Article Text

Download PDFPDF

Should a paper with erroneous interpretations based on invalid measurements be published?
  1. E Yano
  1. Department of Hygiene and Public Health, Teikyo University School of Medicine, Tokyo, Japan
  1. Correspondence to:

Statistics from

Request Permissions

If you wish to reuse any or all of this article please use the link below which will take you to the Copyright Clearance Center’s RightsLink service. You will be able to get a quick price and instant permission to reuse the content in many different ways.

In response to Mr Lee’s comment1 which follows previous responses2,3 and my paper,4 I offer further explanation to resolve an apparent misunderstanding of the validity and reliability of cotinine/creatinine ratio (CCR) measurement and his mishandling of the formula of misclassification. I also express concerns about the lack of scientific integrity in his reporting5 of the Japanese spousal study, including his authorship.

As I demonstrated,4 all indices of nicotine exposure (ambient room, personal sampler monitors, and salivary cotinine) were well correlated but correlated poorly with CCR, raising doubts about the validity of the CCR measurement. Yet Lee maintains that CCR measurement in this study was the gold standard for distinguishing true smokers from falsely reporting smokers.

There are several possibilities about why the CCR measurement may have been invalid and unreliable in this study. In 1991 when I sent the urine samples to the RJ Reynolds laboratory (where the measurement was performed), I was informed that all the dry ice sent with the sample had sublimated before it reached the laboratory. This suggests that the sample was not maintained at low temperature before analysis. Cotinine measurement is temperature sensitive and measurement after the sample is exposed to high temperature can make the measurement inaccurate.6

As I calculated,4 the misclassification and reverse misclassification were equally high suggesting inappropriateness of the CCR measurements as the gold standard. Lee’s neglect of reverse misclassification thus allows him to claim an inflated false negative rate of smoking. Lee continues to justify his misclassification formula by referring to his previous use of the formula. However, this formula is dependent on the prevalence of smoking among the study population and thereby artificially inflates the misclassification rate of populations with low smoking prevalence. By way of illustration, consider two hypothetical populations of 1000 people each with smoking rates of 10% (A) and 30% (B). Suppose that, due to the inaccurate CCR measurement, just 3% of true smokers are classified as non-smokers by erroneously low CCR and 3% of true non-smokers are classified as smokers by erroneously high CCR (for the sake of simplification, I assume no false reports by the subjects). We will get the results shown in table 1.

Table 1

 Hypothetical populations with 3% inaccurate CCR measurement

As can be seen, Lee’s formula for misclassification is dependent on the prevalence of smoking. With only a slight (3%) inaccuracy in CCR measurement, he can thereby easily get more than three times higher (21% v 6%) “misclassification” in a population with lower smoking prevalence, such as with Asian women.

After a long discussion between Proctor and me, Proctor finally understood and accepted my point on the misclassification formula.7 Our final draft of the misclassification paper,8 which Proctor sent to me on 9 November 1992 with my name as a sole author, clearly mentioned the high proportion of misclassification in both sides (self reported non-smoking subjects with high CCR and self reported smokers with low CCR).

Lee insists that reverse misclassification is relatively unimportant in his “abundant” mathematical publications. However, I note that he seems to have realised his mistake of using 28/106 as the misclassification rate of self reported smokers in his original study,5 having quietly switched to 28/98 for this rate1 after I pointed out his confusion. Despite his claim that reverse misclassification is implausible, it was observed as a fact.

Lee states that as far as he is aware “the data never belonged to Yano”. He should be aware that I developed the questionnaire, and selected the study areas and subjects. I supervised the survey at the study area (Shizuoka), erroneously referred to in Lee’s paper as “Shizoka”.5 I planned and ordered the data input, performed the data analysis, and sent the disc to Proctor. On learning from the experience of possible sample damage (from dry ice sublimation) by the commercial shipment at the first phase study in 1991, I even transported the second phase samples myself to the RJ Reynolds laboratory, in Winston Salem, North Carolina, where CCR was measured. I discussed the scientific content of the study with Proctor many times and he accepted my points7 and revised the draft many times, always with my name as the author, and never with Lee’s. As can be seen in the final draft,8 Proctor and I reached a certain agreement on the misclassification formula and the importance of the reverse misclassification rate.

Because Lee never participated in the actual survey it may be that he was unaware of details of the research such as the integrity of the sample which may have seriously affected the interpretation of results. Nor did he participate in the discussion which led Proctor and I to a deeper understanding of the analysis.7 Despite this, still Lee claims that because he proposed the research project, he has a right to sole authorship regardless of who actually conducted the research. This is a unique idea that few scientists would accept.

Lee states: “Had I not published the paper it seems that the findings would never have appeared in the public domain at all. Did Yano also have sole rights to suppress the findings?” Again, I remind Lee that Proctor and I agreed that the results did not indicate high misclassification in self report non-smokers but some failure in the study.7,8 What both Proctor and I prepared for publication, although Proctor ceased to contact me before we could reach a final agreement, was totally different from what Lee eventually published.5 I consider that a description of a failed study involving the inaccurate measurement of CCR was undeserving of publication. Moreover, as a scientist committed to truth, I have a responsibility to be critical of a report with erroneous interpretations based on invalid measurements.