Utility of Restricted Mean Survival Time for Analyzing Time to Nursing Home Placement Among Patients With Dementia

This cohort study assesses the utility of restricted mean survival time as a method for quantifying time to nursing home placement among patients with dementia.


Introduction
Delaying nursing home (NH) placement for patients with dementia is an important goal.The effect of a treatment for dementia on NH placement is conventionally summarized as a hazard ratio, risk difference, or median time difference.2][3] Despite its advantages over conventional measures (Table), to date, RMST has not been applied to clinical trials of dementia treatments.We assessed the utility of RMST in analyzing time to NH placement among patients with dementia using the Donepezil and Memantine in Moderate to Severe Alzheimer's Disease (DOMINO-AD) trial as an example. 4,5

Results
Over 4 years, 162 participants were placed in an NH.The Kaplan-Meier curves show that the risk of NH placement among those who discontinued donepezil treatment was initially higher (36.8% vs 21.4% at 1 year) but later became similar to that for those who continued the treatment (76.7% vs 76.5% at 4 years), suggesting a withdrawal effect (Figure).This delay in NH placement could not be inferred or quantified from the hazard ratio (discontinuation vs continuation: 1.29; 95% CI, 0.95-1.75) or risk difference (0.2%; 95% CI, −12.5% to 13.2%) (Table ).Differences in the median NH-free time

Discussion
In this cohort study, unlike hazard ratios and risk differences, RMST quantified a benefit of donepezil continuation in terms of an associated gain or loss in NH-free days.The median NH-free time difference, although intuitive, was insensitive to early or late NH placement and had a wider 95% CI than that did RMST.The 95% CI of the RMST difference did not rule out the possibility of a clinically meaningful delay (up to 240 days) in NH placement.Although RMST analysis assumes noninformative censoring and requires a priori determination of a time window, it does not require the proportional hazards assumption of a Cox proportional hazards regression model, which was violated in the DOMINO-AD trial. 5Because our objective was to assess use of RMST methods, one limitation of our analysis was that it was not intended to comprehensively evaluate benefits vs harms (eg, need for a pacemaker) associated with donepezil or to offer new clinical insights.Nonetheless, for patients with Alzheimer disease, the majority of whom may ultimately require NH placement, our findings suggest that RMST may complement conventional measures of treatment effect by helping to estimate a potential delay in NH placement.

JAMA Network Open | Geriatrics
Restricted Mean Survival for Time to Nursing Home Placement in Patients With Dementia

Table .
6easures of Treatment Effect for Nursing Home Placement in Clinical Trials of Treatments for Dementia aThis cohort study consisted of a post hoc analysis aimed to assess use of RMST methods; thus, we did not obtain data from the DOMINO-AD investigators but reconstructed data by scanning Kaplan-Meier curves from the study by Howard et al5using WebPlotDigitizer software (https://automeris.io/WebPlotDigitizer)andapplying a validated algorithm.6Thisalgorithm reconstructs patient-level data based on the number of patients at risk and the magnitudes and locations of steps in the Kapan-Meier curves.The Hebrew SeniorLife/Advarra Institutional Review Board deemed that institutional review board review was not necessary because the study used reconstructed data.This study followed the Strengthening the Reporting of Observational Studies in Epidemiology (STROBE) reporting guideline.
bHazard ratioHazard rate and ratio of 2 hazard rates (without a reference hazard rate) are not intuitive to interpret, proportional hazards are assumed (treatment effect is constant over time), and statistical power depends on the number of events 1.29 (0.95 to 1.75) the Kaplan-Meier curve, the time window needs to be prespecified, statistical power depends on the exposed follow-up time, and RMST can provide a more precise estimate (ie, narrow CI) in case of a low event rate −108 (−240 to 23) Abbreviations: DOMINO-AD, Donepezil and Memantine in Moderate to Severe Alzheimer's Disease; RMST, restrictive mean survival time.aTreatment effect estimates were calculated from reconstructed DOMINO-AD trial data (see the Methods section for details).bDonepezil discontinuation vs continuation.cFor treatment discontinuation vs continuation, the risk difference was 36.8% vs 21.4% at 1 year and 76.7% vs 76.5% at 4 years.Figure.Continuation and Discontinuation of Donepezil and Nursing Home (NH) Placement-Free Survival in the Donepezil and Memantine in Moderate to Severe Alzheimer's Disease Trial Open Access.This is an open access article distributed under the terms of the CC-BY License.JAMA Network Open.2021;4(1):e2034745.doi:10.1001/jamanetworkopen.2020.34745(Reprinted) January 28, 2021 1/3 Downloaded From: https://jamanetwork.com/ on 09/23/2023 Methods