Background and objective 

Post-operative analgesic consumption is often used as a surrogate measure for pain; analyses of mean data assume a Gaussian distribution and use parametric statistics to assess statistical differences, often in small samples. We used a large individual patient dataset to examine the distribution of analgesic consumption, the validity of such analyses and alternative dichotomous outcomes.

Methods 

Analysis of individual patient data from 913 patients over 48 post-operative hours in five randomised trials. Patients had either epidural injection of placebo or morphine (as sulphate and extended release epidural morphine) and use of patient-controlled analgesia. Post-operative fentanyl consumption was calculated over 0–24, 24–48 and 0–48 h.

Results 

The distribution of analgesic consumption for all patients over the periods 0–24, 24–48 and 0–48 h was exponential. Most patients used less than 750 μg fentanyl over 48 h; 34% used over 1000 μg fentanyl (100 mg morphine), 13% over 2000 μg and 5% over 3000 μg. Mean, median and mode were very different; 20% of patients consumed almost 60% of post-operative analgesic, and standard deviations were generally larger than means. A useful dichotomous outcome was less than 750 μg fentanyl consumed over 48 h, a level associated with very good or excellent patient pain rating. Use of very good or excellent patient pain rating differentiated between different doses of epidural morphine.

Conclusion 

Because of a highly skewed distribution, post-operative analgesic consumption is an uncertain method of measuring analgesic efficacy of an intervention designed to limit pain during and after surgery.