We used propensity-score matching to optimize the relevant covariates between TZD users and nonusers (D’Agostino, 1998 (link)). The propensity score for each participant was estimated using non-parsimonious multivariable logistic regression, with TZD use as the dependent variable. We included 35 clinically related covariates as independent variables (Table 1 ). The nearest-neighbor algorithm was adopted to construct matched pairs, assuming the standardized mean difference (SMD) value <0.1 to be a negligible difference between the study and comparison cohorts.
We used crude and multivariable-adjusted Cox proportional hazards models to compare outcomes between TZD users and nonusers. The results were presented as hazard ratios (HRs) and 95% confidence intervals (CIs) for TZD users compared with nonusers. This study is based on the intention-to-treat hypothesis. To calculate the observed risks, we censored the participants until the date of respective outcomes, death, or at the end of follow-up on December 31, 2018, whichever came first. The Kaplan–Meier method and log-rank tests were used to compare the cumulative incidences of hospitalization for all-cause pneumonia, bacterial pneumonia, IMV, and death due to pneumonia during the follow-up time between TZD users and nonusers. We compared the risk of hospitalization for all-cause pneumonia among different subgroups of age, sex, comorbidities, medications (rosiglitazone, pioglitazone, and others) for clinical applicability of results. We also assessed the cumulative duration (<153, 153–549, ≧550 days) and dose (<2,940, 2,940–10,009, ≧10,110 mg) of pioglitazone for the risks of hospitalization for all-cause pneumonia, bacterial pneumonia, IMV, and death due to pneumonia compared with no-use of TZDs to explore the dose relationship. We performed a stratified analysis to see the effect of TZD vs. non-TZD in the risk of all-cause pneumonia stratified by the subgroups of metformin use vs. no-use, SU use vs. no-use, DPP-4 inhibitor use-vs. no-use trying to determine whether other hypoglycemic agents have effect on pneumonia risk; stratified by patient’s resident areas of the Northern, Central, Southern, and Eastern Taiwan trying to see whether the different environmental exposures have different effect on pneumonia risk.
A two-tailed value of p <0.05 was considered significant. SAS (version 9.4; SAS Institute, Cary, NC, United States) was used for statistical analysis.
We used crude and multivariable-adjusted Cox proportional hazards models to compare outcomes between TZD users and nonusers. The results were presented as hazard ratios (HRs) and 95% confidence intervals (CIs) for TZD users compared with nonusers. This study is based on the intention-to-treat hypothesis. To calculate the observed risks, we censored the participants until the date of respective outcomes, death, or at the end of follow-up on December 31, 2018, whichever came first. The Kaplan–Meier method and log-rank tests were used to compare the cumulative incidences of hospitalization for all-cause pneumonia, bacterial pneumonia, IMV, and death due to pneumonia during the follow-up time between TZD users and nonusers. We compared the risk of hospitalization for all-cause pneumonia among different subgroups of age, sex, comorbidities, medications (rosiglitazone, pioglitazone, and others) for clinical applicability of results. We also assessed the cumulative duration (<153, 153–549, ≧550 days) and dose (<2,940, 2,940–10,009, ≧10,110 mg) of pioglitazone for the risks of hospitalization for all-cause pneumonia, bacterial pneumonia, IMV, and death due to pneumonia compared with no-use of TZDs to explore the dose relationship. We performed a stratified analysis to see the effect of TZD vs. non-TZD in the risk of all-cause pneumonia stratified by the subgroups of metformin use vs. no-use, SU use vs. no-use, DPP-4 inhibitor use-vs. no-use trying to determine whether other hypoglycemic agents have effect on pneumonia risk; stratified by patient’s resident areas of the Northern, Central, Southern, and Eastern Taiwan trying to see whether the different environmental exposures have different effect on pneumonia risk.
A two-tailed value of p <0.05 was considered significant. SAS (version 9.4; SAS Institute, Cary, NC, United States) was used for statistical analysis.
Free full text: Click here
