Re: Possible involvement of Philip Morris statistical consultant in analysis of Enstrom/Kabat data
Dear Dr Hirschhorn,
If you wish me to answer questions, it would seem more appropriate to write to me directly than to ask the questions in a journal without even drawing the existence of such a letter to my attention. However, I will explain the situation.
I met Enstrom for the first time in 2000 at a meeting which Philip Morris organised in Richmond, Virginia. We both gave talks. One of his talks concerned an analysis based on very long term follow up of the California CPS-I dataset, but was not to do with ETS. I may have said that it would be a good idea to study risk associated with ETS in this dataset – I cannot remember now – but I certainly was not involved in the actual analyses leading to the paper he published with Kabat in the BMJ in 2003. Indeed the first I ever knew of such a paper was when it appeared.
I have, as you say, commented on Enstrom’s publications before. This was as part of a routine series of reviews of relevant epidemiological papers I have conducted for the industry over many years. The fact that I may have criticised one publication does not necessarily imply weaknesses of other publications. The methodological issue with the Enstrom/Kabat paper cited by Bero et al (in the paper to which your letter replies) is that there was no real “unexposed” group in the CPS-I dataset. Surely this is true of virtually all ETS studies? Enstrom and Kabat showed that, among those never smoking men and women in their studies who survived to 1999, those who in 1959 reported they had a spouse who currently smoked were very much more likely to report at home or at work ETS exposure in 1999 than did those who in 1959 reported they had a spouse who had never smoked. Clearly their exposure index did indicate higher ETS exposure in the exposed group and their finding of a lack of relationship with lung cancer, CHD or COPD risk should not be dismissed – it should be taken into consideration along with the evidence from all the other studies.
Your letter suggests that I may "develop" analyses to my "liking", implying that I may give a deliberately biased interpretation of the data. I find this insulting in that I always take great pains to try to give an unbiased scientific view of evidence. I note the massive attempt by the anti-smoking industry to rubbish the Enstrom/Kabat study. Could this reflect the fact that its results were not to their "liking"?