Similarities and differences between patients included and excluded from a randomized clinical trial of vitamin d supplementation for improving glucose tolerance in prediabetes: interpreting broader applicability

BACKGROUND: Randomized Clinical Trial (RCT) designs range from highly selective resulting in lack of external validity to more inclusive, requiring large sample sizes to observe significant results. Few publications, however, have compared excluded to enrolled participants. We aimed to assess our trial’s design based on the effectiveness versus efficacy continuum using the Pragmatic-Explanatory Continuum Indicator Summary (PRECIS) tool and to compare included and excluded patients. METHODS: Fifteen members of endocrinology section completed PRECIS for DIVA (D-Vitamin Intervention in VA) trial; an RCT evaluating vitamin D supplementation in improving dysglycemia in patients with prediabetes. Retrospective chart review compared subjects excluded (OUT) to those included (IN) in RCT. Student’s t and Chi-square tests were used to compare continuous and categorical variables. Additionally, multiple logistic regression was completed. RESULTS: PRECIS scores were nearly universally pragmatic. 178 patients enrolled in DIVA trial were compared with 178 randomly selected patients excluded from study involvement. There was no significant difference between IN and OUT for the majority of the continuous and all of the categorical variables. Multivariate logistic regression identified only the A1c, HDL and Charlson Index as significant predictors of a participant’s inclusion or exclusion. There was higher HDL (51.3(13.9) versus 44.6(10.1), P = 0.001) and Charlson Index (2.85(1.6) versus 2.2(1.17), P = 0.001) for OUT versus IN groups. Subanalysis of excluded patients in A1c range 5.7 to 6.9, had lower BMI (30.7(3.4) versus 32(2.7), P = 0.002) but higher HDL (mg/l: 49.7(11.8) versus 44.6(10.1), P = 0.001) and Charlson index (2.85(1.6) versus 2.2(1.17), P = 0.001) than included participants. Additionally, there was a trend towards higher rates of cancer (22.9 % versus 12.9 %, P = 0.033) but less psychiatric problems (56.2 % versus 72.5 %, P = 0.026) and thiazide diuretic use (18.1 % versus 29.8 %, P = 0.034). CONCLUSION: DIVA trial design appears to favor broad clinical applicability. The majority of objectively compared variables did not different between patients included and excluded from this RCT. Advice based on the evidence from this RCT may be applicable to a larger group of patients than those fitting inclusion/exclusion criteria alone. TRIAL REGISTRATION: ClinicalTrials.gov NCT01375660 (registered 15 June 2011). ELECTRONIC SUPPLEMENTARY MATERIAL: The online version of this article (doi:10.1186/s13063-015-0812-0) contains supplementary material, which is available to authorized users.