J Manipulative Physiol Ther. 1992 Mar-Apr; 15(3): 181-94.
A meta-analysis of clinical trials of spinal manipulation.
Mills College, Oakland, CA 94613.
OBJECTIVE: To assess the efficacy of spinal manipulative therapy (SMT) in the treatment of back pain using meta-analytical techniques. DATA SOURCES: The literature was systematically searched for all studies of SMT through June 1989. The Index Medicus from 1980 was expanded by citation tracking. The Chiropractic Research Archives Collection was utilized as a regularly updated bibliographic source for the location of research publications. A hand search of professional chiropractic journals was also undertaken. STUDY SELECTION: Studies in English with concurrent controls treated by methods other than SMT, including sham, produced 23 randomized controlled clinical trials of the effectiveness of spinal manipulation. Because a single trial might include more than one comparison of treatments, these trials produced a total of 34 mutually exclusive, discrete samples. DATA EXTRACTION: Data were extracted via a standardized coding document by one author and verified by two of the others. Data were independently extracted from a subset of the studies by a blinded research assistant to ensure that coding methods produced acceptable consistency. DATA SYNTHESIS: Effect sizes (Cohen's D index) were calculated for nine outcome variables at eight time points following the initiation of treatment. Thirty-eight of 44 effect sizes indicated that SMT was better than the comparison treatment. It was also found that meta-analysis was an imperfect instrument for the kind of trials that were pooled in this study because the research protocols were highly diverse. Furthermore, because the nature of SMT does not permit an easy use of placebos, true no-treatment control groups were rare. Most studies compared SMT to an alternative treatment. This probably obscured the effectiveness of SMT since the comparison treatments were presumably also effective. CONCLUSIONS: SMT proved to be consistently more effective in the treatment of low back pain than were any of the array of comparison treatments. The analysis provided some suggestion that manipulation, as such, is more effective than mobilization, as such. For the future, it is suggested that researchers strive for more consistent measures in terms of explicit descriptions of the nature of SMT, the times of post-treatment assessments and the nature of outcome measures. Only then can meta-analysis fulfill its potential in this clinical area.