Nordström P, Pedersen NL, Gustafson Y, et al. Risks of myocardial infarction, death, and diabetes in identical twin pairs with different body mass indexes. JAMA Intern Med 2016;176:1522–9.
To Download a PDF of the Full Article:
Objective. To evaluate whether higher BMI alone contributes to risk of cardiovascular disease (CVD) and death.
Study design. Cohort study of weight-discordant monozygotic twin pairs
Setting and participants. This study took place in Sweden, using a subset of data from the Swedish Twin Registry and the Screening Across Lifespan Twin (SALT) study, which aimed to screen Swedish twins born prior to 1958 for the development of “common complex diseases.” From a total of 44,820 individuals, the current study limited to a subset of 4046 monozygotic twin pairs where both twins had self-reported height and weight data, and where calculated body mass index (BMI) was discordant between the twins, defined as a difference > 0.01 kg/m2. No other inclusion or exclusion criteria are mentioned. Data for the study were collected from several different sources, including telephone interviews (eg, height and weight, behaviors such as physical activity and smoking), national registries on health conditions (eg, myocardial infarction [MI], stroke, diabetes) or prescriptions (eg, diabetes medications), the national causes of death register, and a nationwide database containing socioeconomic variables (eg, income and education). The primary exposure of interest for this study was weight status, categorized as “leaner” or “heavier,” depending on the relative BMI of each twin in a given pair. “Leaner” twins were assumed to have lower adiposity than their “heavier” counterparts, and yet to have identical genetic makeup, thereby allowing the authors to eliminate the contribution of genetic confounding in evaluating the relationship between weight status and CVD risk. The classification system could mean that one person with a BMI of 26 kg/m2 would be placed in the “leaner” category if their twin had a BMI of 28, while someone else in another twin pair but also with a BMI of 26 kg/m2 might be classified in the “heavier” category if their twin had a BMI of 22. Twin pairs were followed for up to 15 years to assess for incident outcomes of interest, with baseline data collected between 1998 and 2002, and follow-up through 2013.
Main outcome measures. The primary outcome of interest was the occurrence of incident MI or death from any cause. As above, these outcomes were assessed using national disease and death registries spanning 1987-2013, and ICD-9 or -10 codes of interest. A secondary outcome of incident diabetes was also specified, presumably limited to development of type 2 diabetes mellitus, and identified using the same datasets, as well as the national prescription registry. Kaplan-Meier curves for incident MI and death were constructed comparing all “leaner” twins against all “heavier” twins, and Cox proportional hazards modeling was used to compare the hazard of the primary composite outcome between groups. Logistic regression was used to evaluate the odds of each outcome including diabetes incidence, and several models were built, ranging from an unadjusted model to one adjusting for a number of lifestyle factors (eg, smoking status, physical activity), baseline health conditions, and sociodemographic factors.
The authors separately examined risk of MI/death in the subgroup of twins where the “heavier” twin had a BMI ≤ 24.9 kg/m2 at baseline (ie, despite being labeled “heavier” they still had a technically normal BMI), and examined the impact of weight trajectory prior to the defined baseline (eg, they were able to incorporate into models whether someone had been actively gaining or losing weight over time prior to the baseline exposure categorization). The authors also conducted several sensitivity analyses, including running models excluding twins with < 1 year of follow-up in an effort to insure that results of the main analysis were not biased due to differential loss to follow-up between exposure categories.
Results. Of the 4046 twin pairs in this study, 56% (2283 pairs) were female, and mean (SD) age at baseline was 57.6 (9.5) years. Race/ethnicity was not reported but presumably the vast majority, if not all, are non-Hispanic white, based on the country of origin. In comparing the group of “heavier” twins to “leaner” twins, several important baseline differences were found. By design, the “heavier” twins had significantly higher mean (SD) BMI at study baseline (25.9 [3.6] kg/m2 vs. 23.9 [3.1] kg/m2) and reported greater increases in BMI over the 15–20 years preceding baseline (change since 1973 was +4.3 [2.9] BMI units for “heavier” twins, vs. +2.6 [2.6] for “leaner” twins). Smoking status differed significantly between groups, with 15% of “heavier” twins reporting they were current smokers versus ~21% of “leaner” twins. “Leaner” twins were also slightly more active than their “heavier” counterparts (50.4% reported getting “rather much or very much” exercise versus 46.5%). The groups were otherwise very similar with respect to marital status, educational level, income, and baseline diagnoses of MI, stroke, diabetes, cancer or alcohol abuse.
In fully adjusted models over a mean (SD) 12.4 (2.5)-year follow-up, “heavier” twins had a significantly lower odds of MI or death (combined) than “leaner” twins (odds ratio [OR] 0.75, 95% CI 0.63–0.91). Because the “heavier” vs. “leaner” dichotomy did not map to clinical definitions of overweight or obesity, the investigators also examined this primary outcome among subgroups with more clinical relevance. Being “heavier” actually had the greatest protective effect against MI/death (OR 0.61, 95% CI 0.46–0.80) among pairs where the so-called “heavier” twin had a normal BMI (< 25.0 kg/m2), and this subgroup appeared to be driving the overall finding of lower odds of MI/death in the “heavier” group as a whole. This pattern was underscored when examining the subgroup of twin pairs where the “heavier” twin had a BMI ≥ 30 kg/m2 at baseline – in this group the protective effect of being “heavier” disappeared (OR 0.92, 95% CI 0.60 to 1.42). Besides not always reflecting clinically relevant weight categories, the “heavier” vs. “leaner” twin dichotomy could, in some cases, amount to a very small difference in BMI between twins (anything > 0.01 unit counted as discordant). As such, the investigators sought to examine whether their results held up when looking at pairs with a higher threshold for BMI discordance (1.0 to 7.0 units or more difference between twins), finding that risk of MI or death did not increase among the “heavier” group in these more widely split twin pairs, even when adjusting for smoking status and physical activity.
In contrast to the MI/mortality analyses, “heavier” twins did have significantly greater odds of developing diabetes during follow-up compared to their “leaner” counterparts (OR 1.94, 95% CI 1.51 to 2.48, adjusted for smoking and physical activity). Also unlike the MI/death analyses, this relationship of increased diabetes risk among “heavier” twins was enhanced by increasing BMI dissimilarity between twins, and among twins who had been gaining weight prior to baseline BMI measurement.
Sensitivity analyses excluding twins with less than 1 year of follow-up did not result in changes to the main findings—“heavier” twins still had similar odds of MI/death as “leaner” twins.
Conclusion. The authors conclude that among monozygotic twin pairs, where the possibility for genetic confounding has been eliminated, obesity is not causally associated with increased risk of MI or death, although the results do support an increased risk of developing incident diabetes among individuals with higher BMI.
Obesity is a known risk factor for many chronic conditions, including diabetes, osteoarthritis, sleep apnea, and hypertension . However, the relationship between obesity and cardiovascular outcomes, particularly coronary artery disease and death from heart disease, has been more controversial. Some epidemiologic studies have demonstrated reduced mortality risk among patients with obesity and heart failure, and even among those with established coronary artery disease—the so-called “obesity paradox” . Others have observed that overweight older adults may have lower overall mortality compared to their normal weight counterparts . On the other hand, it is known that obesity increases risk for diabetes, which is itself a clear and proven risk factor for CVD and death.
As the authors of the current study point out, genetic confounding may be a potential reason for the conflicting results produced in studies of the obesity–CVD risk relationship. In other words, patients who have genes that promote weight gain may also have genes that promote CVD, through pathways independent of excess adipose tissue, with these hidden pathways acting as confounders of the obesity–CVD relationship. By studying monozygotic twin pairs, who have identical genetic makeup but have developed differential weight status due to different environmental exposures, the investigators designed a study that would eliminate any genetic confounding and allow them to better isolate the relationship between higher BMI and CVD. This is an important topic area because, at a population level, we are faced with an immense number of adults who have obesity. Treatment of this condition is resource intense and it is critical that patients and health care systems understand the potential risk reduction that will be achieved with sustained weight loss.
The strengths of this study include the use of a very unique dataset with longitudinal measures on a large number of monozygotic twin pairs, and the authors’ ability to link this dataset with nationwide comprehensive datasets on health conditions, health care use (pharmacy), sociodemographics, and death. Sweden’s national registries are quite impressive and permit these types of studies in a way that would be very difficult to achieve in the United States, with its innumerable separate health care systems and few data sources that contain information on all citizens. Because of these multiple data sources, the authors were able to adjust for some important lifestyle factors that could easily confound the weight status-MI/death relationship, such as smoking and physical activity. Additionally, their models were able to factor in trajectory of weight on some individuals prior to baseline, rather than viewing baseline weight only as a “snapshot” which could risk missing an important trend of weight gain or loss over time, with important health implications.
There are several limitations of the study that are worth reviewing. First, and most importantly, as pointed out in a commentary associated with the article, the categorization of “leaner” and “heavier” can be somewhat misleading if the true question is whether or not excess adiposity is an independent driver of cardiovascular risk . BMI, at the individual level, is not an ideal measure of adiposity and it does not speak to distribution of fat tissue, which is critically important in evaluating CVD risk . For example, 2 siblings could have identical BMIs, but one might have significantly more lean mass in their legs and buttocks, and the other could have more central adipose tissue, translating to a much higher cardiovascular risk. Measures such as waist circumference are critical factors in addition to BMI to better understand an individual’s adipose tissue volume and distribution.
Although the authors did adjust for some self-reported behaviors that are important predictors of CVD (smoking, exercise), there is still potential for confounding due to unscreened or unreported exposures that differ systematically between “leaner” and “heavier” twins. Of note, smoking status—probably the single most important risk factor for CVD—was missing in 13% of the cohort, and no imputation techniques were used for missing data. Another limitation of this study is that its generalizability to more racial/ethnically diverse populations may be limited. Presumably, the patients in this study were non-Hispanic white Swedes, and whether or not these findings would be replicated in other groups, such as those of African or Asian ancestry, is not known.
Finally, the finding that “heavier” twins had greater odds of developing diabetes during follow-up is certainly consistent with existing literature. However, it is also known that diabetes is a strong risk factor for the development of CVD, including MI, and for death . This raises the question of why the authors observed an increased diabetes risk yet no change in MI/death rates among heavier twins. Most likely the discrepancy is due to inadequate follow-up time of incident diabetes cases. Complications of diabetes can take a number of years to materialize, and, with an average of 12 years’ total follow-up in this study, there simply may not have been time to observe an increased risk of MI/death in heavier twins.
Applications for Clinical Practice
For patients interested in weight loss as a way of reducing CVD risk, this paper does not support the notion that lower body weight alone exerts direct influence on this endpoint. However, it reinforces the link between higher body weight and diabetes, which is a clear risk factor for CVD. Therefore, it still seems reasonable to advise patients who are at risk of diabetes that improving dietary quality, increasing cardiorespiratory fitness, and losing weight can reduce their long-term risk of CVD, even if indirectly so.
—Kristina Lewis, MD, MPH
1. Jensen MD, Ryan DH, Apovian CM, et al. 2013 AHA/ACC/TOS guideline for the management of overweight and obesity in adults: a report of the American College of Cardiology/American Heart Association Task Force on Practice Guidelines and The Obesity Society. Circulation 2014;129(25 Suppl 2):S102–138.
2. Antonopoulos AS, Oikonomou EK, Antoniades C, Tousoulis D. From the BMI paradox to the obesity paradox: the obesity-mortality association in coronary heart disease. Obes Rev 2016;17:989–1000.
3. Flegal KM, Kit BK, Orpana H, Graubard BI. Association of all-cause mortality with overweight and obesity using standard body mass index categories: a systematic review and meta-analysis. JAMA 2013;309:71–82.
4. Davidson DJ, Davidson MH. Using discordance in monozygotic twins to understand causality of cardiovascular disease risk factors. JAMA Intern Med 2016;176:1530.
5. Amato MC, Guarnotta V, Giordano C. Body composition assessment for the definition of cardiometabolic risk. J Endocrinol Invest 2013;36:537–43.
6. The Emerging Risk Factors Collaboration, Seshasai SR, Kaptoge S, et al. Diabetes mellitus, fasting glucose, and risk of cause-specific death. N Engl J Med 2011;364:829–41.