The Authors Reply

Abstract

Yu-Wen Hu raised a question about immortal time effect and how it relates to patients with end-stage renal disease who are prescribed Chinese herbal medicine.1Hu Y.-W. Chinese herbal medicine use and risk of end-stage renal disease in patients with chronic kidney disease: is there an immortal time bias?.Kidney Int. 2016; 90: 227-228Abstract Full Text Full Text PDF PubMed Scopus (2) Google Scholar This concern is common, particularly in observational studies that focus on the effect of treatment. Our reply follows. First, a time-dependent analysis classifying a patient's treated and untreated period is commonly used to describe immortal time. This analysis has some limitations; one is that the interruption of treatment should be subject to random censoring.2Suissa S. Immortal time bias in observational studies of drug effects.Pharmacoepidemiol Drug Saf. 2007; 16: 241-249Crossref PubMed Scopus (361) Google Scholar Unfortunately, no patient in our cohort who started dialysis in the treatment period implied that interruption of Chinese herbs was against random censoring. This is why we used time stratification. Second, active-comparator design is also suggested for handling immortal time bias.3Kowall B. Stang A. Rathmann W. et al.No reduced risk of overall, colorectal, lung, breast, and prostate cancer with metformin therapy in diabetic patients: database analyses from Germany and the UK.Pharmacoepidemiol Drug Saf. 2015; 24: 865-874Crossref PubMed Scopus (83) Google Scholar By defining an active comparator as one treated with other herb formulas, we have consistent results in our published data (Table 1). Third, we recalculated follow-up time for the treated group since treatment was initiated and considered the immortal time period as untreated.4Liu J. Weinhandl E.D. Gilbertson D.T. et al.Issues regarding ‘immortal time’ in the analysis of the treatment effects in observational studies.Kidney Int. 2012; 81: 341-350Abstract Full Text Full Text PDF PubMed Scopus (44) Google Scholar This approach tends to underestimate the treatment effect; however, we had an unadjusted cause-specific hazard ratio of 0.73 (95% confidence interval 0.65–0.82) and 0.90 (95% confidence interval 0.80–1.01) after adjusted covariates.Table 1CSHRs of end-stage renal disease by use of various classes of prescribed Chinese herbal medicineClassesBefore matchedPropensity score-matchedaCSHR (95% CI)aModel adjusted for age, sex, insurance amount, region, urbanization of residence, comorbidities, Charlson comorbidity index score, diabetic drugs, antihypertensive drugs, nonsteroidal anti-inflammatory drugs, analgesic drugs other than nonsteroidal anti-inflammatory drugs, anti-lipid drugs, number of outpatient visits, and various classes of prescribed Chinese herbal medicine.P valueCSHR (95% CI)P valueTonic formulas1.07 (0.85–1.35)0.551.04 (0.8–1.34)0.77Blood-regulating formulas0.73 (0.59–0.91)0.0050.81 (0.63–1.04)0.09Heat-clearing formulas0.88 (0.7–1.09)0.230.94 (0.73–1.21)0.63Exterior-releasing formulas0.85 (0.68–1.06)0.150.84 (0.65–1.09)0.19Dampness-dispelling formulas1.47 (1.18–1.83)<0.0011.31 (1.03–1.68)0.03Wind dampness–dispelling formulas0.70 (0.56–0.87)0.0010.70 (0.55–0.89)0.004Phlegm-dispelling formulas0.88 (0.69–1.13)0.310.8 (0.6–1.06)0.11Sedative formulas0.97 (0.75–1.25)0.811.04 (0.78–1.4)0.77Qi-regulating formulas0.67 (0.52–0.86)0.0020.78 (0.59–1.05)0.10Harmonizing formulas0.66 (0.52–0.84)<0.0010.63 (0.48–0.83)0.001Downward draining formulas0.97 (0.76–1.23)0.770.86 (0.65–1.14)0.29Dryness-relieving formulas0.88 (0.69–1.13)0.320.89 (0.67–1.18)0.42Cold-dispelling formulas1.22 (0.96–1.55)0.111.14 (0.86–1.51)0.37Exterior- and interior-releasing formulas0.90 (0.7–1.17)0.440.72 (0.52–0.99)0.05Astringent formulas0.88 (0.67–1.15)0.350.88 (0.64–1.22)0.45Purgative formulas1.63 (1.25–2.12)<0.0011.71 (1.26–2.33)<0.001Cough-suppressing and panting-calming formulas0.98 (0.73–1.32)0.910.93 (0.66–1.33)0.71Liver-pacifying and wind-extinguishing medicinals1.09 (0.79–1.51)0.601.19 (0.81–1.74)0.37Summer heat-clearing formulas0.73 (0.52–1.02)0.060.66 (0.44–0.99)0.05Orifice-opening formulas0.66 (0.39–1.12)0.130.78 (0.43–1.4)0.40Shen-calming formulas0.70 (0.37–1.34)0.280.62 (0.25–1.52)0.29Formulas that treat abscesses and sores0.83 (0.53–1.28)0.400.89 (0.54–1.48)0.66Antiparasitic formulas1.13 (0.61–2.1)0.700.83 (0.36–1.93)0.67Interior-warming formulas0.87 (0.12–6.29)0.89<0.01 (0.00 to >100)0.97Emetic formulas<0.01 (0.00–>100)0.97<0.01 (0.00 to >100)0.99Undetermined formulas0.86 (0.7–1.07)0.180.88 (0.68–1.13)0.30This table only includes user group and reset follow-up time to the date of treatment initiation.aCSHR, adjusted cause-specific hazard ratio; CI, confidence interval; CSHR, cause-specific hazard ratio.a Model adjusted for age, sex, insurance amount, region, urbanization of residence, comorbidities, Charlson comorbidity index score, diabetic drugs, antihypertensive drugs, nonsteroidal anti-inflammatory drugs, analgesic drugs other than nonsteroidal anti-inflammatory drugs, anti-lipid drugs, number of outpatient visits, and various classes of prescribed Chinese herbal medicine. Open table in a new tab This table only includes user group and reset follow-up time to the date of treatment initiation. aCSHR, adjusted cause-specific hazard ratio; CI, confidence interval; CSHR, cause-specific hazard ratio. With the findings above, immortal time bias might have little effect on the association of Chinese herb medicine use with end-stage renal disease in our study. The causal relationship between Chinese herb medicine and renal progression or protection needs further study. Chinese herbal medicine use and risk of end-stage renal disease in patients with chronic kidney disease: is there an immortal time bias?Kidney InternationalVol. 90Issue 1PreviewRecently, Lin et al. reported that Chinese herbal medicine use is associated with a significantly reduced risk of end-stage renal disease in patients with chronic kidney disease, with a hazard ratio of 0.41.1 Furthermore, there seems to be a dose-response relationship—the longer the duration of treatment, the lower the risk of end-stage renal disease. However, “immortal time bias” might explain this result.2,3 Full-Text PDF Open Archive