American Journal of Epidemiology Advance Access originally published online on January 27, 2008
American Journal of Epidemiology 2008 167(6):751; doi:10.1093/aje/kwm399
| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
LETTERS TO THE EDITOR |
RE: "HEPATITIS C VIRUS INFECTION AND THE DEVELOPMENT OF TYPE 2 DIABETES IN A COMMUNITY-BASED LONGITUDINAL STUDY"
From the Department of Biostatistics and Epidemiology, City of Hope National Medical Center, Duarte, CA 91010
(e-mail: cbehrendt{at}coh.org)
With their 7-year, prospective study of incident type 2 diabetes in a community with high prevalence of hepatitis C virus (HCV), Wang et al. (1) have generated valuable data for studying the contribution of HCV to the development of diabetes. However, I believe that the statistical methodology was inappropriate and that two of the three stated conclusions are erroneous.
Because of baseline characteristics (age 
65 years in 24.7 percent, hepatitis B virus in 13.3 percent, HCV in 18.7 percent), the cohort studied by Wang et al. was at risk of not only diabetes but also death without prior diabetes. The number of such deaths during follow-up was not reported but was likely substantial. Standard methods of survival analysis assume that event types are independent, in which case it is appropriate to censor events other than the event of interest (2). This assumption is violated when subjects are at risk of event types that are dependent ("competing risks"), where one type of event precludes the other (i.e., diabetes and death without diabetes) or the two are alternate manifestations of one phenomenon (i.e., toxicity and lack of efficacy). When death is a competing outcome, censoring deaths limits the analysis to describing the risk of event in imaginary persons who can never die, not in persons who could fail from either the event or death (3). Because the probabilities of survival, death, and the event of interest together equal 1, censoring deaths has the effect of overestimating the survival probability, the cumulative incidence of the event of interest, or both (2). Most importantly, if both competing outcomes are associated with the risk factor of interest, censoring deaths can distort the association between the risk factor and the event of interest (i.e., HCV and diabetes) (4). To avoid biased results, appropriate methodologies and statistical programs have been developed (4–6).
The current study and prior cross-sectional studies that stratified by age have found that the association between HCV and diabetes is significant only in persons below the age of 60 or 65 years (1, 7, 8). Wang et al. (1) attribute this finding to effect modification by age, concluding that "the younger the persons ..., the greater the risk [that HCV will cause diabetes]" (1, p. 200). However, a likelier interpretation is that the finding is an artifact of survival bias against older persons with advanced HCV disease, who may have been disproportionately removed by death both before and during the study.
By concluding that "HCV infection and overweight or obesity showed a synergetic effect on the development of diabetes" (1, p. 200), the authors misinterpret the simple additive effect of two risk factors (overweight or obesity and HCV) as a synergistic interaction. On the contrary, as shown in table 4 of their paper (1), HCV confers a similar hazard of diabetes in each category of body mass index; thus, no interaction between body mass index and HCV is evident. Moreover, no term for such an interaction appears in their final multivariate model (table 5).
 |
ACKNOWLEDGMENTS |
|---|
Conflict of interest: none declared.
|
References |
|---|
 TOP References |
|---|
- Wang CS, Wang ST, Yao WJ, et al. Hepatitis C virus infection and the development of type 2 diabetes in a community-based longitudinal study. Am J Epidemiol (2007) 166:196–203.
[Abstract/Free Full Text] - Kim HT. Cumulative incidence in competing risks data and competing risks regression analysis. Clin Cancer Res (2007) 13:559–65.
[Abstract/Free Full Text] - Logan BR, Zhang MJ, Klein JP. Regression models for hazard rates versus cumulative incidence probabilities in hematopoietic cell transplantation data. Biol Blood Marrow Transplant (2006) 12:107–12.[Web of Science][Medline]
- Gray RJ. A class of K-sample tests for comparing the cumulative incidence of a competing risk. Ann Stat (1988) 16:1140–54.
- Fine JP, Gray RJ. A proportional hazards model for the subdistribution of a competing risk. J Am Stat Assoc (1999) 94:496–509.[CrossRef][Web of Science]
- Gray B. The cmprsk package. Vienna, Austria: The Comprehensive R Archive Network, Vienna University of Economics and Business Administration. (2007) (http://cran.r-project.org/doc/packages/cmprsk.pdf).
- Wang CS, Wang ST, Yao WJ, et al. Community-based study of hepatitis C virus infection and type 2 diabetes: an association affected by age and hepatitis severity status. Am J Epidemiol (2003) 158:1154–60.
[Abstract/Free Full Text] - Behrendt CE, Ruiz RB. Hyperglycemia among persons with hepatitis C: not the classical diabetic phenotype. Diabetes Res Clin Pract (2006) 71:68–74.[CrossRef][Web of Science][Medline]
CiteULike 
Connotea 
Del.icio.us What's this?
This article has been cited by other articles:
|
|
 |
|
  C.-S. Wang, S.-T. Wang, and P. Chou THREE AUTHORS REPLY Am. J. Epidemiol., March 15, 2008; 167(6): 751 - 752. [Full Text] [PDF]
|
|
| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||