Propensity score matching and stratification using multiparty data without pooling.

Matching and stratification based on confounding factors or propensity scores (PS) are powerful approaches for reducing confounding bias in indirect treatment comparisons. However, implementing these approaches requires pooled individual patient data (IPD). The research presented here was motivated by an indirect comparison between a single-armed trial in acute myeloid leukemia (AML), and two external AML registries with current treatments for a control. For confidentiality reasons, IPD cannot be pooled. Common approaches to adjusting confounding bias, such as PS matching or stratification, cannot be applied as 1) a model for PS, for example, a logistic model, cannot be fitted without pooling covariate data; 2) pooling response data may be necessary for some statistical inference (e.g., estimating the SE of mean difference of matched pairs) after PS matching. We propose a set of approaches that do not require pooling IPD, using a combination of methods including a linear discriminant for matching and stratification, and secure multiparty computation for estimation of within-pair sample variance and for calculations involving multiple control sources. The approaches only need to share aggregated data offline, rather than real-time secure data transfer, as required by typical secure multiparty computation for model fitting. For survival analysis, we propose an approach using restricted mean survival time. A simulation study was conducted to evaluate this approach in several scenarios, in particular, with a mixture of continuous and binary covariates. The results confirmed the robustness and efficiency of the proposed approach. A real data example is also provided for illustration.

Improved survival with enasidenib versus standard of care in relapsed/refractory acute myeloid leukemia associated with IDH2 mutations using historical data and propensity score matching analysis.

BACKGROUND: The present study evaluated the relative survival benefits associated with enasidenib and current standard of care (SoC) therapies for patients with relapsed/refractory (R/R) acute myeloid leukemia (AML) and an isocitrate dehydrogenase 2 (IDH2) mutation who are ineligible for hematopoietic stem cell transplantation (HSCT). METHODS: Propensity score matching (PSM) analysis compared survival outcomes observed with enasidenib 100 mg daily in the phase I/II AG221-C-001 trial and SoC outcomes obtained from a real-world chart review of patients in France. RESULTS: Before matching, enasidenib (n = 195) was associated with numerically improved overall survival (OS) relative to SoC (n = 80; hazard ratio [HR], 0.82; 95% confidence interval [CI], 0.61-1.11). After matching and adjusting for covariates (n = 78 per group), mortality risk was significantly lower with enasidenib than with SoC (HR, 0.67; 95% CI, 0.47-0.97). The median OS was 9.26 months for enasidenib (95% CI, 7.72-13.24) and 4.76 months for SoC (95% CI, 3.81-8.21). Results remained robust across all sensitivity analyses conducted. CONCLUSIONS: PSM analyses indicate that enasidenib significantly prolongs survival relative to SoC among patients with R/R AML and an IDH2 mutation who are ineligible for HSCT. Future prospective studies are needed to validate these findings using other data sources and to assess the comparative efficacy of enasidenib for other treatment outcomes.

Randomization with a posteriori constraints: description and properties.

The use of randomization for assigning patients to treatment groups in clinical trials is firmly acknowledged as providing the best quality results. Two standard methods are used in order to achieve well-balanced groups with respect to prognostic factors (i.e. factors influencing the disease outcome): stratification and minimization. Stratification is recommended when the number of strata is not too high--otherwise, minimization is preferred. However, minimization may compromise blinding (since the search for balance is performed a priori) and, furthermore, use of the technique has been questioned by the European Agency for the Evaluation of Medicinal Products. We have developed a new procedure for adaptive randomization, which we have named 'randomization with a posteriori constraints'. By using a search for balance a posteriori, this procedure ensures that patient groups are similar with respect to prognostic factors while being less vulnerable to selection bias. The aim of this work was to describe the new method and to compare it (using simulations) with stratification and minimization. In the case of trials with few prognostic factors, the recourse to minimization or 'randomization with a posteriori constraints' does not appear to be useful. In such a context, stratification has suitable properties and its simplicity of implementation encourages its use. However, when the number of prognostic factors is higher, 'randomization with a posteriori constraints' is less predictable than minimization and the chance of imbalance is lower than for stratification. In conclusion, 'randomization with a posteriori constraints' with an adequate threshold seems to be a good compromise between minimization and stratification.