Peter Lee wants to maintain scientific integrity in research, but his understanding of “integrity” seems very different to mine. I was a named principal investigator on the study. Professor Adlkofer advised him that “no-one was better informed of the progress of the study…than Dr Yano…Professor Wynder and I…would find it advisable if Dr Yano published the results as lead author”.¹ Despite this, he used my data in his paper without my consent, crediting me only with “assistance provided in Japan”.² Now, beyond acknowledging that I organised the study and collected the data, and asserting that the charge of misappropriation is unjustified without offering an explanation, he makes no comment on his failure to contact me. Let the record stand then, that he used my data without my consent, and when given an opportunity to comment on this, failed to acknowledge this or apologise.

The conclusions of Lee’s review of misclassification bias³ were not supported by this study where there were more self reported smokers with low (< 100 ng/mg) cotinine/creatinine ratio (CCR) compared to self reported non-smokers with high CCR. Proctor apparently momentarily agreed (“I certainly agree that in its original form it overstated some points”), with the draft of 31 October 1992 withdrawing the conclusion of higher misclassification of female smokers as non-smokers in Japan.⁴

Lee writes that I claim that the “misclassification of self-reported smokers as non-smokers (8/78  =  10.3%) and of non-smokers as smokers (28/318  =  8.8%) are similar” arguing that “this erroneously uses self report, not CCR, as the ‘gold standard’. One uses CCR to detect true smokers and hence calculate misclassification rates”.

Here Lee seems confused with the calculation formula. His definition of misclassification was obtained by dividing those with > 100 ng/mg CCR (n  =  28) by self reported non-smokers (n  =  318). He should certainly apply the same rule to the other side by dividing those with < 100 ng/mg CCR (n  =  8) by self reported smokers (n  =  78). If Lee claims that CCR is the gold standard, then why did he not use it to also detect true non-smokers? He keeps looking away from the fact that the misclassification rate for self reported smokers in this study is higher than the other western study.⁵ I do not know the reason for this. However, Lee’s conclusion of high false non-smokers can be obtained simply by just measuring CCR poorly.

Here, I am compelled to point out that although the smoking status of each sample was not disclosed to those in the tobacco industry who analysed the sample, they knew the purpose of the project. Poor measurement of CCR caused by sample deterioration during transit to the USA or for other reasons could result in both high false non-smokers and false smokers.

Lee defends the “reliability” of his use of CCR to detect misclassified smokers. As he should know, reliability is a measure of the consistency of results after repeated measurements using a test or instrument. What Lee calls “reliability” in his reply refers to validity. Tests of validity examine how closely the results of a measurement correspond to the true state of the phenomenon being measured. Lee’s statement that “with values > 100 ng/mg seen in 90% of smokers and only 9% of non-smokers” indicates that the validity of self reported smoking status in the study was about 90% when > 100 ng/mg of CCR is used as gold standard of smoking—nothing more, nothing less. It was similarly valid in detecting falsely claimed non-smokers as well as falsely claimed smokers. By confusing basic terminology, Lee fails to understand that he cannot use the CCR, his gold standard for smoking, in examining the validity (his reliability) of self reported smoking status. Moreover, when he challenged the self reports of non-smokers, he used CCR, but he did not use it to verify self reported smokers.

Lee claims I overlooked his statement on confounding factors in his report which may have associated environmental tobacco smoke exposure with poor health outcomes by pathways other than causal ones. However, Lee cited confounding results only when the tendency of the (non-significant) data favoured his hypothesis. He did not, for example, consider that households with no smokers consumed less juice and more smoked fish than those with smokers.

Finally, I am astonished to learn that while Lee was involved in discussions with Proctor about the study protocol which included both a phase I and II, that he obtained only the results of phase I and, on his account, has never seen the phase II data.