The existing understanding of how CST works contradicts modern theories of health and disease, diminishing its plausibility for clinical effectiveness. To establish CST as a recommended practice in clinical settings, compelling results from prospective studies with adequate sample sizes at low risk of bias are necessary. The findings from this meta-analysis indicate that the current body of evidence supporting CST does not reach this standard. This is in line with conclusions found in previous reviews and meta-analyses [15, 16, 18, 20, 21, 23]. A meta-analysis by Haller et al. (2020) found a significant effect of CST on pain and disability, this is in agreement with the presented work. Of note, only six of ten studies included in Haller et al. met inclusion criteria and were reanalyzed in this current report. The only other identified positive evidence, presented in a meta-analysis by Jiang et al. (2023), found a significant effect of CST on short hamstring syndrome. This indication was not assessed in this report as no related trials meeting inclusion criteria were identified.
While significant findings were reported by most studies, per-study effect sizes calculated here indicate that few significant effects were found. This was true of both individual effect sizes as well as per-study aggregate effect sizes. One possible explanation for this discrepancy is the occurrence of post-hoc statistical analysis and post-hoc outcome selection in the assessed articles—two common concerns identified during the risk of bias assessment. Another contributing factor may be that few studies calculated an effect size or other standardized measure of efficacy. Many of the included studies also did not correct for multiple outcome measures or repeated measures despite high numbers of outcomes and statistical tests. Additionally, subgroup analysis of primary outcomes indicates that there is no evidence for a significant effect of CST.
When all outcomes and timepoints were considered, potential effects on pain, disability, and neonatal structure were observed. While notable in individual analyses, this observation is tempered by various limitations. These limitations include a high level of bias, multiple measures, multiple time points, wide prediction intervals in calculated effect sizes, inadequate blinding and evidence of post-hoc data analysis. Additionally, p-curve analysis of disability-related outcomes does not support true evidential value, though there is not enough evidence to definitively state that none exists. Consequently, caution is advised in interpreting any positive effects identified as the likelihood of significant positive bias affecting the outcomes is considerable. As a result, the current findings suggest a lack of conclusive evidence supporting the efficacy of craniosacral therapy for the indications explored within this meta-analysis.
Despite significant effect sizes and support from p-curve analysis, outcomes related to neonatal structure and somatic pain displayed prediction intervals that broadly crossed zero. Such a wide prediction interval suggests substantial uncertainty regarding the treatment’s potential effectiveness in practical application or in subsequent studies. Additionally, these outcome categories are not significant when only primary outcomes are considered. More research may help clarify whether there is any benefit from the use of CST for these populations and indications.
The results of the outcome-based meta-analysis were affected by the decision to exclude outliers. Initially, an algorithm flagged studies in which the effect size fell outside the confidence interval of the group effect size. Influence analysis was conducted in order to confirm these results, which agreed with the initial detection. This process is limited by the variation within and between studies, differing outcomes, and multiple indications. The supplemental material contains results that include outliers, demonstrating only one significant difference in Neonate health, behavior (Supplement 2C). However, considering the large contribution to heterogeneity and the high bias of the outlier studies, the calculated effect sizes excluding outliers are likely closer to reality. It should also be noted that subgroup analysis of primary outcomes was conducted including outliers.
This analysis is limited by problems inherent in the included studies, including repeated measures and multiple time point measures. Efforts to correct for this based on recommendations by Morris et al (2002) [45] still led to a violation of the assumption of independence, meaning confounding factors may be present. Additionally, a number of the included studies did not provide detailed descriptions of the techniques used and many studies mixed CST with other osteopathic treatments. Both these factors may have impacted the results, though likely in the positive direction. Given the overall negative outcome, this did not result in reinterpretation of the findings, though care should be taken when interpreting the three positive effects identified.
The limitations in the available literature warrant reservation in considering CST as part of evidence-based treatment plans until substantially higher-quality evidence emerges. The role and scope of CST should be reevaluated in modern osteopathic medicine. While the findings from this meta-analysis have the potential to inform future research directions, it becomes challenging to advocate for continued exploration of CST considering the predominantly negative outcomes which persist despite various methodological concerns producing probable positive bias. Further research exploring the physiologic mechanisms behind CST could help resolve some of the controversy surrounding its use and provide insight into potential clinical applications.