Statistics from Altmetric.com
We appreciated Safiri and Mansournia for their interest and comment on our recent study.1 2 Safiri et al 1 mentioned that we did not explain the steps for the selection of confounders that have been included in our fully adjusted models. We acknowledged that due to the limited number of the words for the manuscript required by the journal, the content for the explanation for selection of confounders was not included. Actually, all confounders in our study were selected according to the published literature,3–5 and these were the most important risk factors for mortality. In total, 27 confounders were included in our fully adjusted models, and the adjustment for confounding was sufficient.
As suggested by Safiri and Mansournia,1 we examined the balance of confounding variables between the glucosamine users and non-users by using standardised mean difference. The most of confounding variables were considered balanced between the two groups (table 1), suggesting that the application of propensity scores was appropriate. We evaluated the proportional hazard assumptions for propensity scores and the outcomes, and no evidence of a violation of the assumption was observed. When we put propensity scores into the models as a covariate to adjust for baseline differences, the results were not substantially changed.
The proportional hazard assumptions for glucosamine use and outcomes were evaluated for all models, and no violation of the assumption was found. Due to 10 models for five outcomes used in our study, we did not provide the hazard curves and p values in the article.2 However, we stated clearly that no violation of the assumption was found in the section of statistical analysis in the article.2
We did not provide p values for the differences in the baseline characteristics of the participants between the two groups due to a very large sample size. Even if there was a minor difference, the test also showed statistically significant results. All the p values for the differences in the baseline characteristics were less than 0.001 in this study, and all the baseline variables were included in the fully adjusted models.
Finally, in this study, we excluded participants who withdrew from the study (1299) and those with missing data on the use of glucosamine (6160). We do agree that differential reasons for loss to follow-up could lead to selection bias. However, the baseline characteristics of the participants excluded and included in this study were generally similar, and no differential reason for loss to follow-up was found.
Patient consent for publication
Contributors All authors critically reviewed the manuscript for important intellectual content.
Funding The authors have not declared a specific grant for this research from any funding agency in the public, commercial or not-for-profit sectors.
Competing interests None declared.
Patient and public involvement Patients and/or the public were not involved in the design, conduct, reporting or dissemination plans of this research.
Provenance and peer review Commissioned; internally peer reviewed.
If you wish to reuse any or all of this article please use the link below which will take you to the Copyright Clearance Center’s RightsLink service. You will be able to get a quick price and instant permission to reuse the content in many different ways.