Am Fam Physician. 2024;109(2):104
Author disclosure: Dr. Raoof reports serving as a speaker, consultant, or advisory board member for Verily Life Sciences, Exact Sciences, GRAIL, GLG, PRIMED, and AstraZeneca. Dr. Klein reports current employment with GRAIL.
To the Editor:
We would like to comment on several statements in the editorial on multicancer early detection tests by Drs. Doubeni and Castle.1
The editorial states that performance characteristics of these tests are unclear for people in primary care. Two studies assessed multicancer early detection tests in average-risk populations: the PATHFINDER study in 6,662 asymptomatic adults and the DETECT-A study in 10,000 women.2,3 PATHFINDER reported a specificity of 99.5% and a positive predictive value of 43%.2 The DETECT-A study, based on two blood tests, reported a specificity of 98.9% and a positive predictive value of 19.4%.3
Furthermore, the editorial states that multicancer early detection tests have not been studied to distinguish indolent from lethal lesions. A study by Chen, et. al, showed that one of two prospectively studied multicancer early detection tests preferentially detects aggressive cancers.4
The editorial continues by stating that individual cancers are rare and that most people undergoing screening will not benefit and are at risk of potential harms. This highlights the statistical value of multicancer early detection tests: the utilization of aggregate prevalence. Tests for multiple cancers have a lower number needed to screen and a higher positive predictive value than a screening test for a single cancer type.5
The editorial states that screening and detection during the occult preclinical phase can create an illusion of longer survival. Lead-time bias is addressed by randomized controlled trials such as the ongoing NHS-Galleri trial, which includes more than 140,000 patients.6
In Reply:
We thank Drs. Raoof and Klein for their comments. The PATHFINDER and DETECT-A studies were feasibility investigations and not designed for accurately assessing sensitivity and specificity.1,2 Sensitivity and specificity determinations require establishing the presence or absence of cancer with a gold standard test administered to all participants at the same time as the multicancer early detection test, which neither study did.1 In both studies, differential verification, selection, or spectrum biases would overestimate test accuracy and limit generalizability.3,4 The PATHFINDER study included people with hereditary cancer predisposition and history of cancer; therefore, positive predictive values would be lower in a general screening population than inferred from that study. The sample was also fairly homogeneous (92% White participants). These studies are useful for establishing feasibility but unsuitable for establishing accuracy or clinical utility in primary care.
Tests with high specificity tend to have high false-negative rates. The sensitivity reported for CANCERSEEK in the DETECT-A study was 27%.2 Current multicancer early detection tests have lower sensitivity for earlier-stage cancers, which are the main targets of screening. Differentiating cancer precursors or early cancers that will be fatal in a person's lifetime from indolent lesions that are nonfatal is the goal of clinical prevention.5 Those questions cannot be answered by a study such as Chen, et al., that used previously diagnosed cancer cases.
Screening in clinical prevention is a series of care processes that extend beyond performing the test.4 Although combining outcomes across all detectable cancers could yield higher statistical power for a given sample, the effectiveness of screening is not based solely on the ability to detect cancer, but on the ability to prevent premature death from cancer.6 For some cancers (i.e., ovarian cancer and pancreatic cancer), studies have shown that screening results in net harm or does not demonstrate a mortality benefit. Multicancer early detection tests should be studied to demonstrate the ability to decrease the risk of dying from an individual cancer or multiple cancers.1
We agree that randomized controlled trials are the ideal to demonstrate whether multicancer early detection tests provide more benefits than harms, and we look forward to the results of the NHS-Galleri study and others. Such studies should be designed and powered to show the effect on mortality and the harms of testing, and need many years of follow-up. We need such evidence to support using multicancer early detection tests in routine screening of asymptomatic people.1
