Eva Ørnbøl

Background SF-36 Physical and Mental Component Summaries (PCS and MCS) are increasingly used as main measures of outcome in clinical trials. Yet, studies in different patient populations have demonstrated serious problems in interpreting these summary scores.

We are conducting a randomised controlled trial on cognitive behavioural therapy in patients with severe and chronic functional disorders. Based on a pilot study and baseline data, we have assessed the performance of the summary scores.

Aim To demonstrate problems in the orthogonal factor solution for PCS and MCS and to assess other methods of summarising the SF-36 subscales.

Method Subjects were 102 patients included in our trial and 10 patients from the pilot study. We examined three methods of summarising physical and mental health subscales of the SF-36: the original summary scores PCS and MCS, an alternative summary scoring based on an oblique factor solution and the summary components from the RAND-36 HSI.

Results Pilot study: Improvement on subscales of physical health was not reflected by the original PCS. The three methods showed different results with regard to individual changes over time.

Baseline data: Surprisingly, SF-36 PCS and MCS showed moderate negative correlation (r=-0.48, p=0.000), which is contradictory to the orthogonal construction of the Component Summaries. SF-36 MCS was significantly higher than the Mental Health Component from the RAND-36 HSI (95%CI 39.7-44.7 and 32.0-35.8, respectively), while mental health subscales showed no significant differences.

Conclusion The negative weights of the orthogonal factor solution distorted the results in both study samples. As outcome measures, PCS and MCS should be interpreted with caution in patients with functional disorders and in other patient groups with interacting physical and mental health problems. One may benefit from using other scoring methods such as the RAND-36 HSI to validate the results.