Randomized controlled trial of an internet-based self-guided hand exercise program to improve hand function in people with systemic sclerosis: the Scleroderma Patient-centered Intervention Network Hand Exercise Program (SPIN-HAND) trial

A Systematic Review of the Statistical Methods Adopted for Analyzing Follow-Up Data in Cohort Multiple Randomized Controlled Trial

BACKGROUND

The cohort multiple randomized controlled trial (cmRCT) can tackle some of the weaknesses of an RCT which has triggered the interest of researchers considerably over time. Several challenges persist regarding the methods of analyzing such valued data. The paucity of international recommendations concerning the statistical methods for analyzing trial data has led to a variety of strategies further complicating the result comparison. Our aim was to review the different cmRCT analysis methods since cmRCT was first proposed in 2010.

METHODOLOGY

A search for full-length studies presenting statistical analysis of the data collected adopting a cmRCT design was conducted on PubMed, Cochrane Library, EMBASE, JSTOR, Scopus, MEDLINE, and ClinicalTrials.gov.

RESULTS

Out of 186 studies screened, we selected 22 for the full-text screening and 11 were found eligible for data extraction. All 11 studies were conducted in high-income countries, reflecting the design being underutilized in other settings. All of the studies were found to have used intention-to-treat (ITT) analysis with four of them utilizing instrumental variables (IV) analysis or a complier average causal effect (CACE). Randomization was noted often to be interchangeably used for random selection. Sample size calculation was not clearly specified in the majority of the studies.

CONCLUSION

Clarity regarding the distinction between an RCT and a cmRCT is warranted. The fundamental difference in design, which leads to certain biases that need to be taken care of by adopting IV or CACE analysis, has to be understood before taking up a cmRCT.

Differential Item Functioning on the Cochin Hand Function Scale Among People With Systemic Sclerosis by Language, Sex, and Disease Subtype: A Scleroderma Patient-Centered Intervention Network (SPIN) Cohort Study

OBJECTIVE

To evaluate the degree that the Cochin Hand Function Scale (CHFS) generates scores that are comparable across language, sex, and disease subtype.

METHODS

We included participants enrolled in the Scleroderma Patient-centered Intervention Network (SPIN) Cohort who completed the CHFS at their baseline assessment between April 2014 and September 2020. Confirmatory factor analysis (CFA) was used to test unidimensionality, and multiple indicator multiple cause (MIMIC) models were used for differential item functioning (DIF) analysis based on language, sex, and disease subtype. Both intraclass correlation coefficient (ICC) and Pearson's correlation were calculated using factor scores obtained from unadjusted and DIF-adjusted MIMIC models to evaluate agreement and correlation between scores.

RESULTS

A total of 2,155 participants were included. CFA with covarying error terms supported a good fit of the model (χ2 [127] = 1,754.671; P < 0.001; Tucker-Lewis index = 0.985; comparative fit index = 0.987; root mean square error of approximation = 0.077). Nine items displayed statistically significant DIF for language of administration, 10 items for sex, and 10 items for disease subtype. However, the overall impact of DIF was negligible when comparing factor scores that did and did not account for DIF (ICC = 0.999; r = 0.999).

CONCLUSION

The CHFS has score comparability in systemic sclerosis regardless of participants' language, sex, and disease subtype.