Thursday, December 15, 2011

Karolinska Institute Study: Heart of Failure

I have described in detail (here, here, and here) studies from Stockholm’s Karolinska Institute (KI) that have driven smokeless tobacco regulation worldwide. These studies contain important and troubling discrepancies that KI researchers have refused to address. Rather, they continue to defy international scientific principles of data sharing, which stipulate that scientific results are open to challenge by other scientists to determine their accuracy and integrity.

Some of the same KI researchers, led by Gabriel Arefalk, recently authored another study, alleging that snus use causes heart failure (here). The article was published in the European Journal of Cardiovascular Prevention and Rehabilitation.

As in the past, KI focused on snus users in the Swedish Construction workers cohort. This time, construction workers who were snus users didn’t have a significantly higher risk for heart failure than nonusers. But the researchers also analyzed the Uppsala Longitudinal Study of Adult Men (ULSAM), reporting that snus use “was associated with a more than doubled risk for subsequent heart failure” in this group.

It appears that the ULAM analysis was tailor-made to produce the desired result, and it is evident that the KI researchers employed highly unorthodox methods. I attempted to draw attention to these serious problems by submitting a letter to the journal editor. My submission was rejected, “on grounds that this commentary is a re-review of the analyses of the paper.” Ignoring my systematic description of the study’s deficiencies, the editor suggested that I “obtain data and write another scientific paper for the journal.” This is particularly ironic, since I have tried for three years to obtain data from KI.

Following is my letter to the European Journal of Cardiovascular Prevention and Rehabilitation.

To the Editor:
Study of Snus Use and Heart Failure: Problems Requiring Resolution

The article by Arefalk et al. (1) reported that snus use was “a significant predictor of heart failure” among 70 year-old men in the Uppsala Longitudinal Study of Adult Men (ULSAM) (hazard ratio, HR = 2.09, 95% confidence interval, CI = 1.00 – 4.39). However, there are numerous problems that raise questions about the validity of the findings.

Among the 78 snus users at baseline, 62 (79%) were current smokers; in the referent group of 998 snus non-users, only 175 (18%) were current smokers. The large difference in smoking makes it extremely unlikely that the results were fully adjusted for current smoking, which is an important confounding factor (2). This problem was compounded by the authors’ highly unorthodox adjustment procedure.

The authors apparently did not utilize conventional categories of current, former and never smoking, even though it was clearly possible to do so according to the ULSAM questionnaire (3). Instead, they stated that “[s]moking was adjusted for by using a current smoking dose variable [none, < 10 or > 10 cigarettes per day] as well as a pack-year variable [never, < 33 and > 33 pack-years].” But this created a major problem: the 504 former smokers were combined with the 335 never smokers in the current smoking dose variable, and they were combined with the 237 current smokers in the pack-years of smoking variable (Table 1, Column labeled Total Sample). Thus, in the authors’ adjusted models former smokers were simultaneously pooled both with never smokers and with current smokers, two groups that ought to be mutually exclusive.

Although Arefalk et al. failed to define or categorize current and former smokers in the ULSAM cohort, the latter group appeared in the discussion: “…we performed a secondary analysis…further subdividing former smokers into those who quit smoking less than vs. more than 10 years before baseline.” Mentioning former smokers in the last third of the discussion without ever defining them is incomprehensible. In addition, the authors’ 10-year cut-point for smoking cessation was different from the 5-year cut-point in the questionnaire (3). The authors need to provide a complete explanation of the analysis that they employed.

Arefalk et al. reported that there were 78 snus users, which is consistent with the number on the ULSAM website (questionnaire item Z480, “Do you use snus?”; yes, n=79)(3). They also described 237 current smokers (167 moderate and 70 heavy), which was consistent with responses to the question asked during a euglycemic hyperinsulinemic clamp investigation, “Do you smoke?” (Z085; yes, n=245), but not with responses to the questionnaire item “Do you smoke?”(Z158; yes, n=173) (3). It would be helpful to understand the differences in these responses.

Arefalk et al. also estimated the risk of heart failure among snus users in the Swedish Construction Workers Cohort (CWC), and they evaluated the effect of dose among the 75 current snus users with heart failure (Table 4). They concluded that “[n]o clear dose-response relationship was observed…” In contrast, Arefalk et al. reported that “the limited sample size did not permit the study of dose-response relations” in the ULSAM cohort. This important information, which was available for the 14 snus users with heart failure (3), may be very informative even if it was inadequate for a formal analysis. For example, perhaps all 14 cases of heart failure were among users of very low doses. Arefalk et al. should not withhold this information.

There were numerous formatting errors in the tables. For example, in Table 1 the variables Diabetes prevalence, ECG-left ventricular hypertrophy, Body mass index, Office systolic blood pressure and Antihypertensive medication use were listed under Pack-years of smoking, and the variables Myocardial infarction before baseline and Myocardial infarction during follow-up were listed under Alcohol use. There were similar errors in Table 4.

Arefalk et al. concluded that they “observed an increased risk for subsequent heart failure” among male snus users in the ULSAM, but the conclusion is neither legitimate nor persuasive until the authors resolve the fundamental questions about the analysis.


1. Arefalk G, Hergens M-P, Ingelsson E, Ärnlöv J, Michaëlsson K, Lind L, Ye W, Nyrén O, Lambe M and Sundström J. Smokeless tobacco (snus) and risk of heart failure: results from two Swedish cohorts. European Journal of Cardiovascular Prevention and Rehabilitation 2011, DOI: 10.1177/1741826711420003 (here).

2. Dunlay SM, Weston SA, Jacobsen SJ, and Roger VL. Risk factors for heart failure: a population-based case-control study. American Journal of Medicine 2009; 122: 1023-1028 (here).

3. ULSAM-70 Questionnaire and Response Statistics (here).

Brad Rodu
Professor of Medicine
Endowed Chair, Tobacco Harm Reduction Research
University of Louisville

No comments: