This review reflects comments by Louis Fisher, Helen Curtis, Rose Higgins and Brian MacKenna generated as part of a preprint journal club. Review synthesised by Louis Fisher.
The COVID-19 pandemic led to the postponement of cancer screening programmes and reductions in diagnostic testing. It is possible that this could have resulted in delayed diagnosis of cancer and an increase in avoidable cancer deaths. This study investigates if there were any changes across the UK in the incidence rate of screening, consultations, diagnostic testing and diagnoses of breast, colorectal, lung and prostate cancers during the COVID-19 pandemic period. It uses data from a subset of UK general practices available in CPRD Gold between January 2017 and December 2021. Modelling pre-pandemic data, the authors also provide an estimation of the number of missed cancer diagnoses during the lockdown period.
The authors find that the first UK national lockdown resulted in reductions in screening, diagnostic tests and referrals for the cancers investigated, especially for mammograms, colonoscopies and visits to breast surgeons. This reduction persisted beyond the lockdown period, up until December 2021. The authors estimate that there were 62,000 missed cancer diagnoses in the UK between March 2020 and December 2021.
The results are an important indication of the need to manage the backlog in screening/diagnostic services post-pandemic. Below we highlight some points which we feel could make this manuscript stronger.
CPRD Gold includes all general practices using Vision EHR software, which in the most recent release includes ~5% of all UK general practices 1. The geographic distribution of these practices varies and is ever changing. Given that Vision EHR software is currently only used in 10 practices in England 2 and the data profile referenced is Herrett et al from 2015 3, could the authors comment on the representativeness of the study? This is important to understand given the extrapolation of the predicted missed diagnoses to the entire UK population.
In Figure 4, the expected rate of colorectal and lung cancer appear to increase from March 2020. Why is this? This is unexpected given the pre-lockdown rates.
We appreciate the availability of the codelists and study code used in this study, but it is not clear how the codelists were developed. Can the authors provide more detail on this to allow for a clearer understanding of the clinical activity intended to be captured by each codelist?
The authors highlight that the shift of cancer diagnoses to secondary care in the pandemic could have resulted in reduced/delayed coding in primary care. Do they think recording activity could also have been impacted by temporary suspension of QOF during the pandemic?
Similarly, not all of the cancers investigated have a national cancer screening programme in the UK. There is no national screening programme for pancreatic cancer and one for lung cancer has only recently been recommended and is slowly being rolled out in England. Could this have affected the observed incidence rates?
The post-lockdown period analysed is short, only continuing until December 2021. Updating the results with more recent data prior to publication would allow insight into continued reductions in diagnostic testing/screening and would highlight any rebound in diagnoses. A Longer time period would also provide enough time to accrue for investigation of any changes in stage at diagnosis.
Is it possible to see the pattern of diagnoses broken down by country?
In the methods section, the predicted number of missed diagnoses was described as:
"Number of ‘missing’ diagnoses were calculated as the difference between forecasted (observed) and expected number of incident cancer diagnoses during each time-period"
Should this be
"Number of ‘missing’ diagnoses were calculated as the difference between forecasted (expected) and observed number of incident cancer diagnoses during each time-period"
The number of patients excluded from the initial study population is described in Table S2
11M patients are removed from the source population because "No observation time available during study period (2017 - Dec 21)". It is not clear in the manuscript who this is removing. Could the authors describe this filter in more detail, as it removes a large number of patients. Were patients required to be registered with the same practice for the entire period?
There are also two exclusion steps for removing patients on age that don’t remove anyone. From the methods it appears there is no age restriction.
We feel the supporting literature could be improved by including some more recent work. Some examples include
Modelling of avoidable cancer deaths due to diagnostic delays.
Impact of the pandemic on lung cancer screening
Colorectal screening in the Netherlands predicted minimal long term impact of screening suspension
Figure 1 could be better represented as a timeline. There is no information on the y-axis.
Inclusion of rates as well as counts in this Table 1 would allow for easier comparison between the different time periods (which are for different lengths of time)
The figure axes/labels/legends could be increased in size for better legibility.
Figure 4. Is the time period in the legend correct? Should it be January 2017 to February 2020?
The author declares that they have no competing interests.