Suchergebnisse

Failure to account for time‐dependent treatment use when developing a prognostic model can result in biased future predictions. We reviewed currently available methods to account for treatment use when developing a prognostic model. First, we defined the estimands targeted by each method and examined their mechanisms of action with directed acyclic graphs (DAGs). Next, methods were implemented in data from 1,906 patients; 325 received selective β‐blockers (SBBs) during follow‐up. We demonstrated seven Cox regression modeling strategies: (a) ignoring SBB treatment; (b) excluding SBB users or (c) censoring them when treated; (d) inverse probability of treatment weighting after censoring (IPCW), including SBB treatment as (e) a binary or (f) a time‐dependent covariate; and (g) marginal structural modeling (MSM). Using DAGs, we demonstrated IPCW and MSM have the best properties and target a similar estimand. In the case study, compared to (a), approaches (b) and (e) provided predictions that were 1% and 2% higher on average. Performance (c‐statistic, Brier score, calibration slope) varied minimally between approaches. Our review of methods confirmed that ignoring treatment is theoretically inferior, but differences between the prediction models obtained using different methods can be modest in practice. Future simulation studies and applications are needed to assess the value of applying IPCW or MSM to adjust for treatments in different treatment and disease settings.

WOS: 000401058500006 ; PubMed ID: 28339826 ; Background. During the follow-up in a randomized controlled trial (RCT), participants may receive additional (non-randomly allocated) treatment that affects the outcome. Typically such additional treatment is not taken into account in evaluation of the results. Two pivotal trials of the effects of hemodiafiltration (HDF) versus hemodialysis (HD) on mortality in patients with end-stage renal disease reported differing results. We set out to evaluate to what extent methods to take other treatments (i.e. renal transplantation) into account may explain the difference in findings between RCTs. This is illustrated using a clinical example of two RCTs estimating the effect of HDF versus HD on mortality. Methods. Using individual patient data from the Estudio de Supervivencia de Hemodiafiltracion On-Line (ESHOL; n = 902) and The Dutch CONvective TRAnsport STudy (CONTRAST; n = 714) trials, five methods for estimating the effect of HDF versus HD on all-cause mortality were compared: intention-totreat (ITT) analysis (i.e. not taking renal transplantation into account), per protocol exclusion (PPexcl; exclusion of patients who receive transplantation), PPcens (censoring patients at the time of transplantation), transplantation-adjusted (TA) analysis and an extension of the TA analysis (TA(ext)) with additional adjustment for variables related to both the risk of receiving a transplant and the risk of an outcome (transplantation-outcome confounders). Cox proportional hazards models were applied. Results. Unadjusted ITT analysis of all-cause mortality led to differing results between CONTRAST and ESHOL: hazard ratio (HR) 0.95 (95% CI 0.75-1.20) and HR 0.76 (95% CI 0.59-0.97), respectively; difference between 5 and 24% risk reductions. Similar differences between the two trials were observed for the other unadjusted analytical methods (PPcens, PPexcl, TA) The HRs of HDF versus HD treatment became more similar after adding transplantation as a time-varying covariate and including transplantation-outcome confounders: HR 0.89 (95% CI 0.69-1.13) in CONTRAST and HR 0.80 (95% CI 0.62-1.02) in ESHOL. Conclusions. The apparent differences in estimated treatment effects between two dialysis trials were to a large extent attributable to differences in applied methodology for taking renal transplantation into account in their final analyses. Our results exemplify the necessity of careful consideration of the treatment effect of interest when estimating the therapeutic effect in RCTs in which participants may receive additional treatments. ; EuDial working group; European Nephrology and Dialysis Institute; Catalan Society of Nephrology; Fresenius Medical Care and Gambro through the Catalan Society of Nephrology; Dutch Kidney Foundation [C02.2019]; Fresenius Medical Care, The NetherlandsNetherlands Government; Gambro Lundia, Sweden; Dr E.E. Twiss Fund, Roche Netherlands; International Society of Nephrology/Baxter Extramural Grant Program; Netherlands Organization for Health Research and Development (ZONMw)Netherlands Organization for Health Research and Development [170882802]; Health Ministry (Programme Hospitalier de Recherche Clinique) ; The HDF Pooling Project was designed, conducted and analyzed independently of the financial contributors of the individual studies as listed below. Study data were collected and retained by the investigators and were not available for the financial contributors of the individual studies. S.A.E.P. and the representatives of the combined authors of the four studies were financially supported by the EuDial working group. EuDial is an official working group of the European Renal Association-European Dialysis Transplant Association (ERA-EDTA; http://era-edta.org/eudial/European_Dialysis_Working_Group.html). No industry funding was received for any part of or activity related to the present analysis.; The Turkish HDF study was supported by the European Nephrology and Dialysis Institute with an unrestricted grant. The study was performed in Fresenius Medical Care hemodialysis clinics in Turkey. ESHOL was supported by the Catalan Society of Nephrology and by grants from Fresenius Medical Care and Gambro through the Catalan Society of Nephrology. The CONTRAST study was supported by a grant from the Dutch Kidney Foundation (Nierstichting Nederland Grant C02.2019) and unrestricted grants from Fresenius Medical Care, The Netherlands, and Gambro Lundia, Sweden. Additional support was received from the Dr E.E. Twiss Fund, Roche Netherlands, the International Society of Nephrology/Baxter Extramural Grant Program and the Netherlands Organization for Health Research and Development (ZONMw Grant 170882802). The French HDF study was supported by a national grant from the Health Ministry (Programme Hospitalier de Recherche Clinique) as a means to improve care and outcomes in elderly chronic disease patients.