Send letter to journal:
Re: Testing for gene-environment interaction

Dear Editor,

We read with interest the recent article “Smoking is a risk factor for anti-CCP antibodies only in RA patients that carry HLA-DRB1 Shared Epitope alleles” by Dr. Linn-Rasker and colleagues in Leiden.[1] The question of whether true gene-environment interaction exists between genotype (HLA-DRB1 shared epitope, SE) and an environmental exposure, cigarette smoking in this case, both of which are known to greatly increase the risk of developing RA, is central to our understanding of how these factors predispose to RA. Does the risk associated with having both of these risk factors exceed what would be expected if these risks were independent and additive, and how so? Case-only analyses, such as this, are an efficient and sound method for screening for gene-environment interactions, under the assumption of independence of exposure and genotype in the population.[2,3]

In a recent editorial, Ahlbom and Alfredsson explained the concepts of biological versus statistical interaction.[4,5] Biological interaction is seen when two risk factors are involved in the same pathway to development of disease; statistically this is seen as a departure from additivity of disease rates. For example, Padyukov and colleagues in Sweden have demonstrated that a greater than additive interaction exists for the effects of cigarette smoking and the presence of 0, 1 or 2 copies of the SE.[6] To do so, they calculated the attributable proportion (AP), the proportion of RA among those with both exposures that is attributable to an additive interaction between these risk factors.[7] They found that among current smokers with a double copy of the SE, the AP is 0.7, or 70%, (95% CI 0.4, 0.9).[6] Another way of statistically testing for gene-environment interaction is to create a cross-classified variable, or interaction term, in a logistic regression model, and test it for significance, pointing to a greater than multiplicative interaction that is significant.

In their study, Linn-Rasker and colleagues employed a case-only analysis of patients in the Leiden Early RA clinic to evaluate the interaction between smoking and HLA-SE in predicting the risk of Anti-CCP antibodies. Their findings are very interesting and suggest an interaction between the two exposures, but it does not appear that the presence of a gene-environment interaction was statistically tested. We have employed their data to test statistically for the presence of a gene-environment interaction, both greater than additive and greater than multiplicative. Using the data presented in table 2 of their article, the AP, or the proportion of RA cases positive for anti-CCP antibodies which is attributable to an interaction between cigarette smoking and the presence of the SE, is 0.5, or 50%. (Given our limited access to the data, we are not able to use a re-sampling technique to obtain a 95% confidence interval.) When a cross-classified variable (smoking x HLA-SE) is created and employed in a logistic regression model using the data in Table 2, the term is not significant for a multiplicative interaction (p=0.19). Employing the terminology suggested by Ahlbom and Alfredsson,[4,5] the fact that an additive, but not multiplicative, interaction exists is important and suggests that cigarette smoking and HLA SE lie on the same causal pathway to the development of anti-CCP antibodies and RA. Presenting such statistical analyses promotes translational research between epidemiologists and basic scientists.

Karen H. Costenbader, MD, MPH (A)
Lori B. Chibnik, MPH (A)
Lisa Mandl, MD, MPH (B)
Elizabeth W. Karlson, MD (A)

(A) Brigham and Women’s Hospital, Harvard Medical School, Boston, MA
(B) Hospital for Special Surgery, Weill Cornell Medical College, New York, NY

References

1. Linn-Rasker SP, van der Helm-van Mil AH, Van Gaalen FA, Kloppenburg M, de Vries R, le Cessie S, Breedveld FC, Toes RE, Huizinga TW. Smoking is a risk factor for anti-CCP antibodies only in RA patients that carry HLA-DRB1 Shared Epitope alleles. Ann Rheum Dis 2005.

2. Botto LD, Khoury MJ. Commentary: facing the challenge of gene- environment interaction: the two-by-four table and beyond. Am J Epidemiol 2001; 153:1016-20.

3. Gatto NM, Campbell UB, Rundle AG, Ahsan H. Further development of the case-only design for assessing gene-environment interaction: evaluation of and adjustment for bias. Int J Epidemiol 2004; 33:1014-24. 4.

Andersson T, Alfredsson L, Kallberg H, Zdravkovic S, Ahlbom A. Calculating measures of biological interaction. Eur J Epidemiol 2005; 20:575-9.

5. Ahlbom A, Alfredsson L. Interaction: A word with two meanings creates confusion. Eur J Epidemiol 2005; 20:563-4.

6. Padyukov L, Silva C, Stolt P, Alfredsson L, Klareskog L. A gene- environment interaction between smoking and shared epitope genes in HLA-DR provides a high risk of seropositive rheumatoid arthritis. Arthritis Rheum 2004; 50:3085-92.

Send letter to journal:
Re: No association of smoking with anti-CCP antibodies in RA after adjustment for rheumatoid factor

Dear Editor,

We were interested to read the article by Linn-Rasker and colleagues who report that smoking is a risk factor for anti-cyclic-citrullinated (anti-CCP) antibodies only in RA patients that carry HLA-DRB1 shared epitope (SE) alleles.[1] They claim that an interaction is found between tobacco exposure (TE) and carriage of the SE which leads to anti-CCP production in RA. They find a similar effect for presence of rheumatoid factor (RF), but suggest that the interaction is primarily with the anti- CCP antibody response. Although an attractive hypothesis, our analysis of the data presented does not agree with their interpretation.

We have used the data provided in Table 2 to investigate the independent and combined affects of RF, TE and SE (independent variables) on the presence of anti-CCP antibodies (dependent variable) using multivariate logistic regression models. Firstly we examined for evidence of interaction. In a model which includes TE only there is significant association with the presence of anti-CCP (OR 1.6, 95% CI 1.01 – 2.6, p = 0.04). Addition of SE to the model indicates that TE and SE are independently associated with anti-CCP since both are significant (OR 1.7, 95% CI 1.04 – 2.8, p = 0.03, and OR 3.3, 95% CI 2.01 – 5.6, p < 0.0001, respectively). However, if the interaction term (product of TE and SE) is added into the model together with the main effects (TE and SE separately) there is no association between the interaction term and anti-CCP (p = 0.2). This does not rule out the possibility of an additive effect of TE and SE but argues against a multiplicative interaction. In a similar way we have previously shown that RF production is associated with an additive, but not multiplicative interaction between smoking and the HLA- DRB1*0401 SE allele.[2]

We have also investigated more closely the influence of RF on the association between TE and anti-CCP. Analysis of SE+ patients only in a logistic regression model without adjustment for RF suggests that TE is associated with the presence of anti-CCP (OR 2.1, 95% CI 1.2 – 3.9, p = 0.016). However inclusion of RF as well as TE in the same model reveals a strong association with RF (OR 16.1, 95% CI 7.6 – 34.1, p < 0.00001), while the association with TE loses significance (OR 1.7, 95% CI 0.9 – 3.6, p = 0.17).

Our analyses indicate that the apparent association between TE and anti-CCP can be explained by the confounding association of RF with smoking. The dominant association of RF with anti-CCP can be seen in the last section of Table 2 where non smokers carry as much risk of developing anti-CCP (OR 3.83) as smokers (OR 3.86) if they are positive for both the SE and RF. Further examination of the data in Table 2 also shows that, with or without tobacco exposure, SE negative patients who are RF positive actually have a higher risk of developing anti-CCP antibodies than SE positive smokers who are RF negative (68% v 43.3%, OR 2.7 and 66.7% v 43.3%, OR 2.5, respectively).

Our analyses of these data are consistent with a previous preliminary report on the association of RF and SE, but not smoking, with anti-CCP.[3] This was a study on 271 RA patients in which we also examined the association of HLA-DRB1*0401 with the presence of anti-CCP. Our results showed that patients who were smokers were more likely to be positive for anti-CCP antibodies than those who had never smoked. However, inclusion of RF, together with smoking status as independent variables in logistic regression analyses caused loss of significance between smoking and anti- CCP, while RF remained significantly associated (OR 6.2, 95% CI 1.9 -10.2, p<0.0001). Additional inclusion of HLA-DRB1*0401 status in a logistic regression model demonstrated that this was also strongly associated with anti-CCP (OR 4.7, 95% CI 1.9 -11.4, p = 0.0007) independent of RF. These data indicate that although RA smokers are more likely to be positive for anti-CCP this appears to be primarily due to an association with RF in patients who smoke.

Derek L. Mattey
Staffordshire Rheumatology Centre
University Hospital of North Staffordshire
Stoke-on-Trent
Staffordshire
UK
ST6 7AG

David Hutchinson
Royal Cornwall Hospital
Truro
Cornwall
UK
TR1 3LJ

Correspondence to Dr DL Mattey. d.l.mattey{at}keele.ac.uk

References

1. Linn-Rasker SP, van der Helm-van Mil AHM, Van Gaalen FA, Kloppenburg M, de Vries R, LE Cessie S, et al. Smoking is a risk factor for anti-CCP antibodies only in RA patients that carry HLA-DRB1 shared epitope alleles. Ann Rheum Dis published online 13 Jul 2005;doi:10.1136/ard.2005.041079.

2. Mattey DL, Dawes PT, Clarke S, Fisher J, Brownfield A, Thomson W, Hajeer AH, Ollier WER. Relationship among the HLA-DRB1 shared epitope, smoking, and rheumatoid factor production in rheumatoid arthritis. Arthritis Rheum 2002;47:403-407.