We read with interest the paper by George et al on a multimodal intervention involving manual therapy, exercise, and education for low back pain and pelvic pain during pregnancy. Although the authors concluded that musculoskeletal and obstetric management (MOM) in midpregnancy is more beneficial to patients than standard obstetric care, they acknowledged several important study limitations including the absence of active or placebo comparators, the inability to independently assess individual components of the multimodal intervention, and likely the enrollment of patients motivated to achieve successful results, thereby having an impact on the generalizability of results.

However, other aspects of study design, analysis, and reporting raise further questions. The eligibility criteria specified women with fetuses of 24-28 weeks’ gestation; however, data on gestational age at baseline (cited in Table 1) suggest that more than half of the patients were prematurely enrolled. Furthermore, in the absence of recorded logs, patients in the MOM group may have received up to 10 or more weekly treatments by 33 weeks’ gestation as opposed to only 4-6 treatments as reported.

The reported power analysis allowed for 20% attrition, suggesting that an intention-to-treat analysis was not used. The Consolidated Standards of Reporting Trials diagram reported that 19% overall and 24% in the MOM group were lost to follow-up. The impact of these relatively high levels of attrition on study outcomes was not assessed and reported. Furthermore, the power analysis was reportedly based on a numerical rating scale (NRS) and the Quebec Task Force Disability Questionnaire (QDQ) as the primary outcomes. Yet it appears that the sample size to achieve statistical power of 0.9 was initially computed for the QDQ (n = 120) and that the statistical power achievable with 120 patients on the NRS was secondarily computed and reported (power, 0.96). Correspondingly, the ClinicalTrials.gov registry lists the QDQ as the only primary outcome, with the NRS and other variables as secondary outcomes.

It is commendable that the authors used a Bonferroni correction to reduce the likelihood of type 1 errors in testing multiple hypotheses; however, the rationale for using 10 in the denominator to compute the appropriately adjusted P value is unclear. It is also puzzling that P values for interaction were reported for both the MOM and standard obstetric care groups (cited in Tables 2 and 3; presumably, 3 P values for interaction in Table 3 were misplaced). Overall, these methodological issues and reporting inconsistencies diminish the potential impact of this important study and call into question the significance of several reported findings.