Overdiagnosis: causes and consequences in primary health care

Patient case scenarios

Definition

Overdiagnosis refers to the diagnosis of a condition that otherwise would not have caused symptoms or death.⁶ In other words, it is the detection of a condition without any possible benefit of early treatment to the person with the condition (Figure 1).⁷

• Download figure
• Open in new tab

Figure 1.

Example of overdiagnosis

Reproduced from Hersch et al with permission.⁷

What is not overdiagnosis?

Misdiagnosis and false positives are related but distinctly different concepts. Misdiagnosis is an incorrect diagnosis of a condition, often owing to lack of diagnostic specificity—for example, cellulitis when a patient has swelling and redness of the great toe owing to gout. False positives occur when an initial test result is positive for an abnormality, such as microcalcification seen on mammography, but a follow-up test (eg, biopsy) result reveals no disease. Misdiagnosis and false positives can also lead to harms due to subsequent investigations and unnecessary treatment.

Consequences

Overdiagnosis can lead to serious harms—psychological and behavioural effects of labeling; consequences of subsequent testing (including invasive tests), treatment, and follow-up; and financial effects on the individual who is overdiagnosed and on society. Overtreatment following overdiagnosis can lead to clinically important consequences, including death from the side effects of treatment—for example, sepsis in a patient undergoing chemotherapy for treatment of an overdiagnosed cancer. Higher rates of myocardial infarction and suicide have been reported in men with prostate cancer in the year after diagnosis.^8,9

Overdiagnosis is self-perpetuating. Those who might have been overdiagnosed, such as Linda, encourage others to undergo testing without considering the potential harms of testing and subsequent workup or treatment.

Prerequisites

Overdiagnosis is much more likely to occur in conditions with heterogeneous progression, including those with very slow or no progression to symptoms or death (Figure 2).³ Also essential for overdiagnosis is a test that detects the condition in the early asymptomatic state. To be effective, screening requires a long asymptomatic phase and a test to detect the condition during that time. As not all individuals with an asymptomatic condition will ever become symptomatic, overdiagnosis is inherent to any form of effective screening. The proportion of overdiagnosed cases differs for each disease (Table 1).^10–18

• Download figure
• Open in new tab

Figure 2.

Variable progression of disease

Reproduced from Welch and Black with permission.³

Factors leading to increasing overdiagnosis

Increased rates of testing (increased frequency or wider use of the test in low-risk groups), implementation of higher-sensitivity tests, physician or patient desire to not miss a diagnosis, widened disease definitions (eg, lower threshold for diabetes diagnosis), and financial incentives (eg, payment for increased testing or screening referral) can all lead to increased overdiagnosis.¹⁹

Overdiagnosis can arise when criteria for a disease are modified to include more individuals. This is often done without confirmation of benefit for a substantial portion of those labeled with the disease based on the new definitions.²⁰

Increasing use and overuse of tests can lead to detection of unrelated incidentalomas, a version of overdiagnosis. For example, increasing use of imaging tests (eg, computed tomography [CT] scans, including CT colonography) leads to detection of unrelated asymptomatic conditions such as renal carcinoma or a small aortic aneurysm.¹¹ Consequently, it was demonstrated recently that nephrectomies were linked to the number of scans done, not to the actual incidence of renal cancer. Overtreatment through surgery is now recognized to be one of the risks of excessive CT imaging.

Overdiagnosis can also occur when there is concurrent disease that leads to death and the overdiagnosed disease would not have had time to progress to being symptomatic, as is likely for both the cases described above.

Not limited to cancer screening and detection

Although first described in the setting of cancer diagnosis, overdiagnosis can occur with many conditions,²¹ especially those that have a long indolent phase. Nevertheless, more studies on overdiagnosis have focused on cancer diagnosis than other conditions.

Determining a diagnosis that has clinical consequences and mitigating overdiagnosis

Unfortunately, science has not advanced enough to be able to distinguish whether a condition in a particular individual will lead to clinical consequences. However, attempts are being made to distinguish very slowly progressive or nonprogressive versions of different conditions, as they are more likely to represent overdiagnosis and could be followed up without any treatment.⁵ Examples include the recent recognition and increase in surveillance without treatment of low-risk prostate cancers and ongoing studies to determine when breast DCIS (ductal carcinoma–in situ) can be followed without treatment.²² There are proposals to remove the word cancer from the naming of low-grade and premalignant neoplasms and instead reclassify them as indolent lesions of epithelial origin.²³

Overdiagnosis is real and common

It is expected that the intervention arms of randomized controlled trials (RCTs) of screening would initially demonstrate higher detection and incidence rates. If this were all due to early detection of diseases, it should lead to equivalent decreased incidence later on. In addition, there should be a concomitant decreased incidence of advanced disease. Also, if the screening intervention detects precancer conditions and prevents development of cancers or disease, then there would be an overall reduction in cumulative incidence of the disease.²⁴ Thus, if there were no overdiagnosis, after a long enough follow-up, the overall incidence of cases would be the same in both groups or decreased in the screening group. The Canadian National Breast Screening Study demonstrates that the cumulative increased incidence in the screening and intervention arm usually persists without a later decrease in incidence even when long-term follow-up is available; this is an example of overdiagnosis in the screened group.²⁵ Detecting pre-disease or early disease and changing the outcome is the goal of screening. Unfortunately, there is always a balance between cases in which it might be possible to change the outcome and overdiagnosed cases where we are causing harm.

The pattern in clinical practice most frequently seen and indicative of overdiagnosis is when trends over time demonstrate an increasing incidence, especially of early-stage disease with no or minimal progression to serious disease (ie, advanced disease or cause-specific mortality). As an example, thyroid ultrasound became available in 1980. The incidence of thyroid cancer has been increasing in Canada since the early 1990s, with minimal reduction in mortality due to thyroid cancer (Figure 3).¹⁰ This pattern suggests overdiagnosis due to diagnostic testing.

• Download figure
• Open in new tab

Figure 3.

Age-standardized thyroid cancer incidence and mortality rates in Canada from 1970 to 2012, depicting increasing incidence from the 1990s with no concomitant change in mortality

Adapted from Topstad and Dickinson with permission.¹⁰

Estimates of extent

The extent of overdiagnosis is estimated from follow-up within RCTs, observational studies (such as ecological studies and trends of incident diagnosis and concomitant death over time), and modeling studies.²⁶ Each method has limitations. For RCTs, estimates require long-term follow-up (which might not be available) and assessment under ideal experimental settings. Ecological studies are vulnerable to unidentified confounding. There are many assumptions made for input variables in modeling studies. Indeed, all study designs make inherent assumptions, which lead to varying estimates.

In addition, estimates vary because different formulas are used to calculate overdiagnosis. As an example, a review on prostate-specific antigen screening for prostate cancer¹⁷ used 2 different denominators: all screening-detected prostate cancers during the screening phase of the trials, and prostate cancers diagnosed overall, with the numerator for both calculations being the excess number of cancer cases diagnosed as determined during long-term follow-up. The first approach estimated the percentage of overdiagnosed screening-detected cancers to be 50.4% in the ERSPC (European Randomized Study of Screening for Prostate Cancer) trial. The second approach estimated the percentage of all cancer cases that were overdiagnosed to be 33.2% in the ERSPC trial. Other studies have used incident cases in the unscreened group as the denominator; this approach leads to higher overdiagnosis estimates compared with using the higher number of cancers diagnosed in the screened groups as the denominator. The lowest estimates are obtained when the entire screened population is used in the denominator.^27,28

There is ongoing debate on the best method to measure overdiagnosis.^27,28 Currently, it is recommended to quote the range of magnitude of overdiagnosis, the denominator used, and the method of estimation used.

Suggestions for family physicians

Family physicians should discuss the possibility of overdiagnosis in shared decision making. Overdiagnosis can harm patients; therefore, it needs to be discussed with patients before they enter the cascade of screening. A study from Australia reported that more than 90% of women who had mammography and 82% of men who had prostate-specific antigen screening were not informed of the issue of overdiagnosis. When this concept was explained to patients, a large percentage considered this information crucial to decision making.²⁹ When women aged 48 to 50 were informed of the possible extent of overdiagnosis, they were much less likely to undergo mammographic screening.⁷ Explaining the concept of overdiagnosis is difficult because it is inherently counterintuitive. Stating this up front and using the image of a turtle to explain that some cancers progress slowly can be helpful to foster understanding. An example of such an approach can be found in “The Prostate Specific Antigen (PSA) Test” video by Dr Mike Evans.³⁰

A recent review has described public health efforts to counter the drivers of overdiagnosis, which is worthwhile reading.³¹

Returning to our cases