Avancées dans la détection précoce du cancer grâce aux tests sanguins

Titre: Multi-cancer early detection tests for general population screening: a systematic literature review

Résumé: BackgroundGeneral population cancer screening in the UK is limited to selected cancers. Blood-based multi-cancer early detection (MCED) tests aim to detect potential cancer signals from multiple cancers in the blood. The use of an MCED test for population screening requires a high specificity and a reasonable sensitivity to detect early-stage disease, so that the benefits of earlier diagnosis and treatment can be realised. ObjectiveTo undertake a systematic literature review of the clinical effectiveness evidence on blood-based MCED tests for screening. MethodsComprehensive searches of electronic databases (including MEDLINE and Embase) and trial registers were undertaken in September 2023 to identify published and unpublished studies of MCED tests. Test manufacturer websites and reference lists of included studies and pertinent reviews were checked for additional studies. The target population was individuals aged 50 to 79 years without clinical suspicion of cancer. Outcomes of interest included test accuracy, number and proportion of cancers detected (by site and stage), time to diagnostic resolution, mortality, potential harms, health-related quality of life (HRQoL), acceptability and satisfaction. Risk of bias was assessed using the QUADAS-2 checklist. Results were summarised using narrative synthesis. Stakeholders contributed to protocol development, report drafting, and interpretation of review findings. ResultsOver 8000 records were identified. Thirty-six studies met the inclusion criteria: one ongoing randomised controlled trial (RCT), 13 completed cohort studies, 17 completed case-control studies and five ongoing cohort or case-control studies. Individual tests claimed to detect from three to over 50 different types of cancer. Diagnostic accuracy of currently available MCED tests varied substantially: Galleri(R) (GRAIL) sensitivity 20.8% to 66.3%, specificity 98.4% to 99.5% (3 studies); CancerSEEK (Exact Sciences) sensitivity 27.1% to 62.3%, specificity 98.9% to 99.1% (2 studies); SPOT-MAS (Gene Solutions) sensitivity 72.4% to 100%, specificity 97.0% to 99.9% (2 studies); TruCheck (Datar Cancer Genetics) sensitivity 90.0%, specificity 96.4% (1 study); CDA (AnPac Bio) sensitivity 40.0%, specificity 97.6% (1 study). AICS(R) (Ajinomoto) screens for individual cancers separately, so no overall test performance statistics are available. Where reported, sensitivity was lower for detecting earlier stage cancers (Stage I-II) compared with later stage cancers (Stage III-IV). Studies of seven other MCED tests at an unclear stage of development were also summarised. LimitationsStudy selection was complex; it was often difficult to determine the stage of development of MCED tests. The evidence was limited; there were no completed RCTs and most included studies had a high overall risk of bias, primarily owing to limited follow-up of participants with negative test results. Only one study of Galleri recruited asymptomatic individuals aged over 50 in the USA, however, study results may not be representative of the UK general screening population. No meaningful results were reported relating to patient relevant outcomes, such as mortality, potential harms, HRQoL, acceptability or satisfaction. ConclusionsAll currently available MCED tests reported high specificity (>96%). Sensitivity was highly variable and influenced by study design, population, reference standard test used and length of follow-up. Future workFurther research should report patient relevant outcome and consider patient and service impacts. PROSPERO registration number: CRD42023467901 FundingThis project was funded by the National Institute for Health and Care Research (NIHR) Evidence Synthesis programme (NIHR161758) and will be published in full in the Health Technology Assessment Journal; Vol. XX, No. XX. See the NIHR Journals Library website for further project information. PLAIN LANGUAGE SUMMARYCancer screening is only available for some cancers. New tests that look for signs of cancer in blood (blood-based multi-cancer early detection tests) are being developed; they aim to detect multiple different cancers at an early stage, when they are potentially more treatable. Taking account of stakeholder feedback, we reviewed all studies assessing the effectiveness of blood-based multi-cancer early detection tests for cancer screening. We thoroughly searched for relevant studies and found over 8000 records. We included 30 completed studies and six ongoing studies of 13 different tests. None of the studies were good quality, mainly because they didnt properly check whether the test result might have been incorrect and participants with a negative test result actually had cancer. Most studies included participants who are different from the general UK population that would be likely to have this type test for cancer screening. None of the studies reported meaningful results for patient-relevant outcomes, such as death, potential harms, quality of life and acceptability. We found 14 completed studies assessing six tests that are currently available: Galleri(R) (GRAIL), CancerSEEK (Exact Sciences), SPOT-MAS (Gene Solutions), TruCheck (Datar Cancer Genetics), CDA (AnPac Bio) and AICS(R) (Ajinomoto). All of the tests were quite good at ruling out cancer, but their accuracy for finding cancer varied a lot, mostly because of differences in the study methods and characteristics of the included participants. The tests were better at finding more advanced cancers, which are potentially less curable than early cancers, so more research is needed to know whether tests would actually save lives. Better designed studies including participants similar to those who might get the test in the real world, and which report on patient-relevant outcomes and properly consider patient experience and impact on services, are needed. Several new studies are planned or underway. SCIENTIFIC SUMMARYO_ST_ABSBackgroundC_ST_ABSGeneral population cancer screening in the UK is limited to selected cancers (cervical, breast, bowel and, for some high-risk individuals, lung). Most other cancers are detected after presentation of symptoms, when the disease tends to be at a more advanced stage and treatment options may be more limited. Blood-based multi-cancer early detection (MCED) tests aim to detect potential cancer signals (such as circulating cell-free deoxyribonucleic acid [cfDNA]) from multiple cancers in the blood. The use of an MCED test as a screening tool in a generally healthy, asymptomatic population, requires a high specificity and a reasonable sensitivity to detect early-stage disease, so that the benefits of earlier diagnosis and treatment can be realised. An MCED test embedded within a national population-based screening programme, in addition to existing cancer screening programmes, may increase the number of cancers diagnosed at an earlier stage. However, identification of cancers with no effective treatments, even at an early stage, may have no improvement on mortality or health-related quality of life (HRQoL). In addition, early screening of healthy people for such a wide range of cancers, and the expected lengthy time to diagnostic confirmation, may create anxiety and lead to unnecessary follow-up tests, when false positive test results occur. ObjectivesThe aim of this project was to conduct a systematic review to assess the accuracy and clinical effectiveness, acceptability and feasibility of blood-based MCED tests for population-based screening. MethodsComprehensive searches of electronic databases (including MEDLINE and Embase) and trial registers were undertaken in September 2023. Test manufacturer websites and reference lists of included studies and pertinent reviews were checked for additional relevant studies. Published and unpublished prospective clinical trials and cohort studies of blood-based MCED tests for screening were sought. Studies assessing blood-based tests for assessing prognosis or therapeutic decision-making in patients with cancer were not eligible for inclusion. The target population was individuals aged 50 to 79 years without clinical suspicion of cancer and who had not been diagnosed with, or received treatment for, cancer within the last three years. As insufficient relevant studies were identified within the target population, studies that included patients known to have cancer (i.e., case-control studies) and studies that included individuals with a different age range were included. Outcomes of interest were test accuracy (including sensitivity, specificity, positive and negative predictive values), number and proportion of cancers detected (by site and stage), mortality, time to diagnostic resolution, incidental findings, additional tests and procedures, potential harms, HRQoL, acceptability and satisfaction. A standardised data extraction form was developed and piloted. Data on the intervention(s), participant characteristics, setting, study design, reference standard test(s) used, and relevant outcomes were extracted from included studies by one reviewer and independently checked by a second reviewer. Risk of bias and applicability were assessed using the QUADAS-2 (Quality Assessment of Diagnostic Accuracy Studies) checklist by one reviewer and independently checked by a second. Disagreements were resolved through discussion. Results were summarised using narrative synthesis. Stakeholders contributed to protocol development, report drafting, and interpretation of review findings. ResultsThe electronic searches identified 8,069 records; 228 full texts were further reviewed. Eleven additional records were identified from searching MCED test manufacturer websites. Study selection was complex; it was often difficult to determine whether studies assessed technologies at an early stage of development, or the final or near-final version of the test. Thirty-six studies, evaluating thirteen MCED tests or technologies, met the review inclusion criteria: one ongoing randomised controlled trial (RCT), 13 completed cohort studies, 17 completed case-control studies, four ongoing cohort studies and one ongoing case-control study. Studies assessed the following MCED tests: Galleri(R) test (GRAIL, Menlo Park, California), CancerSEEK (Exact Sciences, Madison, Wisconsin), SPOT-MAS (Gene Solutions, Ho Chi Minh City, Vietnam), TruCheck (Datar Cancer Genetics, Beyreuth, Germany), CDA (Cancer Differentiation Analysis; AnPac Bio, Shanghai, China) and AICS(R) test (AminoIndex Cancer Screening; Ajinomoto, Tokyo, Japan). Other MCED technologies included in the review, that were at an unclear stage of development and did not appear to be available for use, were: Aristotle(R) (StageZero Life Sciences, Richmond, Ontario), CancerenD24 (manufacturer unknown), OncoSeek(R) (SeekIn Inc, San Diego, California), SeekInCare(R) (SeekIn Inc, San Diego, California), OverC (Burning Rock Biotech, Guangzhou, China), Carcimun-test (Carcimun Biotech, Garmisch-Partenkirchen, Germany) and SpecGastro test (manufacturer unknown). Technologies that appeared to be at a very early stage of development did not meet the inclusion criteria for the review. Individual MCED tests and technologies claimed to detect from three to over 50 different types of cancer. Owing to the differences in the number of cancer types detected, study design and populations, statistical pooling of results was not considered appropriate. Studies of MCED tests available for useThe risk of bias assessment identified substantial concerns with the included studies. Case-control studies have a high risk of bias in the QUADAS-2 patient selection domain. Almost all of the studies had a high risk of bias in the flow and timing domain, however, this is difficult to avoid when the reference standard for positive test results involves invasive testing, as it is not practical or ethical to undertake such invasive tests in participants with a negative MCED (index) test result. Only one study was undertaken in the UK, although this was in individuals in whom cancer was suspected, so not reflective of the general cancer screening population. Cancer risk and the availability of general population cancer screening programmes differ worldwide, which will impact the applicability of results of the included studies to the UK. Ethnicity and socioeconomic status of included participants was not well reported in the included studies. There were also concerns about the applicability of the CancerSEEK test, which has since been modified (now called Cancerguard) and is undergoing further assessment. The applicability of the SPOT-MAS, Trucheck, CDA and AICS tests assessed in the included studies was unclear. Outcomes relating to MCED test performance (i.e., test accuracy and number of cancers detected by site and/or stage) were reported in most studies. Overall test sensitivity and specificity are not directly comparable across different MCED tests, owing to the differences in the number of cancer types each test claims to detect. Diagnostic accuracy varied substantially (95% confidence interval [CI] shown in brackets): Galleri (3 studies) Sensitivity: 20.8% (14.0 to 29.2) to 66.3% (61.2 to 71.1) Specificity: 98.4% (98.1 to 98.8) to 99.5% (99.0 to 99.8) CancerSEEK (2 studies) Sensitivity: 27.1% (18.5 to 37.1) to 62.3% (59.3 to 65.3) Specificity: 98.9% (98.7 to 99.1) to 99.1% (98.5 to 99.8) SPOT-MAS (2 studies) Sensitivity: 72.4% (66.3 to 78.0) to 100% (54.1 to 100) Specificity: 97.0% (95.1 to 98.4) to 99.9% (99.6 to 100) TruCheck (1 study) Sensitivity: 90.0% (55.5 to 99.7) Specificity: 96.4% (95.9 to 96.8) CDA (1 study) Sensitivity: 40.0% (95% CI 12.2 to 73.8) Specificity: 97.6% (95% CI 96.8 to 98.2) AICS screens for individual cancers separately; sensitivity ranged from 16.7% (95% CI 3.0 to 56.4) for ovary/uterus cancer to 51.7% (95% CI 34.4 to 68.6) for gastric cancer. Sensitivity by cancer stage was only reported in some studies of Galleri and CancerSEEK. Sensitivity was considerably lower for detecting earlier stage cancers (Stage I-II) compared with later stage cancers (Stage III-IV). Amongst the Galleri studies, sensitivity for detecting Stage I-II cancer ranged from 27.5% (25.3 to 29.8) to 37.3% (29.8 to 45.4) and sensitivity for detecting Stage III-IV cancer ranged from 83.9% (81.7 to 85.9) to 89.7% (84.5 to 93.6). The CancerSEEK cohort study reported sensitivity for detecting Stage I-II cancer of 12.7% (95% CI 6.6 to 23.1) and sensitivity for detecting Stage III-IV cancer of 53.1% (95% CI 36.4 to 69.1). One study of Galleri found that sensitivity was higher in an elevated risk cohort (23.4%, 95% CI 14.5 to 34.4) than a non-elevated risk cohort (16.3%, 95% CI 6.8 to 30.7). Studies of Galleri, CancerSEEK, SPOT-MAS, CDA and AICS reported sensitivity by cancer site and found that it varied substantially, although the total number of participants diagnosed with certain types of cancer was low, so results are difficult to interpret. Screening programme availability: The sensitivity of the MCED tests to detect solid tumour cancers without a current screening programme in the UK was generally higher than the sensitivity to detect cancers with a current screening programme in the UK (breast, cervical and colorectal). However, this was not the case in one study of Galleri and the study of the CDA test, where sensitivity for detecting solid tumour cancers without a current screening programme was lower than for cancers with a current screening programme in the UK. Subgroup results by participant demographic characteristics: One study each of Galleri and CancerSEEK reported MCED test performance by pre-specified subgroups of interest (i.e., age, sex and ethnicity). For CancerSEEK, sensitivity was slightly lower for participants less than 50 years of age, compared to participants aged 50 or over, while for Galleri sensitivity was very similar across the age categories presented. The sensitivity of Galleri was highest for Hispanic participants (63%), and it was lowest (43%) for the small number of participants classified as Other ethnicity in the study. Sensitivity of CancerSEEK ranged from 50% in participants with unknown ethnicities to 70.4% in Asian participants (and cancer was correctly detected by the CancerSEEK test in one Hispanic participant resulting in a sensitivity of 100%). One study using an earlier version of the Galleri test reported results by age and sex for a subset of study participants; cancer signal detection rate was similar in males and females and increased with age for both sexes, however, few details were given on the subset of participants analysed. Only one study of Galleri reported data for participants with a low socioeconomic status. Patient relevant outcomes Only limited results relating to patient relevant outcomes, such as mortality, potential harms, HRQoL, acceptability and satisfaction of individuals screened, were reported in some studies of Galleri, CancerSEEK and AICS. Studies of MCED technologies at an unclear stage of developmentThe risk of bias assessment identified substantial concerns. Most studies were case-control studies so had a high risk of bias in the patient selection domain of QUADAS-2. Most studies also had a high risk of bias in the index test and/or flow and timing domains. All studies were considered to have high or unclear concerns relating to the applicability of study participants, index tests and reference standard tests. Outcomes relating to MCED test performance were reported in most studies. OncoSeek reported the lowest overall sensitivity across all cancer types (47.4%), and CancerenD24 reported the lowest sensitivity in detecting bladder cancer (38.0%). By stage, OverC and SeekInCare reported a sensitivity of 35.4% and 50.3%, respectively, for stage I cancer. The highest sensitivity overall came from the Carcimun-test (88.8%), however, the exclusion of individuals with inflammation is noted as a disadvantage. The SpecGastro test was only developed to detect three types of gastrointestinal cancer (colorectal, gastric, and oesophageal). Stakeholder engagementAt the protocol stage, stakeholders highlighted issues with the implementation of MCED tests, including resource use, impact on existing diagnostic services and wider care pathways, the need to balance benefits with potential risks, and consideration of factors likely to affect test uptake. Stakeholders also reinforced the importance of patient relevant outcomes. Stakeholders commenting on the draft report noted that important details about the potential benefits, harms, and possible unintended consequences of implementing MCED tests in the UK were poorly reported, limiting the relevance of the available evidence for policy decision-making. Other feedback fell into six broad areas: poor applicability and generalisability of available evidence; limitations of the current evidence base; the potential impact of MCED tests on existing screening, diagnostic and treatment pathways; opportunities to enhance services to improve outcomes; acceptability and potential impact on populations offered and/or receiving screening, and; targeting specific groups to support early identification and improve outcomes. ConclusionsLimited evidence is available on the potential for early detection of treatable cancers, and the consequences of introducing screening with an MCED test in a UK population. There were no completed RCTs identified for any of the MCED tests and most included studies had a high overall risk of bias, primarily owing to limited follow-up of participants with negative test results. There were concerns about the applicability of the participants in most studies. Only one study of Galleri recruited asymptomatic individuals aged over 50 years but it was conducted in the USA, therefore, study participants and results may not be representative of a UK screening population. All currently available MCED tests (Galleri, CancerSEEK, SPOT-MAS, TruCheck, CDA and AICS) reported high specificity (>96%) which is essential if an MCED test is to correctly classify people who do not have cancer. Sensitivity was variable and influenced by study design, population, reference standard test used and length of follow-up. Sensitivity also varied by cancer stage; where reported, MCED tests had considerably lower sensitivity to detect earlier stage cancers (Stage I-II). Sensitivity also appeared to vary substantially for different cancer sites, although results are limited by small patient numbers for some cancer sites. The sensitivity of most of the MCED tests to detect solid tumour cancers without a current screening programme in the UK was higher than their sensitivity to detect cancers with a screening programme in the UK (breast, cervical and colorectal). Where reported, differences in test accuracy by age and sex were small. Whilst some differences were observed by ethnicity, these results should be interpreted with caution as the majority of participants recruited to studies were White and numbers of participants from other ethnic groups were small. Evidence on seven MCED technologies which were at an unclear stage of development and did not appear to be available for use were briefly summarised; most were evaluated in case-control studies and had a high risk of bias, all studies had high or unclear applicability concerns. No meaningful results were reported relating to patient relevant outcomes, such as mortality, potential harms, HRQoL, acceptability or satisfaction. Recommendations for researchRCTs with sufficiently long follow-up, reporting outcomes that are directly relevant to patients, such as mortality/morbidity, safety, and HRQoL, are needed and some are planned or underway. Research is also needed on the resource implications of MCED tests on NHS services, risk of over-treatment and cost-effectiveness of implementing MCED tests for screening in the UK. Study registrationThis study is registered as PROSPERO CRD42023467901. Funding detailsThis project was funded by the National Institute for Health and Care Research (NIHR) Evidence Synthesis programme (NIHR161758) and will be published in full in the Health Technology Assessment Journal; Vol. XX, No. XX. See the NIHR Journals Library website for further project information.