The importance of distinguishing between the odds ratio and the incidence rate ratio in GWAS

BMC Medical Genetics

Asger Hobolth¹  Mette Nyegaard³  Carsten Bøcker Pedersen²  Berit Lindum Waltoft¹ 

[1] Bioinformatics Research Center, Aarhus University, C.F. Møllers Allé 8, Aarhus C, 8000, Denmark;The Centre for Integrated Register-based Research, Aarhus University, CIRRAU, Arhus University, Aarhus, Denmark;Department of Biomedicine, Aarhus University, Vennelyst Boulevard 4, Aarhus C, 8000, Denmark

关键词: Rare disease assumption;    Conditional logistic regression;    Logistic regression;    Competing risk;    Matched case-control study;    Study design;    Genome wide association study;   

Others  :  1223151
DOI  :  10.1186/s12881-015-0210-1

 received in 2014-06-10, accepted in 2015-08-10,  发布年份 2015

【 摘 要 】

Background

In recent years, genome wide association studies have identified many genetic variants that are consistently associated with common complex diseases, but the amount of heritability explained by these risk alleles is still low. Part of the missing heritability may be due to genetic heterogeneity and small sample sizes, but non-optimal study designs in many genome wide association studies may also have contributed to the failure of identifying gene variants causing a predisposition to disease. The normally used odds ratio from a classical case-control study measures the association between genotype and being diseased. In comparison, under incidence density sampling, the incidence rate ratio measures the association between genotype and becoming diseased. We estimate the differences between the odds ratio and the incidence rate ratio under the presence of events precluding the disease of interest. Such events may arise due to pleiotropy and are known as competing events. In addition, we investigate how these differences impact the association test.

Methods

We simulate life spans of individuals whose gene variants are subject to competing events. To estimate the association between genotype and disease, we applied classical case-control studies and incidence density sampling.

Results

We find significant numerical differences between the odds ratio and the incidence rate ratio when the fact that gene variant may be associated with competing events, e.g. lifetime, is ignored. The only scenario showing little or no difference is an association with a rare disease and no other present associations. Furthermore, we find that p-values for association tests differed between the two study designs.

Conclusions

If the interest is on the aetiology of the disease, a design based on incidence density sampling provides the preferred interpretation of the estimate. Under a classical case-control design and in the presence of competing events, the change in p-values in the association test may lead to false positive findings and, more importantly, false negative findings. The ranking of the SNPs according to p-values may differ between the two study designs.

【 授权许可】

   
2015 Waltoft et al.

【 图 表 】

Fig. 1.

Fig. 2.

Fig. 3.

Fig. 4.

Fig. 5.

【 参考文献 】

• [1]Stranger BE, Stahl EA, Raj T. Progress and promise of genome-wide association studies for human complex trait genetics. Genetics. 2011; 187(2):367-383.
• [2]Ligthart L, Hottenga JJ, Lewis CM, Farmer AE, Craig IW, Breen G, et al. Genetic risk score analysis indicates migraine with and without comorbid depression are genetically different disorders. Hum Genet. 2014;133(2):173–86.
• [3]Simonson MA, Wills AG, Keller MC, McQueen MB. Recent methods for polygenic analysis of genome-wide data implicate an important effect of common variants on cardiovascular disease risk. BMC Med Genet. 2011; 12:146. BioMed Central Full Text
• [4]Wray NR, Lee SH, Mehta D, Vinkhuyzen AA, Dudbridge F, Middeldorp CM. Research review: polygenic methods and their application to psychiatric traits. J Child Psychol Psychiatry. 2014; 55(10):1068-1087.
• [5]Dudbridge F. Power and predictive accuracy of polygenic risk scores. PLoS Genet. 2013; 9(3): Article ID e1003348
• [6]Andersen PK, Geskus RB, de Witte T, Putter H. Competing risks in epidemiology: possibilities and pitfalls. Int J Epidemiol. 2012; 41(3):861-870.
• [7]Allignol A, Schumacher M, Wanner C, Drechsler C, Beyersmann J. Understanding competing risks: a simulation point of veiw. BMC Medical Research Methodology 2011, 11(86). doi:10.1186/1471-2288-11-86
• [8]Pearce N. What does the odds ratio estimate in a case-control study. Int J Epidemiol. 1993;22(6):1189–92.
• [9]Clayton D, Hills M. Statistical models in epidemiology. Oxford University Press Inc., New York; 1998.
• [10]Sistrom CL, Garvan CW. Proportions, odds, and risk. Radiology. 2004; 230:12-19.
• [11]Wacholder S, Silverman DT, McLaughlin JK, Mandel JS. Seletion of controls in case-control studies. III. Design options. Am J Epidemiol. 1992; 135(9):1042-1050.
• [12]Prentice RL, Breslow NE. Retrospective studies and failure time models. Biometrika. 1978; 65(1):153-158.
• [13]Beyersmann J, Latouche A, Buchholz A, Schumacher M. Simulating competing risks data in survival analysis. Stat Med. 2009; 28(6):956-971.
• [14]Bender R, Augustin T, Blettner M. Generating survival times to simulate Cox proportional hazards models. Stat Med. 2005; 24(11):1713-1723.
• [15]Rosthøj S, Andersen PK, Abildstrom SZ. SAS macros for estimation of the cumulative incidence functions based on a Cox regression model for competing risks survival data. Comput Methods Programs Biomed. 2004; 74(1):69-75.
• [16]Breslow NE, Day NE. Statistical Methods in Cancer Research. Volume I—The Analysis of Case-Control Studies. International Agency for Research on Cancer (IARC Scientific Publications No. 32), Lyon; 1980.
• [17]Hoffmann-Jørgensen J. Probability With a View Towards Statistics, Volume 1. Chapmann & Hall, New York; 1994.
• [18]Fradin DD, Fallin MD. Influence of control selection in genome-wide association studies: the example of diabetes in the Framingham Heart Study. BMC Preceedings. 2009; 3(7):S113. BioMed Central Full Text
• [19]Wang M-H, Shugart YY, Cole SR, Platz EA. A simulation study of control sampling methods for nested case-control studies of genetic and molecular biomarkers and prostate cancer progression. Cancer Epidemiol Biomarkers Prev. 2009; 18(3):706-711.
• [20]Greenland S, Thomas DC. On the need for the rare disease assumption in case-control studies. Am J Epidemiol. 1982; 116(3):547-553.
• [21]Fine JP, Gray RJ. A proportional hazards model for the subdistribution of a competing risk. J Am Stat Assoc. 1999; 94(446):496-509.
• [22]Karon JM, Kupper LL. In defense of matching. Am J Epidemiol. 1982; 116(5):852-866.
• [23]Kupper LL, Karon JM, Kleinbaum DG, Morgenstern H, Lewis DK. Matching in epidemiologic studies: validity and efficiency considerations. Biometrics. 1981; 37(2):271-291.
• [24]Rose S, Laan MJ. Why match? Investigating matched case-control study designs with causal effect estimation. Int J Biostat. 2009; 5(1):1.
• [25]Thomas DC, Greenland S. The relative efficiencies of matched and independent sample designs for case-control studies. J Chronic Dis. 1983; 36(10):685-697.
• [26]Schwartz S, Susser E. Genome-wide association studies: does only size matter? Am J Epidemiol. 2010; 167(7):741-744.
• [27]Pedersen CB, Mortensen PB, Cantor-Graae E. Do risk factors for schizophrenia predispose to emigration? Schizophr Res. 2011; 127(1–3):229-234.