Opioids including morphine and tramadol are widely used in the management of post-operative pain. Nevertheless, some patients also respond to placebo medication [1,2]. Although the management of acute pain has undoubtedly improved through the wide-spread use of patient-controlled analgesia (PCA), patients still exist who experience inadequate analgesia or require doses of analgesics that exceed the standard protocol. Establishing factors that might predict response to PCA treatment and the amount of analgesic consumption could be of advantage and may influence further pain management. The PCA bolus could be adjusted to the patients' need and, in cases of high analgesic consumption and insufficient pain relief, a non-steroidal anti-inflammatory drug (NSAID) could be used to improve analgesia.
In a previous study, we compared the analgesic efficacy of post-operative morphine, tramadol and placebo PCA treatment . Here, we describe a second analysis investigating the groups of responders and non-responders to post-operative PCA treatment. The early treatment period with baseline pain scores and the amount and effect of loading doses were evaluated. The following questions were considered. Which attributes characterize responders and non-responders? Can the loading dose of an analgesic distinguish responders from non-responders? Which variables might influence non-response and is non-response caused by variables such as the ineffectiveness of the drug or the stress of the surgery? Are there any differences between placebo responders and true responders?
Approval of the study design was obtained from the local ethics committee. A prospective, randomized and double-blind design was used. After giving written informed consent, patients older than 18 years (ASA I-II) scheduled for elective intra-abdominal gynaecological or abdominal surgery were included in the study.
On the evening before surgery, patients were instructed in the use of the PCA equipment. General anaesthesia was conducted using a standardized protocol. After recovery from anaesthesia, pain was measured using a visual analogue scale (VAS; 0 denotes no pain, 100 denotes worst pain imaginable). Patients with a post-operative pain intensity of at least 40 were randomized to treatment with tramadol, morphine or placebo. There were 60 patients in each group. Study medication was titrated individually by administering up to 20 mg of morphine, 200 mg of tramadol or placebo intravenously (i.v.). If the pain score decreased by at least 20 on the VAS within the first 30 min from the beginning of titration (primary response), the patients were connected to the PCA device for the following 47.5 h. Patients could demand repetitive i.v. bolus doses of 1 mL of solution equivalent to 2 mg of morphine, 20 mg of tramadol or 1 mL of placebo up to a daily maximum of 100 mg of morphine, 1000 mg of tramadol or 50 mL of placebo. The delivery time of one dose via the PCA pump was 2 min, and the lockout time between two single boluses was set to 5 min. VAS and individual satisfaction with pain management were recorded before the initial bolus and thereafter at 15, 30 and 60 min, hourly for a further 7 h and at intervals of 5 h thereafter.
A patient was considered as a responder if the following two criteria were fulfilled. (1) The pain score decreased by at least 20 during the first 30 min; and (2) the patient was satisfied with pain management during the whole study period. A patient was considered a non-responder if pain scores decreased less than 20 points on the VAS within the first 30 min, if study medication proved to be ineffective during the further study period or if the patient was not satisfied with pain management. These non-responders were switched to morphine by PCA as rescue medication. Patients terminating the double-blind period prematurely or violating the study design were not included in the analysis.
Demographic data, surgery-related variables, pain scores, loading doses and number of PCA boluses were compared between responders and non-responders and between the treatment groups. To differentiate between the different kinds of surgery, five categories were introduced: (1) minor abdominal surgery - herniotomy, appendectomy; (2) intermediate laparoscopic surgery - tubal surgery, surgery of the ovary; (3) major laparoscopic surgery - hysterectomy, cholecystectomy, excision of myomas; (4) vaginal hysterectomy; (5) major abdominal surgery - gastrectomy, colonic surgery, Wertheim's operation.
'Survival analysis' considering all 180 patients used a multiple-sample test to compare the duration of double-blind periods between the treatment groups. Kruskal-Wallis and Mann-Whitney U-tests were used for the statistical evaluation of group differences (demographic data, duration of surgery, baseline pain score). A non-parametric test was used, because base-line pain scores were not normally distributed as a result of the patient's inclusion criteria of a pain score of at least 40 on the VAS. Manova was performed for between-group differences (responders/non-responders, treatment groups). The difference in types of surgery was analysed using a chi-square test, and Pearson product moment was used to test correlations. Differences were considered statistically significant for P<0.05, and the test adjusted by Bonferroni correction.
Ninety-six patients were responders (morphine 45, tramadol 40, placebo 11), and 65 patients (morphine 9, tramadol 14, placebo 42) were non-responders. The remaining 19 patients terminated the trial prematurely by refusal of rescue medication (2/2/2) or violation of the study design (4/4/5).
The Kaplan-Meier double-blind 'survival analysis' (Fig. 1) revealed a significant difference in duration of the period between patients on morphine, tramadol and placebo (P<0.001). Median time intervals were 43.5 h (confidence interval 27.4-36.8 h) for morphine, 33.3 h (23.3-33.4 h) for tramadol and 0.5 h (4.6-11.4 h) for placebo. Thirty-nine patients on placebo (65%) were changed to rescue medication because of inadequate analgesia after the first 30 min. The corresponding data for patients on morphine and tramadol were 7 (11.7%) and 12 (20.0%). There was a parallel decline in the numbers of patients remaining on study medication. The majority of non-responders (89.2%) were identified 30 min after the loading dose. Only seven further patients (M:2, T:2, P:3) were assessed as non-responders during the subsequent study period.
Pain scores after 30 min (Fig. 2) as well as the loading dose (r = 0.42, P<0.001) correlated with pain VAS during the following 8 h. These variables also correlated with the number of PCA boluses delivered (VAS 30 min: r=0.36 P<0.001; loading dose: r=0.32, P<0.001). The ratio of loading doses to further analgesic consumption was 1.1 (0.4-1.6), 2.2 (1.1-3.3) and 3.5 (1.6-6.0) up to the 8th, 23rd and 48th hour, respectively [median (lower-upper quartile)]. Patients' analgesic requirements equalled the loading dose, twice the loading dose and three to four times the loading dose up to the 8th, 23rd and 48th hour.
Comparison of responders and non-responders
Age, weight and height were similar in responders and non-responders (Table 1). The group of non-responders underwent a higher proportion of major surgery (P=0.036), more procedures being categorized as type 4 or 5 (Table 2). This difference was most pronounced in the morphine group (P=0.007). Duration of surgery was longer in non-responders than in responders (P<0.001). These differences were significant in the morphine and placebo groups only.
The pain scores of all responders showed no difference with respect to morphine, tramadol or placebo medication. Equally, all non-responders in the three treatment groups had similar pain scores, although at a higher level than responders (Fig. 3).
A comparison of pain scores in responders vs. non-responders revealed higher baseline VAS for non-responders in the morphine group only (P=0.001). Thirty minutes after the initial bolus, pain VAS were significantly higher in all non-responders (P<0.001) compared with responders. Responders also showed significantly more pain relief than non-responders (P<0.001) when compared with baseline pain score after 30 min (Fig. 4).
Although non-responders were changed to morphine, their pain scores up to the 7th hour showed significantly higher values than responders (P<0.001). Loading doses were significantly higher in non-responders (P<0.001), and non-responders demanded and received more PCA boluses than responders (Table 2).
Comparison of responders to morphine, tramadol or placebo
A total of 96 patients responded to morphine, tramadol or placebo. The three groups differed in their number of responders. However, there was no difference in their demographic data, the duration and type of surgery and baseline VAS (Tables 1 and 2). Pain scores within the first 8 h were similar in all three groups, and there was no difference in pain reduction after 30 min in comparison with baseline pain score (Figs 3 and 4). During the remainder of the study, 11 patients receiving placebo had sufficient pain control and could be classified as responders.
The cumulative delivered PCA doses showed a wide interindividual variability, comparable between the treatment groups. The frequency of dose requirements and the number of boluses delivered were similar in the morphine and tramadol group (Table 1). Up to the 8th hour, placebo responders demanded (P<0.001) and received (P=0.002) significantly more boluses than the true responders. After 23 h, the number of boluses demanded and received was still twice as high in patients on placebo medication (not significant). One bolus of morphine or tramadol produced analgesia for 90 min during the first 8 h; however, in patients given placebo, this time interval was only 41 min (P=0.02). Throughout the first 23 h, one bolus lasted 110 and 120 min for patients on morphine and tramadol, respectively, but only 55 min for patients receiving placebo (not significant). Eight out of 11 patients receiving placebo had no need for any further analgesic after 24 h. Responders in the placebo group used the PCA device for a shorter period of time (median 25 h) than patients on morphine (48 h) or tramadol (48 h).
Comparison of non-responders to morphine, tramadol or placebo
Non-responders to morphine, tramadol and placebo did not differ in their demographic data, but displayed differences in duration and type of surgery and baseline pain scores (Table 1). The small number of morphine non-responders underwent longer surgical procedures.
Pain reduction after 30 min in comparison with base-line pain score was similar between the non-responders (Fig. 4). During the next 23 h, pain scores in all non-responders remained higher compared with responders (P<0.005), although they did receive morphine as a rescue medication.
Loading doses were comparable between the groups of non-responders and, in total, these doses were significantly higher than in responders. Subsequent analgesic consumption remained high for all non-responders (Table 1).
This study compares responders and non-responders to post-operative pain management. As expected, the number of responders to morphine, tramadol and placebo medication differed, which can be explained by the different analgesic efficacy of the drugs. The majority of non-responders were identified after the loading dose, although, during the following study period, seven further patients were assessed as non-responders. Pain scores following the loading dose and the loading dose itself seem to be useful factors in predicting further analgesic consumption and pain scores. Butscher et al. used the initial i.v. bolus dose to predict the patients' need for intermittent intramuscular injections on the ward. In 78 patients investigated by Macintyre and Jarvis , pain scores and loading dose correlated with morphine consumption during the first 24 h after surgery. A careful observation of the loading dose and its effect on pain reduction should become standard in post-operative pain treatment. Further dose regimens can then be adapted to individual needs and requirements. The ratios of the loading doses and further analgesic consumption did not differ between responders and non-responders in this study. However, patients not responding needed greater amounts of analgesic drugs. The loading dose may be of value in assessing further analgesic consumption and may also be used to adjust the analgesic regimen to individual needs. Individual variation has also to be considered because of many additional variables being involved [5,6]. Nevertheless, future assessment of the primary response (loading dose and pain scores during the first half hour) might help to identify patients with extremely low analgesic consumption who will probably not need a PCA device, as well as patients at risk of not responding to pain treatment.
Characterization of responders
Responders showed similar pain reduction irrespective of the drug they received. In contrast, when looking at all 180 patients, the expected difference in analgesic efficacy under true or placebo treatment became obvious.
According to Turner et al., placebos have time-effect curves and effects similar to those of active medications. Analysing the analgesic consumption, responders showed a wide variability in dosage requirements. Placebo responders displayed a similar variability, although the number of patients in this group was small. A wide range of analgesic requirements has been reported by several other authors [5,8-11]. To some extent, pharmacodynamic and pharmacokinetic mechanisms may explain this phenomenon [9,10]. Individual blood and cerebrospinal fluid concentrations of endogenous opioids, neuropeptides and stress hormones [8,12], psychological [9,13,14] and personality traits [15,16] are further contributors to variations in dose requirements .
In this study, the time effect of one PCA bolus was shorter in placebo responders than in morphine or tramadol responders. Responders in the placebo group delivered twice the number of boluses during the first day in comparison with responders receiving genuine medication. The placebo effect apparently, therefore, contributed to about half of the analgesic effect, with genuine medication contributing to the other half. The placebo effect, often described as a non-specific drug effect , is not only produced by the drug, but by the PCA device as well . Keeri-Szanto  suggested that the non-drug component of PCA action, including a sense of control over pain relief, could be as much as 60%.
All non-responders had higher pain scores and more analgesic consumption compared with responders, although those patients received medication with morphine. Reasons that might account for non-responsiveness to i.v. post-operative pain treatment include severity and duration of surgical trauma, poor drug analgesic efficacy and inadequate analgesic doses. Indeed, the small number of morphine non-responders were patients with the longest duration of surgery. Only major procedures of category 4 and 5 were represented in this group. They were associated with the highest baseline pain scores in this group. Alternatively, the morphine loading dose may have been inadequate. Morphine loading doses were consistent with previous studies [4,5]. Some patients may need more than 20 mg of morphine as an initial bolus; Macintyre and Jarvis  administered loading doses of up to 40 mg of morphine. During the PCA period, patients could demand daily doses of up to 100 mg of morphine, and the PCA bolus of 2 mg was relatively large. Morphine consumption in this trial was, however, in agreement with previous studies [19,20]. In the tramadol group, the same reasons may account for non-responsiveness, although the number of non-responders might be higher because tramadol is not as potent as morphine.
The severity of surgical trauma and high baseline VAS did not play a major role in the non-response of most patients receiving placebo. Baseline VAS in placebo non-responders were similar to baseline VAS in responders. The majority of patients receiving placebo probably did not respond because of its absent analgesic effect.
Responders, including the 11 placebo responders, revealed similar behaviour concerning pain scores and loading doses, irrespective of the treatment they received. Non-responders demonstrated significantly higher pain scores over the subsequent 23 h, although they received morphine as rescue medication. Some 89.2% of the non-responders had already been identified after the initial loading dose. This primary response to the loading dose may be useful in predicting analgesic consumption during PCA treatment and in identifying patients at risk of insufficient pain relief.
This project was carried out with support from Grünenthal GmbH, Stolberg, Germany. The authors are gratefully indebted to U. Hennig for valuable assistance throughout the study. The authors also wish to thank the medical and nursing staff of the Departments of Anesthesiology of the Universities of Kiel and Cologne and the Malteser Hospital of Bonn for their help and cooperation, which made this study possible.
1 Beecher HK, Boston MD. The powerful placebo. JAMA
2 McQuay H, Carroll D, Moore A. Variation in the placebo effect in randomised controlled trials of analgesics: all is as blind as it seems. Pain
3 Stamer UM, Maier C, Grond S et al.
Tramadol in the management of postoperative pain: a double-blind, placebo- and drug-controlled study. Eur J Anaesthesiol
4 Butscher K, Mazoit JX, Samii K. Can immediate opioid requirements in the post-anaesthesia care unit be used to determine analgesic requirements on the ward? Can J Anaesth
5 Macintyre PE, Jarvis DA. Age is the best predictor of postoperative morphine requirements. Pain
6 Mitchell RWD, Smith G. The control of postoperative pain. Br J Anaesth
7 Turner JA, Deyo RA, Loeser JD et al.
The importance of placebo effects in pain treatment and research. JAMA
8 Tamsen A, Hartvig PDalström B et al.
Endorphins and ondemand pain relief. Lancet
9 Gourlay GK, Kowalski SR, Plummer JL et al.
Fentanyl plasma concentration - analgesic response relationship in the treatment of postoperative pain. Anesth Analg
10 Burns JW, Hodsmann NBA, McLintock TTC et al.
The influence of patient characteristics on the requirements for postoperative analgesia. Anaesthesia
11 Lehmann KA. Intravenous patient-controlled analgesia: postoperative pain management and research. In: Chrubasik J, Cousins M, Martin E, eds. Advances in Pain Therapy II.
Berlin, Heidelberg: Springer, 1993; 65-93.
12 Sjöström S, Tamsen A, Hartvig P et al.
Cerebrospinal fluid concentrations of substance P and (Met) enkephalin-ARG6
during surgery and patient-controlled analgesia. Anesth Analg
13 Taenzer P, Melzack R, Jeans ME. Influence of psychological factors on postoperative pain, mood and analgesic requirements. Pain
14 Jamison RN, Taft K, O'Hara JP, Ferrante FM. Psychological and pharmacologic predictors of satisfaction with intravenous patient-controlled analgesia. Anesth Analg
15 Boyle P, Parbrook GD. The interrelation of personality and postoperative factors. Br J Anaesth
16 Thomas V, Heath M, Rose D, Flory P. Psychological characteristics and the effectiveness of patient-controlled analgesia. Br J Anaesth
17 Welchew EA. On-demand analgesia. Anaesthesia
18 Keeri-Szanto M. Drugs or drums: what relieves post-operative pain? Pain
19 Owen H, Plummer JL, Armstrong L et al.
Variables of patient-controlled analgesia. 1. Bolus size. Anaesthesia
20 De Kock M, Scholtes JL. Postoperative PCA in abdominal surgery. Analysis of 200 consecutive patients. Acta Anaesthesiol Belg