefs and relationship with patient).33 In the end, using a standardized tool is only one step in a complex diagnostic process.

The analgesic oxycodone might not be as effective in people with advanced dementia as in people without (or with mild to moderate) dementia. Oxycodone is generally known as effective analgesia.11 However, our knowledge regarding pharmacokinetics of analgesics in general, but also for oxycodone, in older patients is limited. This is even truer in the context of dementia and additional multimorbidity. A recent review from Husebo et al.15 about pain treatment in people with dementia concludes that there is still a weak evidence base for all analgesics in dementia, and well-powered randomized controlled trials are lacking. Nevertheless, we have to deal with the issue. Although findings are gathered from healthier and younger participants, thinking in analogy is an appropriate and agreed procedure in Geriatric Medicine—despite the weaknesses. Taking into account the pharmacokinetics of oxycodone, with first analgesic effects after 1 hour in younger people and maximum concentration 3 hours after administration, effects should also be seen after 3 hours in older people.

To date, there is no evidence if oxycodone is similarly effective in people with advanced dementia. Generally, the effectiveness of analgesics is explained by 2 different effects: the analgesic and the placebo effect, which complement each other.2 Placebo effects refer to improvements in patient's symptoms that are attributable to their participation in a therapeutic encounter, with its rituals, symbols, and interactions, generating expectancy and conditioning by the patient.26 Placebo effects can arise, eg, in the context of verbal and nonverbal communication, empathy, touch, diagnostic and therapeutic tools, and hospital environment. The effect is based on activation of specific areas of the brain, related to anticipation and reward, especially the prefrontal cortex and its connections to the rest of the brain.3 In the course of some types of dementia, the placebo effect diminishes and is finally no longer detectable in an advanced stage of the disease, especially when the frontal lobe of the brain is affected. Beside the global cognitive function, impairment of the connectivity between the prefrontal lobes and the rest of the brain seems to play a crucial role in triggering placebo mechanism.3 It can be argued that the fundamental precondition that the analgesic works as effectively in people with advanced dementia as in people without (or with mild to moderate) dementia was not present. According to the recommendations of Benedetti et al.3 who concluded that pain treatment should be adjusted to compensate for the loss of placebo effect, higher doses of oxycodone might be necessary to reach the same effect.

Because of ethical reasons, pre-existing analgesic treatment WHO level I as well as existing long-term medications were continued unchanged. However, effects of pre-existing treatments were estimated as minimal and consequently negligible, because of the strict inclusion criteria. Thus, patients with PAINAD-G scales ≥4 at the enrolment stage of the study indicate insufficient analgesic treatment and therefore were seen as eligible for the study, regardless of an existing analgesic level I. Conversely, those with lower scores, perhaps because of an analgesic level I treatment, were interpreted as sufficiently treated and consequently excluded from the study.

As we know from cognitive unimpaired patients, combinations of peripheral and central-acting analgesics are very common in pain treatment and reflect a usual therapeutic strategy. By contrast, WHO level I analgesics alone are frequently insufficient for providing pain reduction, underlying the frequent need of combination therapy.

From our point of view, we do not believe that the continuation of analgesic I level drugs has had a substantial influence on our results. Finally, study patients seemed to suffer from substantial pain, indicated by PANAD-G scores ≥4. We are convinced that the missing success of the observational tools can not only be explained by screwed distribution of WHO I analgesics for the benefit of the placebo group. However, despite this assumption, we cannot exclude with absolute certainty that the real effect of these pre-existing analgesics was underestimated and may explain at least parts of our findings.

Besides the limitations that were mentioned earlier and despite the particular attention being paid to delirium, we may have overlooked an intercurrent delirium in some of our patients and may not have reliably excluded them from the study using the CAM assessment. If so, some of our results might be explained by dementia and additional delirium, despite all accurateness.

Regardless of the high study standard, intensive training at the beginning of the study and close monitoring during the study by the principal investigator with repeated visits and additional phone contact to each study centre to control the progress of the study, to answer upcoming questions, and to clarify misunderstandings, we cannot absolutely exclude irregularities in the assessments conducted by the different study nurses, which may partially explain our results as well. However, every initially trained study nurse remained loyal to the project, so that there was no loss of qualified assessors.

Participants, eligible for the study, were unlikely to be able to communicate their pain because of cognitive reasons. Not the cognitive impairment level but the inability to explain pain was used as key criteria. This approach contained a considerable risk of including patients with less severe dementia and, as a consequence, might have had an influence on the results. But indeed, the study showed that eligible patients suffered mostly from severe dementia, indicating an MMSE score lower or equal to 10 (>80%). In addition, although the FAST score measures slightly different aspects of dementia compared with the MMSE, namely the additional attention paid to the physical dependency, the FAST demonstrates that only participants with severe or very severe cognitive impairment were selected. However, as secondary analyses suggest, after excluding those with moderate cognitive impairment (classified by the MMSE), results were similar (not significant!) regarding the primary objective.

Many patients had to be screened in the 3 study centres to find eligible noncommunicative participants with advanced dementia—showing that we examined a very special study group. Our results are gathered in special health care settings, namely acute geriatric hospitals. Simple transferability to other settings, such as nursing homes, has to be seen critically.

Finally, our findings refer exclusively to assessment instruments used in the study. A comparable study with other pain observational tools may lead to different results.

In view of these results, does it make sense to continue pain assessment with (these) observational tools in patients with advanced/severe dementia? Doubts are not something new. However, despite certain concerns over the past few years, most authors recommend using these instruments, probably due to a lack of alternatives.12 Nevertheless, our results may show that it is already highly doubtful whether (at least) these instruments are sensitive enough to detect changes after sufficient pain treatment in advanced dementia. Clinicians have to be aware that a score in these instruments does not necessarily mean that a patient is suffering from pain. There is a variety of reasons for observable behaviour. Further investigations are needed to examine whether similar findings can be seen in people living in nursing homes. Nursing home settings have the advantage that a possible confounder, such as a new environment for the participants, could be excluded in the results. Participants, who have already been living longer in this environment, should not be substantially influenced. On the other hand, these instruments increase carers' awareness of pain and improve the discussion within the therapeutic team.18 This might be the most important benefit of these standardised assessments—despite their imperfections. Observational tools can be helpful, but again, a critical appraisal of the overall situation is needed to integrate the results into the entire care process.

Back to Top | Article Outline

Conflict of interest statement

In addition to the grant (see above), A. Lukas and M. Schuler received remuneration in the course of giving individual lectures at Mundipharma GmbH and Grünenthal GmbH over the last 5 years. Neither the Robert Bosch Foundation nor Mundipharma and Grünenthal had any influence on the content of this article. The remaining authors have no conflicts of interest to disclose.

Registration number: DRKS00000525. Universal Trial Number (UTN): U1111-1116-6820.

Back to Top | Article Outline


The authors thank Paul East and Julie Humble for intensively reviewing the manuscript. They also thank Frank Naumann, Nadine Martin, Manuela Meißner, and Stephanie Thiel from the Geriatric Hospital Woltersdorf/Berlin, Michael Grass and Denise Goike from Diakoniekrankenhaus Mannheim, Geriatric department, as well as Christiane Küenzlen-Honold, Karin Rupp, and Antje Wieland from Geriatric Centre Ulm for participating in the study, screening eligible participants and conducting the assessment. The authors would especially acknowledge the invaluable advice and support of Professor Nikolaus, who unfortunately passed away too early without seeing the completion of this study.

The PAINAD-G II study was kindly supported by Mundipharma GmbH, Limburg, Germany. A. Lukas was funded by a Forschungskolleg Geriatrie grant from the Robert Bosch Foundation, Stuttgart, Germany (Project number: 32.5.1141.0039.0). The sponsors had no role in the design, methods, subject recruitment, data collection, analysis, or preparation of the manuscript.

Author contributions: A. Lukas initiated and conceptualized the study, formulated the hypothesis, designed the study, conducted the training, supervised the data collection, analysed the data, and drafted and revised the manuscript. U. Hagg-Grün had the daily responsibility for running the study and collecting data. He was also involved in planning the study, has analysed and interpreted the data, and drafted and revised the manuscript. B. Mayer designed the study, performed the randomization procedure, and was extensively involved in the statistical planning and analyses. He also drafted and revised the manuscript. T. Fischer designed the study, conducted the training, analysed the data, and drafted and revised the manuscript. M. Schuler conceptualized the study, formulated the hypothesis, designed the study, supervised the data collection, analysed the data, and drafted and revised the manuscript. All authors participated in critical revision of the article for intellectual content. All authors have read and approved the final manuscript.

Back to Top | Article Outline


[1]. Basler HD, Hueger D, Kunz R, Luckmann J, Lukas A, Nikolaus T, Schuler MS. Assessment of pain in advanced dementia. Construct validity of the German PAINAD. Schmerz 2006;20:519–26.
[2]. Benedetti F. Placebo analgesia. Neurol Sci 2006;27(suppl 2):S100–102.
[3]. Benedetti F, Arduino C, Costa S, Vighetti S, Tarenzi L, Rainero I, Asteggiano G. Loss of expectation-related mechanisms in Alzheimer's disease makes analgesic therapies less effective. PAIN 2006;121:133–44.
[4]. Ely EW, Truman B, Shintani A, Thomason JW, Wheeler AP, Gordon S, Francis J, Speroff T, Gautam S, Margolin R, Sessler CN, Dittus RS, Bernard GR. Monitoring sedation status over time in ICU patients: reliability and validity of the Richmond Agitation-Sedation Scale (RASS). JAMA 2003;289:2983–91.
[5]. Feldt KS. The checklist of nonverbal pain indicators (CNPI). Pain Manag Nurs 2000;1:13–21.
[6]. Fischer T. Entwicklung eines Instrumentes zum Assessment von Schmerzen bei alten Menschen mit schwerer Demenz. Institut für Medizinische Soziologie, Znetrum für Human- und Gesundheitswissenschaften der Medizinischen Fakultät. Berlin: Charité—Universitätsmedizin Berlin, 2009. p. 192.
[7]. Fischer T. Schmerzeinschätzung bei Menschen mit schwerer Demenz. Das Beobachtungsinstrument für das Schmerzassessment bei alten Menschen mit schwerer Demenz (BISAD). Bern: Verlag Hans Huber, Hogrefe AG, 2012.
[8]. Folstein MF, Folstein SE, McHugh PR. “Mini-mental state.” A practical method for grading the cognitive state of patients for the clinician. J Psychiatr Res 1975;12:189–98.
[9]. Gimbel JS, Richards P, Portenoy RK. Controlled-release oxycodone for pain in diabetic neuropathy: a randomized controlled trial. Neurology 2003;60:927–34.
[10]. Goodglass H, Kaplan E. The Assessment of Aphasia and related disorders. Philadelphia: Lea & Febinger, 1983.
[11]. Guerriero F, Roberto A, Greco MT, Sgarlata C, Rollone M, Corli O. Long-term efficacy and safety of oxycodone-naloxone prolonged release in geriatric patients with moderate-to-severe chronic noncancer pain: a 52-week open-label extension phase study. Drug Des Devel Ther 2016;10:1515–23.
[12]. Hadjistavropoulos T, Herr K, Prkachin KM, Craig KD, Gibson SJ, Lukas A, Smith JH. Pain assessment in elderly adults with dementia. Lancet Neurol 2014;13:1216–27.
[13]. Herr K, Bjoro K, Decker S. Tools for assessment of pain in nonverbal older adults with dementia: a state-of-the-science review. J Pain Symptom Manage 2006;31:170–92.
[14]. Hestermann U, Backenstrass M, Gekle I, Hack M, Mundt C, Oster P, Thomas C. Validation of a German version of the confusion assessment method for delirium detection in a sample of acute geriatric patients with a high prevalence of dementia. Psychopathology 2009;42:270–6.
[15]. Husebo BS, Achterberg W, Flo E. Identifying and managing pain in people with Alzheimer's disease and other types of dementia: a systematic review. CNS Drugs 2016;30:481–97.
[16]. Jordan A, Hughes J, Pakresi M, Hepburn S, O'Brien JT. The utility of PAINAD in assessing pain in a UK population with severe dementia. Int J Geriatr Psychiatry 2011;26:118–26.
[17]. Kaspar R. Psychometrische Beurteilung verhaltensgestützter Schmerzassessments für Menschen mit Demenz—Potenziale von Item-Response-Theorie und Latent Variable Modellen am Beispiel der Verhaltensinventare CNPI und BESD. Fakultät für Verhaltens- und Empirische Kulturwissenschaften. Heidelberg: Ruprecht-Karls-Universität, 2009.
[18]. Kojer M, Schmidl M. Demenz und Palliaitive Geriatrie in der Praxis—Heilsame Betreuung unheilsamer dmenzkranker Menschen. Wien, New York: Springer, 2011.
[19]. Lichtner V, Dowding D, Esterhuizen P, Closs SJ, Long AF, Corbett A, Briggs M. Pain assessment for people with dementia: a systematic review of systematic reviews of pain assessment tools. BMC Geriatr 2014;14:138.
[20]. Lints-Martindale AC, Hadjistavropoulos T, Lix LM, Thorpe L. A comparative investigation of observational pain assessment tools for older adults with dementia. Clin J Pain 2012;28:226–37.
[21]. Lukas A, Barber JB, Johnson P, Gibson SJ. Observer-rated pain assessment instruments improve both the detection of pain and the evaluation of pain intensity in people with dementia. Eur J Pain 2013;17:1558–68.
[22]. Lukas A, Schuler M, Fischer TW, Gibson SJ, Savvas SM, Nikolaus T, Denkinger M. Pain and dementia: a diagnostic challenge. Z Gerontol Geriatr 2012;45:45–9.
[23]. Mandema JW, Kaiko RF, Oshlack B, Reder RF, Stanski DR. Characterization and validation of a pharmacokinetic model for controlled-release oxycodone. Br J Clin Pharmacol 1996;42:747–56.
[24]. McCarney R, Warner J, Iliffe S, van Haselen R, Griffin M, Fisher P. The Hawthorne effect: a randomised, controlled trial. BMC Med Res Methodol 2007;7:30.
[25]. Morello R, Jean A, Alix M, Sellin-Peres D, Fermanian J. A scale to measure pain in non-verbally communicating older patients: the EPCA-2 Study of its psychometric properties. PAIN 2007;133:87–98.
[26]. Morral A, Urrutia G, Bonfill X. Placebo effect and therapeutic context: a challenge in clinical research. Med Clin (Barc) 2017;149:26–31.
[27]. Mosele M, Inelmen EM, Toffanello ED, Girardi A, Coin A, Sergi G, Manzato E. Psychometric properties of the pain assessment in advanced dementia scale compared to self assessment of pain in elderly patients. Dement Geriatr Cogn Disord 2012;34:38–43.
[28]. Prince M, Bryce R, Albanese E, Wimo A, Ribeiro W, Ferri CP. The global prevalence of dementia: a systematic review and metaanalysis. Alzheimers Dement 2013;9:63–75.e62.
[29]. Rat P, Jouve E, Pickering G, Donnarel L, Nguyen L, Michel M, Capriz-Ribiere F, Lefebvre-Chapiro S, Gauquelin F, Bonin-Guillaume S. Validation of an acute pain-behavior scale for older persons with inability to communicate verbally: Algoplus. Eur J Pain 2011;15:198.e191–198.e110.
[30]. Schuler MS, Becker S, Kaspar R, Nikolaus T, Kruse A, Basler HD. Psychometric properties of the German “Pain Assessment in Advanced Dementia Scale” (PAINAD-G) in nursing home residents. J Am Med Dir Assoc 2007;8:388–95.
[31]. Sclan SG, Reisberg B. Functional assessment staging (FAST) in Alzheimer's disease: reliability, validity, and ordinality. Int Psychogeriatr 1992;4(suppl 1):55–69.
[32]. Smith K, Hopp M, Mundin G, Leyendecker P, Bailey P, Grothe B, Uhl R, Reimer K. Single- and multiple-dose pharmacokinetic evaluation of oxycodone and naloxone in an opioid agonist/antagonist prolonged-release combination in healthy adult volunteers. Clin Ther 2008;30:2051–68.
[33]. Snow AL, O'Malley KJ, Cody M, Kunik ME, Ashton CM, Beck C, Bruera E, Novy D. A conceptual model of pain assessment for noncommunicative persons with dementia. Gerontologist 2004;44:807–17.
[34]. Warden V, Hurley AC, Volicer L. Development and psychometric evaluation of the Pain Assessment in Advanced Dementia (PAINAD) scale. J Am Med Dir Assoc 2003;4:9–15.

Pain; Dementia; Validation; PAINAD; Observational tool; Pain assessment

© 2019 International Association for the Study of Pain