- Email: [email protected]
Effects of passive smoking
408
Letters to the Editors
estimate of the upper limit of the percentage of naturally cycling women who experience psychological distress that can be attributed to the premenstrual state. The second point commented on in Dr Dalton’s letter is that “recurrent symptoms cannot be diagnosed on a one-day-only interview”. Our study was not designed to determine recurrence of premenstrual symptoms over several cycles in individual women. However, the excess prevalence of severe negative affect symptoms among women in the premenstrual phase over that among women in the mid-cycle phase (i.e. the attributable risk), which was accounted for solely by women with a history of probable PMS, represents the proportion of women in the population whose psychological distress is linked to the premenstrual state. This group will therefore include women afflicted with PMS as well as women with affective disorders whose symptoms become exacerbated in the premenstrual phase. Since individual women with a clinical diagnosis of PMS (which means that their symptoms are recurrent), may experience some cycles without premenstrual symptom manifestations, the population measure of risk obtained in our study indicates that about 24% (i.e. 4 out of 17, see Table 5), of women with a lifetime history of probable PMS, will likely experience severe distress (i.e. a recurrence), in the premenstrual phase of their current cycle.
We lacked information that would enable us to separate out those women with a diagnosis of affective disorders. However, 18.7% of the women with a history of probable PMS also gave a history of ever having been told by a doctor that they had an emotional disorder, compared with 5.7% of the non-PMS women. Close to the same proportion (i.e. 1 out of 4) of the PMS women who accounted for the observed risk of psychological distress associated with the premenstrual state, also gave a history of emotional disorder. This finding suggests that about one-quarter of the women manifesting severe negative affect attributable to the premenstrual state, were experiencing an exacerbation in the premenstrual phase of a pre-existing affective disorder. S . RAMCHARAN
E. J. LOVE G. H. FICK University
of Calgary Calgary Alberta Canada
REFERENCE 1.
Ramcharan S, Love El, Fick GH, Goldfien A. The epidemiology of premenstrual symptoms in a population-based sample of 2650 urban women: attributable risk and risk factors. J Clin Epidemiol 1992; 45: 377392.
EFFECTS OF PASSIVE SMOKING In the August 1992 issue of the Journal of Clinical Epidemiology (Variance and Dissent, p. 809), statistician Nathan Mantel presents criticisms of the Glantz and Parmley report [l]. That report claims that environmental tobacco smoke causes 53,000 heart and cancer deaths annually among nonsmokers. In their Dissent to Mantel’s criticisms, Glantz and Parmley seize the opportunity to attack Mantel’s character. They paint him against a backdrop of what they refer to as tobacco industry strategy and tactics, stating that “Mantel has played a key role in implementation of such a strategy for over a decade.” Elsewhere in their Dissent, Glantz and Parmley imply that Mantel has used a “common
tactic employed by the tobacco industry,” and they are not surprised that he has “joined other tobacco industry consultants . . . in criticizing our paper . . .” Glantz and Parmley’s attempt to portray Mantel as an important part of tobacco industry strategy does not make sense in light of Mantel’s long-standing view on active smoking. This view dates back to 1958 when he was one of the authors of a study reporting an association between active smoking and lung cancer in women [2]. Furthermore, in a 1989 published report, Mantel stresses that, “There is little residue of doubt about the causal role of cigarette smoking in lung cancer, and even in other morbidities and mortalities” [3].
Letters to the Editors
409
If anything, Mantel’s published views on active smoking make him a foe-not a puppet--of the tobacco industry. On the other hand, he believes strongly that Glantz and Parmley have misinterpreted the evidence, and shown insufficient evidence, that environmental tobacco smoke causes heart and cancer deaths. It is people like Mantel who insure integrity and credibility in the profession of statistics. Glantz and Parmley’s accusation that he is involved in what they describe as tobacco industry tactics is irresponsible and unwarranted. Ordinarily, journals are careful not to publish anything that reflects on the character or integrity of an individual. Perhaps you owe
Mantel a public apology for letting this happen in your journal.
Glantz and Parmley [l], in the acknowledgements in their paper, thanked me for carefully checking their power calculations. Mantel [2], commenting on the paper, stated that I “could not have approved of the statistical power reported”, noting specifically that the cited powers for the studies by Helsing et al. [3] in females and by He [4] appeared to be grossly in error. In reply Glantz and Parmley [5] found Mantel’s assertion “puzzling” since Glantz and I [6] had jointly published a letter in Environment International “confirming this agreement”. I am writing to clarify the true situation. Arising out of correspondence in Environment International [7,8] in which conflicting power figures for my own study on passive smoking and heart disease [9] were published, Glantz and I had a long private correspondence about the issue of power. From this it emerged that Glantz had made a gross error in implementing a computer program and that all his original power calculations had been totally incorrect, including those planned to be used in the Glantz and Parmley paper. Glantz admitted his error [lo], re-ran all his calculations, and produced revised power estimates which I agreed were now more sensible, though in fact still differing somewhat from my own estimates which were derived by a procedure that was statistically more defensible. Nevertheless, as Mantel [2] correctly points out, errors, obvious at a glance to any trained statistician, still exist in the paper as it appeared in Circul@ion. The stated powers and confidence limits for the Helsing et al. [3] and He [4]
studies clearly violently conflict. For the Helsing study the conflict has arisen because Glantz’s revised power estimate of 92% was incorrectly printed as 2%. Glantz and Parmley have not only failed to proof-read adequately, but have not even bothered to check their data when told there were errors. The power for the He study [4] may in fact be approximately correct. Here the error lies in the published 95% confidence limits which themselves conflict with the sample size and other published data from the study. I pointed out this problem with He’s paper (described in fuller detail elsewhere [l 11) to Glantz in our correspondence, but he totally ignored it, and included the erroneous material anyway. It concerns me that I should be improperly cited as agreeing to results which are grossly in error. It concerns me even more that an author of a published paper on a subject of importance should have such an apparent cavalier disregard for the accuracy of the material he presents.
MARTHA PJXSKE Darien, CT U.S.A.
REFERENCES 1. Glantz SA, Parmley WW. Passive smoking and heart
disease-epidemiology, physiology and biochemistry. Circulation 1991; 83: l-12. 2. Haenszel W, Shimkin MB, Mantel N. A retrospective study of lung cancer in women. J Nat1Cancerhst 1958; 21: 825-842. 3. Mantel N. The passive smoking myth. Presentend Futureof IodoorAir Quality, Proc BrusselsConference,
14-16 February 1989. Bieva CJ, Courtois Y, Govaerts M, Eds. Amsterdam: Excerpta Medica; 1989: 155-159.
P. N. LEE Independent Consultant in Statistics Sutton, Surrey SM25DA England
REFERENCES
I. Glantz SA, Parmley WW. Passive smoking and heart disease-epidemiology, physiology and biochemistry. Circdation 1991; 83: l-12. 2. Mantel N. Dubious evidence of heart and cancer deaths due to passive smoking. J Ciin Epiden~iol1992; 45: 809-813.
Letters to the Editors
estimate of the upper limit of the percentage of naturally cycling women who experience psychological distress that can be attributed to the premenstrual state. The second point commented on in Dr Dalton’s letter is that “recurrent symptoms cannot be diagnosed on a one-day-only interview”. Our study was not designed to determine recurrence of premenstrual symptoms over several cycles in individual women. However, the excess prevalence of severe negative affect symptoms among women in the premenstrual phase over that among women in the mid-cycle phase (i.e. the attributable risk), which was accounted for solely by women with a history of probable PMS, represents the proportion of women in the population whose psychological distress is linked to the premenstrual state. This group will therefore include women afflicted with PMS as well as women with affective disorders whose symptoms become exacerbated in the premenstrual phase. Since individual women with a clinical diagnosis of PMS (which means that their symptoms are recurrent), may experience some cycles without premenstrual symptom manifestations, the population measure of risk obtained in our study indicates that about 24% (i.e. 4 out of 17, see Table 5), of women with a lifetime history of probable PMS, will likely experience severe distress (i.e. a recurrence), in the premenstrual phase of their current cycle.
We lacked information that would enable us to separate out those women with a diagnosis of affective disorders. However, 18.7% of the women with a history of probable PMS also gave a history of ever having been told by a doctor that they had an emotional disorder, compared with 5.7% of the non-PMS women. Close to the same proportion (i.e. 1 out of 4) of the PMS women who accounted for the observed risk of psychological distress associated with the premenstrual state, also gave a history of emotional disorder. This finding suggests that about one-quarter of the women manifesting severe negative affect attributable to the premenstrual state, were experiencing an exacerbation in the premenstrual phase of a pre-existing affective disorder. S . RAMCHARAN
E. J. LOVE G. H. FICK University
of Calgary Calgary Alberta Canada
REFERENCE 1.
Ramcharan S, Love El, Fick GH, Goldfien A. The epidemiology of premenstrual symptoms in a population-based sample of 2650 urban women: attributable risk and risk factors. J Clin Epidemiol 1992; 45: 377392.
EFFECTS OF PASSIVE SMOKING In the August 1992 issue of the Journal of Clinical Epidemiology (Variance and Dissent, p. 809), statistician Nathan Mantel presents criticisms of the Glantz and Parmley report [l]. That report claims that environmental tobacco smoke causes 53,000 heart and cancer deaths annually among nonsmokers. In their Dissent to Mantel’s criticisms, Glantz and Parmley seize the opportunity to attack Mantel’s character. They paint him against a backdrop of what they refer to as tobacco industry strategy and tactics, stating that “Mantel has played a key role in implementation of such a strategy for over a decade.” Elsewhere in their Dissent, Glantz and Parmley imply that Mantel has used a “common
tactic employed by the tobacco industry,” and they are not surprised that he has “joined other tobacco industry consultants . . . in criticizing our paper . . .” Glantz and Parmley’s attempt to portray Mantel as an important part of tobacco industry strategy does not make sense in light of Mantel’s long-standing view on active smoking. This view dates back to 1958 when he was one of the authors of a study reporting an association between active smoking and lung cancer in women [2]. Furthermore, in a 1989 published report, Mantel stresses that, “There is little residue of doubt about the causal role of cigarette smoking in lung cancer, and even in other morbidities and mortalities” [3].
Letters to the Editors
409
If anything, Mantel’s published views on active smoking make him a foe-not a puppet--of the tobacco industry. On the other hand, he believes strongly that Glantz and Parmley have misinterpreted the evidence, and shown insufficient evidence, that environmental tobacco smoke causes heart and cancer deaths. It is people like Mantel who insure integrity and credibility in the profession of statistics. Glantz and Parmley’s accusation that he is involved in what they describe as tobacco industry tactics is irresponsible and unwarranted. Ordinarily, journals are careful not to publish anything that reflects on the character or integrity of an individual. Perhaps you owe
Mantel a public apology for letting this happen in your journal.
Glantz and Parmley [l], in the acknowledgements in their paper, thanked me for carefully checking their power calculations. Mantel [2], commenting on the paper, stated that I “could not have approved of the statistical power reported”, noting specifically that the cited powers for the studies by Helsing et al. [3] in females and by He [4] appeared to be grossly in error. In reply Glantz and Parmley [5] found Mantel’s assertion “puzzling” since Glantz and I [6] had jointly published a letter in Environment International “confirming this agreement”. I am writing to clarify the true situation. Arising out of correspondence in Environment International [7,8] in which conflicting power figures for my own study on passive smoking and heart disease [9] were published, Glantz and I had a long private correspondence about the issue of power. From this it emerged that Glantz had made a gross error in implementing a computer program and that all his original power calculations had been totally incorrect, including those planned to be used in the Glantz and Parmley paper. Glantz admitted his error [lo], re-ran all his calculations, and produced revised power estimates which I agreed were now more sensible, though in fact still differing somewhat from my own estimates which were derived by a procedure that was statistically more defensible. Nevertheless, as Mantel [2] correctly points out, errors, obvious at a glance to any trained statistician, still exist in the paper as it appeared in Circul@ion. The stated powers and confidence limits for the Helsing et al. [3] and He [4]
studies clearly violently conflict. For the Helsing study the conflict has arisen because Glantz’s revised power estimate of 92% was incorrectly printed as 2%. Glantz and Parmley have not only failed to proof-read adequately, but have not even bothered to check their data when told there were errors. The power for the He study [4] may in fact be approximately correct. Here the error lies in the published 95% confidence limits which themselves conflict with the sample size and other published data from the study. I pointed out this problem with He’s paper (described in fuller detail elsewhere [l 11) to Glantz in our correspondence, but he totally ignored it, and included the erroneous material anyway. It concerns me that I should be improperly cited as agreeing to results which are grossly in error. It concerns me even more that an author of a published paper on a subject of importance should have such an apparent cavalier disregard for the accuracy of the material he presents.
MARTHA PJXSKE Darien, CT U.S.A.
REFERENCES 1. Glantz SA, Parmley WW. Passive smoking and heart
disease-epidemiology, physiology and biochemistry. Circulation 1991; 83: l-12. 2. Haenszel W, Shimkin MB, Mantel N. A retrospective study of lung cancer in women. J Nat1Cancerhst 1958; 21: 825-842. 3. Mantel N. The passive smoking myth. Presentend Futureof IodoorAir Quality, Proc BrusselsConference,
14-16 February 1989. Bieva CJ, Courtois Y, Govaerts M, Eds. Amsterdam: Excerpta Medica; 1989: 155-159.
P. N. LEE Independent Consultant in Statistics Sutton, Surrey SM25DA England
REFERENCES
I. Glantz SA, Parmley WW. Passive smoking and heart disease-epidemiology, physiology and biochemistry. Circdation 1991; 83: l-12. 2. Mantel N. Dubious evidence of heart and cancer deaths due to passive smoking. J Ciin Epiden~iol1992; 45: 809-813.