Wednesday, February 9, 2011

The Revolving-Door Cohort at the Karolinska Institute

The Karolinska Institute (KI) in Stockholm is Sweden’s premier medical research institution, and among the most well known and respected in the world. For several years a group of researchers in the KI’s Department of Medical Epidemiology and Biostatistics has published numerous studies of the health risks related to snus use.

As I stated in a recent commentary published in Sweden (link here), “the health risks identified in the KI studies are the raison d’être for the ban on snus in the E.U. (except Sweden), but it has also had a profound impact on tobacco regulation in the U.S. and other countries,” making access to snus difficult or impossible.

The KI studies have profoundly influenced regulatory actions all over the world, but they are compromised by important, troubling and unresolved discrepancies that KI researchers have refused to address. This post describes the biggest problem in detail.

The KI studies are based on over 300,000 male Swedish construction workers who enrolled in a health program from 1971 to 1992, including roughly 135,000 workers who were enrolled in the program during the 1971-1974 period. This group is important, because there are potentially serious questions about the adequacy of tobacco use information for these workers. In fact, the information is so questionable that KI researchers have included and excluded them in a revolving door fashion in published analyses over the last several years.

The story goes back to 1994, when KI investigator Gunilla Bolinder and colleagues reported that snus use was a risk factor for cardiovascular diseases. (abstract here). In that study Dr. Bolinder studied only the construction workers from the 1971-74 group (hereafter, the “Bolinder” cohort). That report, one of only a few linking snus use with heart disease and stroke, had some obvious but inexplicable technical problems, which I raised in a 1995 letter to the editor of the journal (here). Dr. Bolinder did not resolve these problems.

Fast-forward 13 years to 2007, when KI investigators Luo et al. published a high profile study in The Lancet finding that snus use was a risk factor for pancreas cancer (here). In their analysis, Luo et al. excluded ALL of the workers in the Bolinder cohort “because of ambiguities in the coding of smoking status” of participants. In other words, Luo tossed out the 135,000 workers of the Bolinder cohort. The justification for these exclusions was an “unpublished” observation by Zendehdel, another KI investigator.

I published a letter to the editor of The Lancet (here) observing that, if the Bolinder cohort was deficient, then the validity of the Bolinder study was suspect. Luo et al. responded that perhaps the exclusions were not warranted (here), citing a KI study that was about to be published. That study, which found that snus use was associated with esophageal and stomach cancer, included the Bolinder cohort (here). The lead author on that study was Zendehdel.

Confusing? Absolutely. First, KI researchers cited Zendehdel as the justification for excluding the Bolinder cohort, but then Zendehdel included the Bolinder cohort in his published study.

To illustrate how many times KI researchers have subjected the Bolinder cohort to revolving door treatment, here is a list of publications.

The Bolinder Revolving Door Cohort At the Karolinska Institute
YearFirst AuthorJournalDiseaseBolinder Cohort In/Out
1994BolinderAmerican Journal of Public HealthCardiovascularIn
2005OdenbroBritish Journal of CancerSkin cancerIn
2007LuoLancetPancreas cancerOut
2007OdenbroBritish Journal of DermatologyMelanomaIn
2007FernbergCancer ResearchLeukemiaIn
2007HergensJournal of Internal MedicineHeart attackOut
2008ZendehdelInternational Journal of CancerGastrointestinal cancerIn
2008HergensJournal of Internal MedicineHypertensionOut
2010CarlensAmerican Journal of Respiratory and Critical Care MedicineInflammatory DiseasesOut
2011NordenvallInternational Journal of CancerColon-rectal cancerIn

This is unacceptable from a scientific perspective, but there are many other problems. For example, it appears that Dr. Bolinder’s definition of snus use was inconsistent with later definitions. I’ll have more to say about this in a future post.

No comments: