On my critical understandings about 'modern epidemiology'.

Editor's Note: This Essay is the last paper of Olli Miettinen. I discussed it with him until about a week before his death on November 24, 2021. The article had gone through many iterations, as was the case with all his contributions to the European Journal of Epidemiology over the years. As Miettinen writes, it can be viewed as his academic mini-bio. Albert Hofman.

On Progress in Epidemiologic Academia.

Professor Hofman, in his capacity as the Editor-in-Chief of this journal of "epidemiology," invited me to write an essay for it, given that I've been immersed in epidemiologic academia for a good half-century already. He thought that I likely would have something noteworthy to say, based on my personal experience, about the evolution of "epidemiology" in those decades past, and perhaps also in decades yet to come. In this response to Hofman's invitation I naturally focus on my experience with the research-and-teaching (R & T) that are the core business of epidemiologic academia. The big picture of this experience of mine is, alas, one of confusion about the fundamentals of our R & T, all the way to the present. And just as regrettably, I don't foresee any dramatic dissipation of this confusion in the near-term future, except for the possibility that this essay together with the discussion of it in this journal will prompt the called-for reformation of the culture of epidemiologic academia.

Reflections on preventive medicine.

Having thought much about medicine in my career-long effort to understand it and the research for its advancement, I have come to views rather different form the now-prevailing ones in respect to what preventive medicine is about; what epidemiology is in relation to preventive medicine; what distinguishes preventive medicine in preventive healthcare at large; the relation of preventive medicine to public health; the concept of health promotion; and also the core principles of preventive medicine. All of these views I set forth in this article, for the readers' critical reflection.

Lung cancers diagnosed at annual CT screening: volume doubling times.

PURPOSE: To empirically address the distribution of the volume doubling time (VDT) of lung cancers diagnosed in repeat annual rounds of computed tomographic (CT) screening in the International Early Lung Cancer Action Program (I-ELCAP), first and foremost with respect to rates of tumor growth but also in terms of cell types. MATERIALS AND METHODS: All CT screenings in I-ELCAP from 1993 to 2009 were performed according to HIPAA-compliant protocols approved by the institutional review boards of the collaborating institutions. All instances of first diagnosis of primary lung cancer after a negative screening result 7-18 months earlier were identified, with symptom-prompted diagnoses included. Lesion diameter was calculated by using the measured length and width of each cancer at the time when the nodule was first identified for further work-up and at the time of the most recent prior screening, 7-18 months earlier. The length and width were measured a second time for each cancer, and the geometric mean of the two calculated diameters was used to calculate the VDT. The χ(2) statistic was used to compare the VDT distributions. RESULTS: The median VDT for 111 cancers was 98 days (interquartile range, 108). For 56 (50%) cancers it was less than 100 days, and for three (3%) cancers it was more than 400 days. Adenocarcinoma was the most frequent cell type (50%), followed by squamous cell carcinoma (19%), small cell carcinoma (19%), and others (12%). Lung cancers manifesting as subsolid nodules had significantly longer VDTs than those manifesting as solid nodules (P < .0001). CONCLUSION: Lung cancers diagnosed in annual repeat rounds of CT screening, as manifest by the VDT and cell-type distributions, are similar to those diagnosed in the absence of screening.

Survival of patients with stage I lung cancer detected on CT screening.

BACKGROUND: The outcome among patients with clinical stage I cancer that is detected on annual screening using spiral computed tomography (CT) is unknown. METHODS: In a large collaborative study, we screened 31,567 asymptomatic persons at risk for lung cancer using low-dose CT from 1993 through 2005, and from 1994 through 2005, 27,456 repeated screenings were performed 7 to 18 months after the previous screening. We estimated the 10-year lung-cancer-specific survival rate among participants with clinical stage I lung cancer that was detected on CT screening and diagnosed by biopsy, regardless of the type of treatment received, and among those who underwent surgical resection of clinical stage I cancer within 1 month. A pathology panel reviewed the surgical specimens obtained from participants who underwent resection. RESULTS: Screening resulted in a diagnosis of lung cancer in 484 participants. Of these participants, 412 (85%) had clinical stage I lung cancer, and the estimated 10-year survival rate was 88% in this subgroup (95% confidence interval [CI], 84 to 91). Among the 302 participants with clinical stage I cancer who underwent surgical resection within 1 month after diagnosis, the survival rate was 92% (95% CI, 88 to 95). The 8 participants with clinical stage I cancer who did not receive treatment died within 5 years after diagnosis. CONCLUSIONS: Annual spiral CT screening can detect lung cancer that is curable.

Up from 'false positives' in genetic-and other-epidemiology.

Published 'positive' results of epidemiological studies on possible associations (descriptive or causal) are ever more commonly 'false positives' and, thus, false warrants for claiming discovery. More common examination of a multitude of possible associations is widely seen to be the principal cause of this trend. I dispute this explanation and take the principal basis for the trend to be the ever decreasing prior plausibility of the associations that are reported on; and publication bias leading to missing 'negatives' in the published results exacerbates the appearance of the problem. The problem is, however, eminently remediable. We epidemiologists, as a collective of researchers, should leave behind the decision-oriented, inference-denying cult of statistical 'significance' adduced by Neyman and Pearson, and in its stead we should embrace the Fisherian culture of focusing on the production of statistical evidence, for use in inference by our readers. I recommend a simple, objective measure of evidence, suitable for readers' Bayesian-type inferences about the existence of an association.

Computed tomographic screening for lung cancer: the relationship of disease stage to tumor size.

BACKGROUND: The relationship of lung cancer stage to tumor diameter has been identified as a prognostic indicator. We report on the stage-size relationship of these asymptomatic, latent lung cancer cases diagnosed by computed tomographic screening. METHODS: Baseline and repeat screening of 28 689 people following the International Early Lung Cancer Action Program regimen of screening has resulted in 464 diagnoses of lung cancer. Each case was characterized according to tumor diameter, consistency (solid, part solid, or nonsolid), and the presence or absence of identifiable metastases (N0 M0) at the time of diagnosis, regardless of whether it was delayed. RESULTS: For the 436 non-small cell carcinomas, the percentages of cases with no metastases (N0 M0) were 91%, 83%, 68%, and 55% for the categories 15 mm or less, 16 to 25 mm, 26 to 35 mm, and 36 mm or greater, respectively. The gradients in the successive percentages of N0 M0 cases were significantly different (P = .02, 1-sided), except between the last 2 categories, and held for solid nodules, were suggestive for part-solid ones, but were not suggestive for nonsolid ones. For the 28 small cell carcinomas, the percentages of N0 M0 cases were 67% and 23% (P = .01, 1-sided), respectively, for those 25 mm or less compared with those greater than 25 mm. CONCLUSIONS: Lymph node status has a strong relationship to tumor diameter for non-small cell and small cell cancers. The percentages of N0 M0 cases in screen-diagnosed lung cancers are much higher than previously reported in the Surveillance, Epidemiology, and End Results registry. These results provide direct evidence of a stage-size relationship in a screened population.

Computed tomography screening for lung cancer: prospects of surviving competing causes of death.

PURPOSE: The primary objective of this study was to shed light on the frequency of death from a "competing" cause among persons who enter into computed tomography (CT) screening for lung cancer and to determine the 5- and 10-year rates of death from causes other than lung cancer in a cohort of older smokers and former smokers with the initiation of CT screening for lung cancer. PATIENTS AND METHODS: We followed a cohort of 2141 men and women aged 60-75 years with a history of 30-100 pack-years of cigarette smoking who enrolled for CT screening for lung cancer in 1993-2004. The National Death Index retrieval program was used to identify all deaths and causes of death. Follow-up time from the date of the initial CT to death, loss to follow-up, or December 31, 2004, whichever came first, was calculated for each subject. Median duration of follow-up was 50 months (range, 1-133 months). Kaplan-Meier analysis was used to derive the 5- and 10-year survival rates with the exclusion of deaths from lung cancer. RESULTS: The 5- and 10-year survival rates, conditional on not dying from lung cancer, were 96% and 90.7%, respectively. The corresponding 95% confidence intervals were 95%-97% and 88.2%-95.2%, respectively. CONCLUSION: Older smokers and former smokers seeking and receiving CT screening for lung cancer have a low 10-year risk of dying from causes other than lung cancer, and early treatment of screen-diagnosed cancer can be life-saving.

Frequency of coronary artery calcification on low-dose computed tomography screening for lung cancer.

PURPOSE: The purpose of this study was to determine the frequency of coronary artery calcification (CAC) in high-risk people undergoing computed tomography (CT) screening for lung cancer. METHODS: Between 1999 and 2004, we performed CT screening for lung cancer on 4250 participants, all without documented prior cardiovascular disease, using multidetector-row (MD) CT. Of the patients, 1102 underwent imaging with a four-detector-row CT at 120 kVp and 40 mA, with pitch 1.5 and collimation of 2.5 mm in a single breath hold of 15-20 seconds, and 3148 did with an eight-detector-row CT at the same kVp, mA, and pitch settings but with collimation of 1.25 mm. Visualized CACs in each coronary artery (main, left anterior descending, circumflex, and right) were scored separately as 0 (absent), 1 (mild), 2 (moderate), or 3 (severe), yielding a possible score of 0-12 for each person. Frequency distributions by gender, age, and pack-years of smoking were determined. Odds ratios (ORs) were calculated using logistic regression analysis of the prevalence of CAC as a joint function of gender, age, pack-years of smoking, and presence of diabetes. RESULTS: Among the subjects younger than 50 years, positive CAC scores were three times more frequent for men than for women (22% vs. 7%); among those older than 50 years, the frequency increased for both men and women but the increase for women was greater than that for men. The frequency of positive CAC scores increased with increasing pack-years of smoking; it was always higher for men than for women. The ORs were 2.6 for male gender (P<.0001), 3.7 and 9.6 for ages 60-69 years and 70 years or older, respectively, for increasing age (P<.0001 for both), 1.6 and 2.3 for 30-59 pack-years and 60 pack-years or longer, respectively, for increasing pack-years of smoking (P<.0001 for both), and 1.6 for having diabetes (P=.016). CONCLUSION: The CAC score can be derived from ungated low-dose MDCT images. This information can contribute to risk stratification and management of coronary artery disease.

CT screening for lung cancer: significance of diagnoses in its baseline cycle.

PURPOSE: The aim of this study was to assess the significance of Stage I diagnoses of lung cancer in the baseline cycle of screening for this disease, with special reference to the potential for overdiagnosis. METHODS: We reviewed all 69 cases of Stage I lung cancer diagnosis resulting from our baseline CT screening. Among these 69 cases of lung cancer, 24 presented as solid, 30 as part-solid, and 15 as nonsolid nodules. The extent to which these represented genuine malignancy was assessed by a panel of experts on lung pathology, and the "aggressiveness" of these cases was addressed by the criterion of the tumor's volume doubling time being less than 400 days. RESULTS: The expert panel confirmed all 69 cases as representing genuine malignancy. Among the 69 cases without evidence of metastases, the proportion that satisfied the aggressiveness criterion was 60/69=87%. The corresponding proportions by presentation as solid, part-solid, and nonsolid nodule were 23/24 (96%), 27/30 (90%), and 10/15 (67%), respectively. CONCLUSIONS: In baseline CT screening for lung cancer, overdiagnosis of the disease is uncommon, with cases presenting as a nonsolid nodule a possible exception to this.

Women's susceptibility to tobacco carcinogens and survival after diagnosis of lung cancer.

CONTEXT: It has been hypothesized that women are more susceptible to tobacco carcinogens than men, but after diagnosis of lung cancer, they have better survival rates than men. OBJECTIVE: To add to the evidence on the lung cancer risk of women who smoke and their survival after diagnosis of lung cancer, conditional on other prognostic indicators and compared with men of the same age who smoke. DESIGN, SETTING, AND PARTICIPANTS: Nonexperimental, etiologic study with prospective collection of data based on baseline computed tomographic screening for lung cancer and follow-up of diagnosed cases of lung cancer in North America in 1993-2005. A total of 7498 women and 9427 men were screened, all of whom were asymptomatic, aged at least 40 years, and had a history of cigarette smoking. MAIN OUTCOME MEASURES: Comparing women with men, the prevalence odds ratio (OR) for screen-detectable lung cancer (conditional on age and smoking history) and the hazard ratio of fatal outcome of lung cancer (conditional on smoking history, disease stage, tumor cell type, and resection). RESULTS: Lung cancer was diagnosed in 156 women and 113 men (rates of 2.1% and 1.2%, respectively). The prevalence OR comparing women with men was 1.9 (95% confidence interval [CI], 1.5-2.5). The hazard ratio of fatal outcome of lung cancer comparing women with men was 0.48 (95% CI, 0.25-0.89). CONCLUSION: Women appear to have increased susceptibility to tobacco carcinogens but have a lower rate of fatal outcome of lung cancer compared with men.

Feinstein and study design.

Dr. Alvan Feinstein saw himself as the father of "clinical epidemiology" in the modern meaning of this term, of this "new intellectual domain of modern medical science." In this role, he saw himself as drawing from his "clinical sophistication" and from "the rigorous scientific demands" to which "clinicians are accustomed," while "public health" epidemiologists "often use a more arbitrary set of standards." His conception of the scope of clinical epidemiology was remarkably Catholic and the same was the case in respect to cause-effect research in it. In the latter, he was firmly committed to the randomized-trial paradigm, including in his teachings on study design in etiologic research. Characteristically original, many of Dr. Feinstein's study-design ideas remain controversial.

Women's susceptibility to tobacco carcinogens.

STUDY OBJECTIVES: To assess lung cancer risk of smoking women relative to that of equally smoking men. METHODS: The study base was constituted by baseline CT screenings for lung cancer on 1202 women and 1288 men, at least 40 years of age and with at least 10 pack-years of cigarette smoking. The prevalence-odds (incidence-density) ratio contrasting women with men was calculated. Confoundings by age and the particulars of smoking history were controlled in logistic regression analysis. RESULTS: For the prevalence-odds ratio contrasting women with men, upon control of age and smoking history, the point estimate was 2.7 and the 95% interval estimate 1.6-4.7. The diagnosed cancers were of equally 'aggressive' types between the two genders. CONCLUSIONS: At variance with evidence from cohort studies, this evidence from a screening experience calls for further consideration of the hypothesis that women are more susceptible to tobacco carcinogens than are men.

Small stage I cancers of the lung: genuineness and curability.

PURPOSE: To assess the genuineness and curability of diagnosed cases of Stage IA non-small-cell lung cancer of diameter 6-15 mm, with a view to screen-diagnosed asymptomatic cases. METHODS: We identified all cases of diagnosed Stage IA (T1N0M0) primary non-small-cell lung cancer documented in the Surveillance, Epidemiology, and End Results (SEER) registry in 1988-1994. There were 885 such cases of diameter 6-15 mm, 33 of them unresected, 1659 of diameter 16-25 mm, and 782 of diameter 26-30 mm. We determined the 8-year cumulative lung-cancer death rates adjusted for competing causes of death, separately for untreated and fully resected cases. Given the relative underdetection of lymph node metastases of the untreated cases, we also documented the case-fatality rates for the resected with inclusion of those with detected intra-pulmonary, hilar or ipsilateral mediastinal lymph node metastases. RESULTS: The 8-year fatality rate for the diagnosed but untreated cases of lung cancer 6-15 mm in diameter was 87%, for 16-25 mm it was 94%, and for 26-30 mm it was 88%. The corresponding estimates of cure rates with resection were 71, 67, and 55%, respectively. The cure rate estimate for the smallest cancers was 66% when adding those with detected lymph node metastases to the resected series, for those 16-25 mm in diameter it was 65%, and for those 26-30 mm in diameter it was 51%. CONCLUSION: Almost all diagnosed cases of Stage IA lung cancer as small as 10 mm in diameter have a malignant natural course, fatal if not treated, thus representing genuine cancer. Most of these cases are curable by resection.

CT screening for lung cancer Assessing a regimen's diagnostic performance.

PURPOSE: The purpose of this study was to characterize the diagnostic performance of a regimen of CT screening for lung cancer. METHODS: Using a common protocol/regimen of screening, 2968 asymptomatic persons at high risk for lung cancer were enrolled in two studies [Early Lung Cancer Action Projects (ELCAP) I and II] for baseline and annual repeat screening. A total of 4538 annual repeat screenings were performed. The regimen's diagnostic performance was characterized in terms of frequency of positive result of the initial CT as well as of screen-diagnosis and Stage I screen-diagnosis among all diagnoses (interim-diagnoses included), all separately for baseline and annual repeat screenings. RESULTS: The proportions with positive result of the initial CT were 12% and 6% in the baseline and repeat screenings, respectively. The proportions of screen-diagnoses among all diagnoses (interim-diagnoses included) were 97% and 99% in the baseline and repeat cycles, respectively. The corresponding proportions of pre-surgical Stage I screen-diagnoses were 95% and 93%. CONCLUSION: The performance of the ELCAP regimen is quite satisfactory in avoiding over many positive results of the initial CT, and it produces highly promising diagnostic results as for the attainment of cure by early intervention.

Professionalism in medicine.

A Charter on Medical Professionalism (CMA) has just recently been developed internationally, and the Canadian Medical Association is calling for public dialogue on medical professionalism now that reforms in the Canadian system of health care are imminent. We posit that good practices are at issue; we outline the essence of these in general and also specifically in the knowing, teaching and intervening components of practice. We also see challenges not to, but in, medical professionalism - first and foremost in the profession's definition of good practices and the payer's confinement of insurance coverage to these.

Evaluation of screening for a cancer: annotated catechism of the Gold Standard creed.

Even non-scientific people are generally aware of, and commonly also personally affected by, the confusion and debates that arose from two recent systematic reviews and their associated meta-analyses of the published randomized controlled trials (RCTs) that had been designed to assess the usefulness of screening for breast cancer. With nothing about the principles of research on cancer screening learned from this debacle, an RCT on screening for lung cancer has just been launched in the same 'gold standard' framework; and so in about 10 years, and upon expenditure of hundreds of millions of dollars, we are said to be able to learn whether the screening saves lives and whether it thereby is justifiable as a matter of public policy. We here examine the core tenets and precepts in this 'gold standard' line of thinking, and we argue that they are matters of belief at variance with the dictates of reason.

Diagnostic probability function for acute coronary heart disease garnered from experts' tacit knowledge.

OBJECTIVES: Knowing about a diagnostic probability requires general knowledge about the way in which the probability depends on the diagnostic indicators involved in the specification of the case at issue. Diagnostic probability functions (DPFs) are generally unavailable at present. Our objective was to illustrate how diagnostic experts' case-specific tacit knowledge about diagnostic probabilities could be garnered in the form of DPFs. STUDY DESIGN AND SETTING: Focusing on diagnosis of acute coronary heart disease (ACHD), we presented doctors with extensive experience in hospitals' emergency departments a set of hypothetical cases specified in terms of an inclusive set of diagnostic indicators. We translated the medians of these experts' case-specific probabilities into a logistic DPF for ACHD. RESULTS: The principal result was the experts' typical diagnostic probability for ACHD as a joint function of the set of diagnostic indicators. A related result of note was the finding that the experts' probabilities in any given case had a surprising degree of variability. CONCLUSION: Garnering diagnostic experts' case-specific tacit knowledge about diagnostic probabilities in the form of DPFs is feasible to accomplish. Thus, once the methodology of this type of work has been "perfected," practice-guiding diagnostic expert systems can be developed.