RSS Feeds
-
Re: More Thoughts on Multivariate Regression
Submit responseIn Response
We appreciate Dr. Kim’s comments which highlight some of the challenges inherent to combining genetic and conventional risk factors. Such challenges will persist as new loci with modest effects on risk continue to be discovered.
Limited space did not allow us to provide details on how genetic loci and conventional risk factors were related. As Dr. Kim alludes to, obesity is an important risk factor to attend to in understanding the mechanism by which these loci contribute to diabetes development. None of the 10 loci included in the genetic risk score (GRS) were related to body mass index (BMI) or the remaining conventional risk factors (age, smoking, alcohol, physical activity, menopausal status, family history of diabetes)(data not shown). In the women, adjusting for BMI slightly strengthened the association for GRS, but additional adjustment for other covariates did not appreciably alter the results (Table 1). Similar results were observed for men (data not shown). The study design as well as the modest effect of each of these loci on risk impedes our ability to postulate potential mechanisms by which each locus contributes to risk. Nevertheless, the primary purpose of our study was not to uncover mechanisms, but rather to evaluate the combined effects of these loci and conventional risk factors on risk of the disease and our ability to discriminate between diabetics and non-diabetics.
We believe Dr. Kim may have misinterpreted our discussion regarding the discriminative value of the GRS. We were not suggesting that collinearity may have explained the minimal improvement we observed. We were referring to previous studies (1, 2) which incorporated fasting glucose levels or other measures of insulin sensitivity in their clinical risk models and who observed the least discriminative improvement with the addition of genetic information. No correlation between the GRS and conventional risk factors were observed in our study. Although, the mechanism by which these loci contribute to risk have not been established, it remains possible that some act via similar pathways to those of conventional risk factors while others by novel pathways. Our GRS would not capture interactions between individual loci and conventional risk factors if they exist; a limitation we have highlighted in our paper. While a score which accounts for all possible interactions will certainly perform better than a score which does not; to design and validate such a score will be challenge at this time.
Table 1. Association of Candidate SNP Loci and Risk for Type 2 Diabetes Among Women
SNP
Model 1
OR (95% CI)
Model 2
OR (95% CI)
Model 3
OR (95% CI)
rs1111875
1.01 (1.00-1.02)
1.06 (0.95-1.17)
1.05 (0.94-1.17)
rs7756992
1.16 (1.04-1.28)
1.17 (1.04-1.31)
1.13 (1.00-1.28)
rs4402960
1.23 (1.00-1.02)
1.26 (1.12-1.40)
1.24 (1.10-1.39)
rs13266634
1.18 (1.06-1.30)
1.19 (1.06-1.33)
1.17 (1.04-1.32)
rs10010131
1.10 (1.00-1.22)
1.10 (0.99-1.22)
1.09 (0.97-1.21)
rs564398
1.09 (0.99-1.20)
1.12 (1.01-1.24)
1.11 (1.00-1.24)
rs10811661
1.20 (1.06-1.36)
1.14 (1.00-1.31)
1.12 (0.97-1.30)
rs12255372
1.32 (1.20-1.47)
1.36 (1.22-1.52)
1.35 (1.20-1.51)
rs1801282
1.18 (1.02-1.36)
1.19 (1.02-1.39)
1.17 (0.99-1.37)
rs5219
1.17 (1.06-1.29)
1.17 (1.05-1.30)
1.17 (1.04-1.31)
GRS
Continuous
1.15 (1.12-1.19)
1.16 (1.12-1.20)
1.15 (1.11-1.19)
Quintiles 1
1.00 (ref)
1.00 (ref)
1.00 (ref)
2
1.06 (0.86-1.33)
1.25 (0.97-1.60)
1.20 (0.92-1.56)
3
1.62 (1.31-2.01)
1.60 (1.25-2.04)
1.52 (1.17-1.96)
4
2.06 (1.69-2.52)
1.94 (1.53-2.46)
1.78 (1.39-2.29)
5
2.07 (1.70-2.51)
2.46 (1.95-3.10)
2.26 (1.77-2.90)
Model 1: Adjusted for age
Model 2: Adjusted for age and BMI (5 categories)
Model 3: adjusted for age, BMI (5 categories), family history of diabetes (yes, no), smoking (never, past, current), menopausal status [pre- or post-menopausal (never, past, or current hormone use); women only], alcohol (5 categories), and quintiles of physical activity (hours/wk).
References
1. Lyssenko V, Jonsson A, Almgren P, et al. Clinical risk factors, DNA variants, and the development of type 2 diabetes. N Engl J Med. 2008;359(21):2220-32.
2. Meigs JB, Shrader P, Sullivan LM, et al. Genotype score in addition to common risk factors for prediction of type 2 diabetes. N Engl J Med. 2008;359(21):2208-19.
Conflict of Interest:
None declared
-
More Thoughts on Multivariate Regression
Submit responseMany studies have been done to evaluate risks of diabetes associated with comorbidities or behavioral factors (conventional risk factors) (1) and recently also with some genetic loci (2,3). Cornelis et al. presented combined effects of conventional risk factors and genetic loci for the first time. However, the presented data was not enough to grasp the whole picture of the combined effects. The authors did not give information on how genetic loci and conventional risk factors were related and how the effects of genetic loci were adjusted by conventional risk factors in detail. Above all, they did not show how adding BMI adjusted the effects of genetic loci. Obesity is the most important risk factor for diabetes [1]. In this regard, careful examination of the patterns of adjustment of the effect of each and joint genetic loci by BMI could have given clues on how these genetic loci contribute to the development of diabetes.
Secondly, on discussion of the AUC of having diabetes, the authors mentioned a possibility of collinearity between conventional risk factors and GRS (Genetic Risk Score) as an explanation for marginal contribution of adding GRS. However, the author should have calculated correlation between these if they thought that it was a possibility. And as another explanation, the authors also suggested that the effects of GRS could have been mediated through conventional risk factors. But as a matter of fact, this is opposite to their findings in the results where they already showed that the effects of GRS were significant after adjustment of age and BMI and minimally adjusted by adding other risk factors.
References
1. Perry IJ, Wannamethee SG, Walker MK, Thomson AG, Whincup PH, Shaper AG. Prospective study of risk factors for development of non- insulin dependent diabetes in middle aged British men. British Medical Journal 1995;310(6979):560-4.
2. Sladek R, Rocheleau G, Rung J, et al. A genome-wide association study identifies novel risk loci for type 2 diabetes. Nature 2007;445:881- 5.
3. Zeggini E, Weedon MN, Lindgren CM, et al. Replication of genome- wide association signals in UK samples reveals risk loci for type 2 diabetes. Science 2007;316(5829):1336.
Conflict of Interest:
None declared