Statistical analysis
Descriptive statistics were used and expressed as mean (standard deviation), median (range), or count (frequency). Demographic and clinical characteristics were compared between the two groups using thet test and chi-square test. The mean difference and odds ratio with their corresponding 95% confidence intervals (CIs) were used to evaluate the effect size between groups.
Patients with a long history of ITP often received multiple drugs successively or simultaneously. To assess the effect of HCQ on platelet counts in this context, we regressed platelet counts against HCQ treatment using a mixed-effects model with adjustment for potential confounding factors: sex, age at onset, and treatments which may affect platelet counts (including corticosteroids, IVIG, TPO-RAs, rituximab, and splenectomy). In this model, multiple measures within one individual were set as random-effects intercept. Explanatory variables included main effects of HCQ and ANA level, and an interaction term between HCQ and ANA level.
IVIG and rituximab were drugs used in a discontinuous mode, and platelet counts on the initiation date of the two drugs did not reflect their therapeutic effect in the mixed-effects model, which would weaken their effect on platelet counts. Thus, we performed a sensitivity analysis to check the robustness of main findings. In the sensitivity analysis, the effective duration of IVIG was defined to be one week, according to the clinical experience that patients with chronic ITP usually had poor response to IVIG, and the effective duration of rituximab was defined to be one year.
The t test and chi-square test were conducted using the SPSS Statistics software (version 26.0, IBM Corp). The mixed-effects model was performed using R (version 4.1.3, R Foundation, www.r-project.org, Vienna, Austria) with lme4 and lmerTest packages.13, 14 A two-sided p value of less than 0.05 was considered statistically significant.