Lead Time Bias in Medicine and Psychiatry: A Concept Simply Explained

Abstract

Screening for a disease typically results in earlier diagnosis and hence longer life lived with the disease. Earlier diagnosis permits earlier intervention, which is good if it improves functioning and quality of life from an earlier time point, and if it improves the course and outcome of the disease. However, for diseases that are benign, those that are relentlessly progressive, and those that arise late in life, earlier diagnosis and earlier intervention may not meaningfully change course and outcome. In such situations, the early detection and longer life lived with the disease can incorrectly suggest that the screening and early intervention prolonged survival. Expressed otherwise, longer survival may merely be an artifact of starting measurement earlier. This artifact is known as lead time bias. Lead time bias has long been recognized for many medical conditions, including cancers. In psychiatry, lead time bias is clearly associated with survival in dementia, and has recently been investigated in the context of duration of untreated psychosis and outcome domains in schizophrenia. It is important to know about lead time bias because it may play an unrecognized role in shaping health policy and clinical practice.

Bias shapes attitudes and behaviors in daily life. Bias is prevalent in research as well, with dozens of biases described, such as selection bias, information bias, bias related to confounding, and others, including subsets thereof. ¹ Biases in research can result in inaccurate conclusions with adverse consequences for health policy and patient care. Biases are especially common in observational research.

Certain forms of bias, such as confounding, ² were considered in earlier articles in this column. This article explains lead time bias, a form of bias that has long been known in other branches of medicine but is only recently being discussed in psychiatry.

Lead Time Bias

Consider a disease that begins silently at age A. Screening for disease results in early detection, at age B. Without screening, the disease is detected still later, when symptoms cause clinical distress at age C. In screened patients, the interval between B and C allows for early intervention. In many diseases, such early intervention results in a healthier life and improved disease outcomes. However, in some diseases, such as (a) those that are benign, (b) those that progress at the same rate, regardless of early intervention, and (c) those that arise later in life, a medical endpoint or critical outcome (e.g., disease progression, death) occurs due to disease or other causes at approximately the same time point, age D, where “same time point” is reckoned with respect to the time of disease onset at age A.

Expressed otherwise, regardless of screening, early diagnosis, and early intervention, in situations a, b, and c above, the time interval between A and D is approximately the same for all patients.

In situations a, b, and c above, we do not know what age A was; we only know age B or C, and age D. So, in cohort studies, in situations a, b, and c, we discover that screening and early intervention are associated with longer survival to the critical outcome (interval B to D) than conventional care (interval C to D). We wrongly conclude that screening and early intervention improve survival, and change health policy and treatment approaches accordingly. In actuality, screening and early intervention merely shifted diagnosis and intervention, and hence the start of measurement, from C to B without changing the A to D interval. The B to C interval represents lead time bias.

Hypothetical Scenarios

For easier understanding, here are two hypothetical scenarios. With aggressive screening for a disease, detection and intervention occur at age 50 years and patients live for 25 more years, dying at age 75 due to disease or other causes. Without aggressive screening for disease, detection and intervention occur much later, at age 65 years, and patients live for 10 more years, dying at age 75 due to disease or other causes. Thus, life expectancy for this disease is 25 years in patients who are diagnosed early but only 10 years in patients who are diagnosed late, implying that screening improves life expectancy by 15 years. However, in both scenarios, patients die at the same age. The advantage appearing with screening is an artifact of lead time bias.

Lead Time Bias in Medicine and Psychiatry

Lead time bias has been identified in many disorders in medicine and psychiatry. As an example, the false advantage associated with screening can be 10 years or more in prostate cancer, ³ for which condition lead time bias has been reported for progression to high-grade cancer ⁴ as well as for mortality.^5,6 Lead time bias has been reported for other cancers, as well, such as bladder cancer ⁷ and breast cancer. ⁸

In psychiatry, lead time bias was suggested to explain the association between the duration of untreated psychosis and psychosocial decline in schizophrenia, ⁹ though this finding was not confirmed in another study ¹⁰ ; it is possible that the presence and extent of lead time bias depends on the schizophrenia outcome domain under study. ¹¹ As is easy to understand, lead time bias also explains longer life expectancy associated with earlier diagnosis of dementia. ¹²

Important Messages

Lead time bias, on average, is a mathematical certainty because earlier diagnosis by definition will result in a longer period of life lived with diagnosis. Lead time bias is not an argument against screening, early diagnosis, and early intervention when these result in reduction in the period of symptomatic illness; so, shortening of the duration of untreated psychosis is certainly a desirable goal. Lead time becomes problematic only when it is associated with an extended duration of asymptomatic disease and with the psychological stresses associated with the knowledge thereof, and when the bias results in utilization of healthcare resources without improvement in clinical outcomes.

Because lead time bias will contribute to different extents in different patients at different stages of different diseases, the only way to recognize it is through awareness of its possible existence and examination of population data on the course and outcome of the illness under consideration.

Parting Note

Lead time bias is sometimes referred to as lead-in time bias.