|Home | About | Journals | Submit | Contact Us | Français|
While the assiduity of Castleberry et al in compiling and analysing this huge data assemblage is commendable, regrettably, their conclusion that population CT screening is more cost-effective than symptomatic tumour identification at improving lung cancer (LC) outcomes is based on three demonstrably flawed premises:
Because of its import and its critical contribution to the controversy surrounding LC screening, the implications of overdiagnosis deserve elaboration. Although some authors have insisted on its non-existence, advancing in support the well-known lethality of clinically identified LC, it is important to acknowledge that screening identifies a phenotypically less aggressive LC population. A belief in its invariable lethality entails the untenable corollary that, however obtained, a diagnosis of LC confers immunity to death from all other causes. The issue therefore is quantity. In considering the much-disputed point about its frequency, the following should be taken into account. (1) The majority of screen-identified cases are slow-growing stage I adenocarcinomas, whose natural history permits lengthy exposure to competing lethal morbidities, which are particularly common among older smokers. (2) Although volunteers were selected for participation in trials on the basis of their high risk for LC combined with their excellent health and ability to undergo resectional thoracic surgery, competing lethal morbidities were a far more frequent cause of death than LC. For example, in the Mayo Lung Project, non-LC deaths (most of them attributed to coronary artery disease) were sevenfold the deaths due to LC. (3) Individuals disputing the existence of a substantial number of overdiagnosed persons point out the high death rate of persons with stage I LC who decline intervention. This assumption incorrectly imputes LC as the cause of death among many persons whose decision, without doubt, reflects their or their physician's recognition of manifest lethal comorbidities. It is a tautological fallacy to ascribe their deaths to previously diagnosed LC and conclude that stage I LC is therefore invariably lethal.
Overdiagnosis has two insidious effects. First, it favourably biases outcome estimates. As overdiagnosed persons, by definition, die of another cause, their LC survival will be 100% with or without therapy. Thus, their contribution to outcome improvement as reflected in LC survival is entirely spurious. Second, overdiagnosed persons experience the psychological harm and the risks and morbidities of invasive diagnostic procedures and resectional surgery with no possible offsetting benefit. Furthermore, owing to the loss of pulmonary reserve, the courses of their smoking-induced cardiopulmonary comorbidities are foreshortened. Brown et al (1993), using SEER database figures, reported that the non-cancer relative hazard of death in persons with LC was nearly threefold that in persons with colon or breast cancer.
Additional considerations: The cost estimates of population screening are immense. Per 5-year survival, the authors estimate a cost of 100- to 300-thousand dollars. Even if this enhanced survival translated into a reduction in mortality, its justification, considering other health-related obligations and alternative means of reducing LC mortality, would be open to question. More than 90% of the positive tests in CT trials are false positive, that is, the positive predictive value of a positive test is <10%. The emotional and surgical import of false-positive tests merit emphasis: Wilson et al (2008), in a CT screening study of 3642 persons, reported that 41% had non-calcified nodules, 95% of which were non-cancerous. Fifty-four subjects underwent thoracic surgery for LC; half as many (28) underwent thoracic surgery for benign disorders to exclude LC.
In summary, the current evidence indicates no benefit and a high likelihood of harm from mass CT LC screening of the at-risk population.