Evaluation of cancer screening efficacy by means of non-randomized studies
Noel Weiss
Ecologic Studies
The level of screening for a given form of cancer can vary within a population over time, and/or can vary across populations. But only rarely can population-level studies deal with the other factors that vary by time and place – and also bear on the occurrence of or mortality from that cancer – to give rise to an estimate of screening efficacy that is credible. Happily, rare things sometimes do happen – here are two examples.
Cohort Studies - Cancer Mortality
Cohort studies of the efficacy of cancer screening in reducing mortality have produced little useful information. Those that compare survival between persons whose cancer was or was not screen-detected suffer from lead-time and length bias. And those that compare cancer mortality between persons who do and do not undergo screening typically cannot ascertain a crucial characteristic of screened persons – the absence of symptoms or signs of the cancer – in cohort members who have not undergone screening.
Case-Control Studies - Cancer Mortality
Most tests performed for cancer screening also are employed to evaluate persons with signs or symptoms of the cancer in question. The success of these studies hinges largely on the ability to obtain information on the fact of testing in cases and controls, but on the REASON the test was performed.
Cohort Studies - Cancer Incidence
Some investigators have sought to configure such cohort studies to, as much as possible, emulate a randomized trial. Others investigators have instead given priority to maximizing the likelihood that their cohort study will observe an association between screening and reduced cancer incidence if one truly exists. In some instances – perhaps many! – these goals can be in conflict.
Case-Control Studies - Cancer Incidence
These studies focus on receipt of screening during a period of time prior to the presence of the cancer in question, so (in contrast to case-control studies of cancer mortality) need much less to be pre-occupied with distinguishing screening from diagnostic tests.
Gauging the impact of a cancer re-screening interval
It is the rare randomized trial that that employs more than one re-screening interval. Non-randomized studies can assess the occurrence of cancer after a negative screening test, but the results of such studies must be interpreted cautiously when trying to estimate the consequences of choosing different re-screening intervals.