Subclinical cancer diagnosis fallacy

Major reasons for overtesting among physicians are considered to include the incorrect belief that ordering many tests would identify subclinical disease and would always improve prognosis (subclinical diagnosis fallacy), in addition to defensive medicine, lack of access to previous data, and requests from patients.1 However, our survey as well as one in the USA showed physicians did not perceive this fallacy as a cause of overtesting.

In fact, ordering many tests would tend to detect subclinical disease but would not always improve prognosis. Regarding cancer screening, the detection of subclinical diseases is associated with greater risk of overtreatment because of the increased bias of lead time and overdiagnosis. But, many physicians seem not to understand the knowledge of subclinical cancer diagnosis fallacy and such physicians might not recognize this as a potential reason for overtesting.

A nationwide survey was conducted to evaluate the knowledge about cancer screening tests and subclinical cancer diagnosis fallacy in cancer screening, based on a questionnaire developed in the previous study by Gigerenzer et al.2 Our hypothesis was that there would be many physicians with lack of this knowledge. Thus, physicians would tend to perceive that subclinical cancer diagnosis fallacy is not recognized as important reason for overtesting.

Participants in our survey included physicians of primary care, general medicine, or specialty throughout Japan. They were voluntarily invited through members of an online educational community (M‐Three) in January 2018. The sponsor for this community had no role in design and data analysis of the survey. The participant was involved with obtaining a point gift for the online community.

In a total of 940 participants, there were 864 (92%) men and 851 (91%) physicians with at least 10 years of clinical experience. In all participants, 79% were physicians of primary care or general medicine, while 669 participants (71%) worked in hospitals, in which 77 (8.2%) were those working in university hospitals.

First, many physicians in Japan showed limited knowledge of what evidence might prove that a cancer screening test saves lives. Among Japanese physicians, 40% incorrectly responded that finding more cancer cases in screened as opposed to unscreened groups provided the proof (Table 1). But the higher incorrect response (48%) was identified in physicians in the US study.2

Table 1.

Understanding among Japanese physicians about evidence that cancer screening saves life (N = 940)

	Item 1		Item 2		Item 3
Proves	376	40.0%	451	48.0%	466	49.6%
Does not prove	374	39.8%	286	30.4%	234	24.9%
Does not know	190	20.2%	203	21.6%	240	25.5%

Item 1: More cancers are detected in screened populations than in unscreened populations (correct answer: does not prove).

Item 2: Screen‐detected cancers have better 5‐year survival rates than cancers detected because of symptoms (correct answer: does not prove).

Item 3: Mortality rates are lower among screened persons than unscreened persons in a randomized trial (correct answer: proves).

About half of physicians in Japan incorrectly responded that increased survival data prove that screening saves lives (48%), although the US physicians did so in much higher incorrect rate (76%). However, there were only a half of Japanese physicians who correctly responded mortality data provide this proof, compared to those (81%) in the USA. Thus, many physicians could not distinguish between incorrect evidence for screening (increased survival) and correct evidence (lower cancer mortality) in both Japan and the USA.

Next, physicians in Japan were also more likely to report they would definitely recommend the screening test that improved 5‐year survival compared with the one that lowered cancer mortality (21% vs 3%). Both proportions among Japanese physicians were much lower than those among physicians in the USA (69% vs 23%) and this trend might reflect there would be lower involvement with cancer screening among Japanese physicians compared with American.

Third, regarding the test that improved 5‐year survival, physicians received additional information that screening test increased the proportion of cancer cases detected at stage I (from 36% without screening to 54% with screening) and they were asked how to change their behavior about the recommendation to patients. Although this information provides no support for the screening test as early‐stage cancer can be detected more by a harmful screening, 47% of Japanese physicians (68% of American physicians) responded that this information made them more or much more likely to recommend the screening.

For the test that reduced mortality, physicians received additional information that the screening test increased cancer incidence (from 27 to 46 per 1000 persons over 5 years) and they were also asked how to change their behavior about the recommendation to patients. Among Japanese physicians, 28% responded the increased incidence made them “more” or “much more” likely to recommend the test, while 62% of American physicians responded so.

Based on our results combined with the previous study by Gigerenzer et al, many physicians interpreted incorrectly higher survival and increased detection with screening tests as proof that the test saves lives. Some physicians incorrectly believed subclinical diagnosis fallacy that greater detection of early‐stage cancer or increased incidence in screened groups constitutes evidence of the benefit of screening. Physicians are likely to perceive that subclinical cancer diagnosis fallacy is not recognized as important reason for overtesting.

Japanese physicians seemed to be less likely to make a strong recommendation for screening tests even they believed it would be life‐saving compared with American physicians. Educational efforts, such as choosing wisely campaign, are needed to improve understanding of screening evidence and subclinical cancer diagnosis fallacy among physicians not only in the USA but also in Japan.3

CONFLICT OF INTEREST

The authors have stated explicitly that there are no conflicts of interest in connection with this article.