Time-Dependent Bias in Hepatitis C Classification

To the Editors: Implicit assumptions that hepatitis C virus infection antedates that of HIV and that clearance of hepatitis C is rare in HIV-positive individuals has led to many studies treating hepatitis C infection as time invariant. Changes in patterns of acquisition and clearance of the hepatitis C virus no longer support such assumptions. Increasingly, HIV-positive individuals are infected with hepatitis C virus after HIV,1–3 clearance of the hepatitis C virus is more likely with the availability of direct-acting antivirals4, and testing guidelines for hepatitis C virus now include all HIV-positive patients regardless of perceived risk5,6. Therefore, treatment of hepatitis C-positivity as time invariant can lead to time-dependent bias in the estimated impact of hepatitis C infection on clinical outcomes. To assess potential for time-dependent bias, we evaluated the incidence of hepatitis C infection after antiretroviral therapy (ART) initiation and the rate of clearance using data from HIV-positive participants who had started ART and were enrolled in the Ontario HIV Treatment Network Cohort Study, the details of which are available in the eAppendix (https://links.lww.com/EDE/A873). We then examined the impact of time invariant and time-updated hepatitis C classification on the estimated risk of mortality using multivariable proportional hazards models of time from ART initiation to death hypothesizing that treatment of hepatitis C-positivity as time invariant would lead to underestimation of the risk of death. Hepatitis C-positivity was identified from anti-hepatitis C antibody and/or hepatitis C virus RNA laboratory test results obtained from linkage with the Public Health of Ontario Laboratories, through which virtually all confirmatory hepatitis C virus laboratory testing occurs. Participants who had never been tested for hepatitis C virus were considered to be hepatitis C virus-negative in all analyses. Data were left-truncated between ART initiation and cohort enrolment. Additional details are available in the eAppendix (https://links.lww.com/EDE/A873). As of December 2010, 4,555 study participants had initiated ART. The median duration of follow-up was 2.32 years (interquartile range 1.36–8.01); 701 participants died. Definitive hepatitis C virus test results were available for 3,872 individuals (85%); 735 individuals tested hepatitis C-positive. Seventy-nine participants (11%) seroconverted and 120 participants (16%) cleared the virus. Incidence of mortality did not differ by hepatitis C status when it was considered time invariant (Figure). Correctly attributing person years of follow-up to hepatitis C virus-negative classification before hepatitis C seroconversion led to a significant difference in incidence of mortality. This difference became more pronounced after successively accounting for misclassification after clearance and before hepatitis C testing (Figure).FIGURE: Incidence of death by hepatitis C (HCV) status after successively accounting for excess person-years of follow-up attributed to hepatitis C infection when HCV status is considered time-fixed. Person-years of follow-up were re-appropriated successively; accounted for time before seroconversion, then time after clearance of the virus, and finally for the years before a hepatitis C test.The hazard ratio of death associated with time-updated hepatitis C- positive status was 1.98 (95% confidence interval = 1.53–2.57) after adjusting for age, sex, race, injection drug use as risk factor, hepatitis B virus positivity, baseline smoking status, first regimen type, and time-updated CD4 count and viral load. Classification of hepatitis C infection as time invariant resulted in an attenuation of the effect (adjusted hazard ratio = 1.38 [95% confidence interval = 1.07–1.77]), demonstrating that the assumption that hepatitis C infection is time invariant in HIV-positive individuals can lead to substantial bias. This topic has been addressed in the statistical literature7,8 and in relation to other medical research but needs to be highlighted in infectious disease clinical research due to the frequency of time-varying exposures in this setting. In particular, assessing cumulative exposure to the hepatitis C virus, or consideration of hepatitis C status as negative, positive or past infection may be more appropriate to address the complex nature of viral hepatitis coinfection with HIV. Limitations of our study include incomplete and variable testing frequency for the hepatitis C virus, potential false positive and negative tests, and a lack of confirmatory RNA testing for all antibody positive results. However, the potential for bias has been demonstrated using a large, long-standing cohort of a diverse HIV-positive population. Therefore, we advocate the use of time-updated hepatits C status, where possible, or the impact of such time-dependent bias should be expressly discussed, where not. ACKNOWLEDGEMENTS The names of the OHTN Cohort Study Team and other acknowledgements are provided in the eAppendix. Jennifer Gillis Toronto General Research Institute University Health Network Toronto, ON, Canada Curtis Cooper University of Ottawa The Ottawa Hospital Research Institute Ottawa, ON, Canada Ann N. Burchell Sandra Gardner University of Toronto Toronto, ON, Canada Michael Manno Ontario HIV Treatment Network Toronto, ON, Canada Tony Mazzulli Sean B. Rourke University of Toronto Toronto, ON, Canada Janet M. Raboud Toronto General Research Institute Support, Systems and Outcomes Toronto General Hospital Toronto, ON Canada [email protected] and the OHTN Cohort Study Group

Fetched live from OpenAlex and de-inverted. Abstracts are not stored in this database: the inverted indexes are 8.6 GB of the frame’s 9.3 GB of text, and the host has 13 GB free.