Modelled mortality benefits of multi-cancer early detection screening in England

Background Screening programmes utilising blood-based multi-cancer early detection (MCED) tests, which can detect a shared cancer signal from any site in the body with a single, low false-positive rate, could reduce cancer burden through early diagnosis. Methods A natural history (‘interception’) model of cancer was previously used to characterise potential benefits of MCED screening (based on published performance of an MCED test). We built upon this using a two-population survival model to account for an increased risk of death from cfDNA-detectable cancers relative to cfDNA-non-detectable cancers. We developed another model allowing some cancers to metastasise directly from stage I, bypassing intermediate tumour stages. We used incidence and survival-by-stage data from the National Cancer Registration and Analysis Service in England to estimate longer-term benefits to a cohort screened between ages 50–79 years. Results Estimated late-stage and mortality reductions were robust to a range of assumptions. With the least favourable dwell (sojourn) time and cfDNA status hazard ratio assumptions, we estimated, among 100,000 screened individuals, 67 (17%) fewer cancer deaths per year corresponding to 2029 fewer deaths in those screened between ages 50–79 years. Conclusion Realising the potential benefits of MCED tests could substantially reduce late-stage cancer diagnoses and mortality.

. Survival by Cancer Type and Stage.
We carried out survival analyses based on persons for all cancers except sex-specific cancers (prostate, uterus, cervix, and ovary For some combinations of cancer site, stage, and age band, there were no events and therefore survival could not be calculated. In those cases, we imputed data by assuming the survival of the nearest relevant older age band. This is a conservative assumption as survival in the older age group is typically worse. Net survival estimates were considered potentially unreliable when one of the following criteria were met: standard error (SE) > 0.2; net survival estimate > 1 (e.g., in breast cancer); number of patients < 5; and/or difference between upper and lower confidence intervals (CIs) > 20 (as shown in the

Introduction
To consider the implications for the benefit of screening based on a MCED test among cancers with non-sequential stage progression, we developed a mixture model whereby different proportions of cancers that are currently clinically diagnosed at stage IV under usual care are assumed to have either (A) a short stage I (of six months in length), or (B) no previous stages.
These models represent a range of different biological scenarios including cancers metastasizing directly from a preclinical or stage I cancer, as well as metastatic subclones originating from local tumours; the biology of these cancers not only leads to distant metastases before local spread, but also accelerate progression from the earliest stage of disease. For all other cancers, stage progression was assumed to be sequential throughout.

We implemented model
These models were mixed with the main sequential stage progression model, with three different proportions of non-sequential stage progression for the 11 pre-specified cancers: 0.1, 0.25, and 0.5.
In these models, we used the medium dwell time assumption with hazard ratio (HR) = 3 for cfDNA-detectable cancers. We used the national screening programme framework to better understand the long-term benefits of an MCED screening programme with these various proportions of cancers with non-sequential stage progression, as the open cohort may be unduly influenced by the prevalent round of screening.

Results
Results are displayed in Tables 4A and S4B. The number of cancers available to be found via usual care (i.e., without MCED screening) increased with the proportion of cancers that have non-sequential progression. Decreasing the amount of time spent in stage I brought forward in time all cancers that would have otherwise been diagnosed clinically in subsequent years.
Similarly, with increased proportions of cancers with non-sequential progression, the diagnostic yield via MCED screening was smaller. This reflects limited or no opportunities for MCED detection among these cancers, for which the MCED test typically has high sensitivity, at earlier stages. When these cancers are diagnosed via MCED screening at stage IV, no mortality benefit is conferred for these tumours. For diagnosis via MCED screening, the number of cancers diagnosed at a late stage increased as the proportion of non-sequential progression cancers increased. This is because when cancers spend no time in any stages prior to stage IV, intercepting these cancers earlier and reducing the amount of late-stage diagnoses is not possible.
In the most adverse scenario (in which 50% of cancers have non-sequential progression, with no time spent in stages prior to stage IV), 33% of cancers were estimated to be diagnosed late via usual care and MCED screening. The cancer mortality rate in the absence of MCED screening was highest (33%) in the scenario with the most adverse cancer biology, as more cancers with the most rapidly lethal biologies will be diagnosed during the national screening programme period. However, cancer mortality was 31% in the absence of MCED in the national screening programme scenario with a 'fast' dwell time and HR = 3 for cfDNA + tumours (Table 4); this suggests that the most adverse scenario may be too pessimistic. Again, in the most adverse scenario, the reduction in cancer mortality was 13%, which is 6% lower than the cancer mortality reduction in the sequential progression model with a medium dwell time and HR = 3 (Table 4). National screening programme, medium dwell time scenario, HR = 3 for cfDNA status. The total number of cancers found via usual care and via screening with an MCED test with sensitivity as estimated in a case-control study (11) when added to usual care differed between each of the non-sequential stage progression scenarios. This reflects the fact that non-sequential stage progression reduces the effective time spent in the pre-clinical state, which affects both the prevalent round at the start of the programme, and the last round of the programme. National screening programme, medium dwell time scenario, HR = 3 for cfDNA status. The total number of cancers found via usual care and via screening with an MCED test when added to usual care differed between the non-sequential stage progression scenarios. As in Model A, this reflects the fact that non-sequential stage progression reduces the effective time spent in a preclinical state, which affects both the prevalent round at the start of the programme, and the last round of the programme. indeed the benefit gained under the most favourable set of assumptions, is sufficient to outweigh the harms associated with false positives remains to be seen. The resource implications of a MCED screening programme are also currently unknown. We will gain insight into these issues, and more fundamental issues such as the natural history of many of these cancers, in the NHS-Galleri trial. It will be important to revisit the structural assumptions related to natural history when the results of this trial and others become available, and to triangulate the evidence from fundamental studies into the evolution of cancers. Dwell times used in this study (in years) were from a previous publication by Hubbell and colleagues (2021). We renamed the scenarios so that medium, fast and aggressive-fast dwell times in the previous publication corresponds to slow, medium and fast dwell times in this study. To continue the example, if test sensitivity at stage II is 50%, an additional 30% of all cancers will be detectable at this stage (incremental sensitivity). The additional 30% will be intercepted at stage II if screening occurs before the next stage progression.

Cancer Type
The proportion of cancers intercepted at each stage are those that are detectable and for which a screening test is performed prior to stage progression. To obtain the proportion of cancers that go through a screening test in the detectable period, evaluation over the dwell time distributions is required (see equations below). In our analysis, we assumed exponential dwell-time distributions. The intercepted fractions were calculated piecewise by stage-at-diagnosis in usual care. Cancers could progress through more than one stage during the annual screening interval; the probability that a cancer developed and proceeded through all stages between screens was dependent on dwell time distribution.

Equations
This information was presented in the supplemental data section of the original interception model (14). To calculate the proportion of incident cancers intercepted, we considered only cancers detected for a positive duration of time (t) prior to clinical diagnosis. The cumulative distribution function for preclinical duration of these cancers is defined as: For the prevalent round of screening, where the screening interval is effectively indefinitely long, the integral is unrestricted: (1 − ( )) Using the same logic, the probability of a cancer being intercepted at a given stage can be approximated using the cumulative distribution of time in stage instead of the whole detectable time interval.

Figure S8. Cohort Weights in the National Screening Programme.
At 50 years of age, 100,000 participants entered the national screening programme and 'aged' through the programme until they were 79 years old. The cohort weights were applied at each age to account for those who have exited the cohort due to death, emigration, or diagnosis with cancer.  The incidence rate for cfDNA detected cancer by age is displayed for each of the dwell time scenarios. Intuitively, the greatest number of cancers are diagnosed in the 'slow' dwell time scenario, because cancers spend more time in each stage, increasing the likelihood they will be screen-detected. The MCED screen-detected incidence rate increases steadily with age, until age 75 years, when it appears to decrease; this reflects the impact of competing causes of mortality.