We very much appreciate Shen’s1 interest in our paper,2 however, Shen’s comments are misleading.
First, Shen’s recalculation of the number of patients who had a death event in Table 3, using data from Table 1, was incorrect. Our study is longitudinal, with up to 10 years of follow-up. Table 1 represents drug exposure at baseline and Table 3 represents exposure during follow-up. Therefore, direct calculation between these two tables is improper. Shen has recalculated the original data from Tables 1 and 3, and this approach results in a misinterpretation of our data.
Second, as our study is longitudinal with repeated measurements (up to 40 quarters), the event numbers of all-cause and fracture-specific admissions could be larger than the patient numbers, meaning a patient could experience more than one all-cause or fracture-specific admission during the follow-up period. That is also the reason we adopt generalized estimating equation (GEE) models with an autoregressive correlation structure to fit our study design. The “no. of patients” in Table 3 is actually “no. of patients who experience a clinical event in an observational time unit” (3 mo in our study), under the framework of GEE models.