Reply to: Deprescribing is associated with reduced readmission to hospital: An updated meta-analysis of randomized controlled trials

Abstract

We would like to thank Lee and colleagues for their interest in our study.¹ In their Letter,² they raised two key points for discussion: (1) the lack of inclusion of some RCT data in the systematic review and the following meta-analysis; (2) the inclusion of data coming from nonrandomized studies. Accordingly, they conducted a sensitivity analysis on rehospitalization rates restricted to RCTs only, incorporating additional data from two RCTs.^{3, 4}

Concerning the first point, we did not include the MedSafer Study³ in the meta-analysis because it did not meet our selection criteria. Specifically, we restricted inclusion criteria to older patients hospitalized in geriatric or internal medicine wards, while excluding those studies that evaluated specifically deprescribing in other specialized settings (e.g., cardiology wards, benzodiazepines deprescribing in psychiatric wards as well as general practice settings). In the Supplementary Material of the abovementioned Canadian study, it was reported that the study was carried out in medical units and, as such, we assumed that also wards other than geriatrics and internal medicine were included. Studies excluded for similar reasons are shown in Table 1. On the other hand, we did not include the study by Franchi et al.⁴ as it did not explicitly report the number of events (e.g., rehospitalization), while Lee and colleagues included this study only based on estimates employing a Kaplan–Meier tails method.

The second raised issue is the inclusion of nonrandomized data in the meta-analysis. While Lee and colleagues indicated that we also included observational studies, we included only experimental and quasi-experimental studies,⁵ in agreement with previously published meta-analyses.⁶ We did not include observational prospective cohort studies.

We agree that meta-analyses of RCTs only may yield more robust results; however, we also decided to include quasi-experimental studies in our meta-analysis, as our risk of bias assessments of selected RCTs highlighted limitations concerning randomization and patient allocation. For instance, in the study conducted by Wehling et al.⁷ the authors noted that the main reason for heterogeneity between control and intervention groups was related to randomization restrictions due to limited availability of beds in the included wards. Furthermore, we included only six quasi-experimental studies, which overall had a low weight in our meta-analysis.

Based on the considerations above, we further extended the sensitivity analysis conducted by Lee and colleagues by including quasi-experimental studies and reporting each study-specific estimate as a hazard ratio (HR), carefully re-evaluating reported estimates from the studies by Nielsen et al., Gallagher et al., and Legrain et al. If the HR was not explicitly published, we contacted the corresponding authors of the studies included in the systematic review to request the raw data to calculate HRs. If such information was not provided, we calculated the HR indirectly from event probabilities and the associated variance for log HR.⁸ A fixed-effects model was chosen to achieve a pooled estimate as: (a) our objective was to estimate a common effect size for the studied population (i.e., older patients hospitalized in geriatrics and internal medicine wards); (b) the number of studies in specific subgroups was limited; and (c) between-studies heterogeneity was minimal or absent.⁹

The results of the updated meta-analysis confirmed that medication review and deprescribing interventions were statistically associated with a 15% reduced hazard of readmission at 1–3 months as compared with the control group (HR 0.85; 95% CI 0.74–0.98; p = 0.023; I² = 0%) (Table 1). When considering the entire follow-up periods, the analysis still showed a trend toward a reduced hazard, with a dilution effect of longer follow-up periods (HR 0.94; 95% CI 0.89–1.01; p = 0.073; I² = 0%).

To enhance the reliability of our findings, we also transformed all measurements into relative risk (RR) values, as previously done in a Cochrane review on the same topic.¹⁰ Consistently, a statistically significantly reduced probability of readmission in the intervention group was found at 1–3 months (RR 0.86; 95% CI 0.78–0.96; p = 0.006; I² = 0%) and a similar dilution effect due to longer follow-up periods was observed, although the overall risk estimate was statistically significant (RR 0.95; 95% CI 0.91–0.98; p = 0.003; I² = 0%).

In all analyses, medication review and deprescribing interventions were not associated with a reduced mortality rate, confirming our previous findings.

In conclusion, our additional sensitivity analyses, also including quasi-experimental studies and estimating RRs, confirm once more our main findings as well as results from Lee and colleagues, showing a protective effect of medication review and deprescribing on rehospitalization, especially during the first three months after the hospital discharge.

GT supervised the study. AF analyzed the data. MC and SC wrote the first draft of the manuscript. All authors revised the manuscript and have made important intellectual contributions. All authors have approved the manuscript for submission. The corresponding author (GT) attests that all listed authors meet authorship criteria and that no others meeting the criteria have been omitted.

GT has served, over the last 3 years, on advisory boards/seminars funded by Sanofi, MSD, Eli Lilly, Sobi, Celgene, Daichii Sankyo, Novo Nordisk, Gilead, and Amgen; he is also a scientific coordinator of the academic spin-off “INSPIRE srl”, which has received funding from several pharmaceutical companies (i.e., PTC Pharmaceuticals, Kiowa Kirin, Shonogi, Shire, Novo Nordisk, and Daichii Sankyo) for conducting observational studies. Additionally, he is currently a consultant for Viatris in a legal case concerning an adverse reaction to sertraline. None of these listed activities are related to the topic of the article. The other authors have no relevant affiliations or financial involvement with any organization or entity with a financial interest in or financial conflict with the subject matter or materials discussed in the manuscript. This includes employment, consultancies, honoraria, stock ownership or options, expert testimony, grants or patents received or pending, or royalties.

No role.