Modelling the impact of changes to abdominal aortic aneurysm screening and treatment services in England during the COVID-19 pandemic.

BACKGROUND: The National Health Service (NHS) abdominal aortic aneurysm (AAA) screening programme (NAAASP) in England screens 65-year-old men. The programme monitors those with an aneurysm, and early intervention for large aneurysms reduces ruptures and AAA-related mortality. AAA screening services have been disrupted following COVID-19 but it is not known how this may impact AAA-related mortality, or where efforts should be focussed as services resume.
METHODS: We repurposed a previously validated discrete event simulation model to investigate the impact of COVID-19-related service disruption on key outcomes. This model was used to explore the impact of delayed invitation and reduced attendance in men invited to screening. Additionally, we investigated the impact of temporarily suspending scans, increasing the threshold for elective surgery to 7cm and increasing drop-out in the AAA cohort under surveillance, using data from NAAASP to inform the population.
FINDINGS: Delaying invitation to primary screening up to two years had little impact on key outcomes whereas a 10% reduction in attendance could lead to a 2% lifetime increase in AAA-related deaths. In surveillance patients, a 1-year suspension of surveillance or increase in the elective threshold resulted in a 0.4% increase in excess AAA-related deaths (8% in those 5-5.4cm at the start). Longer suspensions or a doubling of drop-out from surveillance would have a pronounced impact on outcomes.
INTERPRETATION: Efforts should be directed towards encouraging men to attend AAA screening service appointments post-COVID-19. Those with AAAs on surveillance should be prioritised as the screening programme resumes, as changes to these services beyond one year are likely to have a larger impact on surgical burden and AAA-related mortality.

Introduction

In March 2020 the initiation of the nationwide “lockdown” to protect against transmission of COVID-19 had a profound effect upon the delivery of routine services provided by the UK National Health Service (NHS). This included a substantial reduction in the number of cardiovascular procedures performed including repair of Abdominal Aortic Aneurysms (AAA) [1]. Furthermore, AAA screening (including surveillance) in most areas of the UK was paused during the lockdown due to concerns about COVID-19 transmission and whilst strategies were considered regarding mitigating risk from delayed surgical intervention (as a result of reduced capacity within the NHS). Ruptured AAA carries a high mortality [2] and screening for AAA is offered to men in their 65th year throughout England via the NHS Abdominal Aortic Aneurysm Screening Program (NAAASP) [3]. Those with small and medium AAA (3.0–5.4 cm) are offered ultrasound surveillance quarterly or annually depending on size, whilst those with large AAA (≥5.5 cm) are referred for consideration of elective surgical repair, before the risk of rupture becomes too high. Circa 300,000 men are offered screening annually, of whom around 1% are found to have an AAA [4], whilst approximately 15,000 men are currently under surveillance in the programme.

During the first lockdown the UK National Joint Vascular Implementation Board [5] suggested that in individuals with AAA measuring 5.5–6.0 cm elective surgery could be delayed for up to 12 months, and those 6.0–7.0 cm for up to 6 months. The capacity to offer elective AAA repair has been severely reduced and the number of elective AAA repairs during the lockdown period fell dramatically to around 12% of pre-COVID procedures per week in April 2020 [6], with a larger proportion >7 cm than pre-COVID-19 (20% versus 10% for elective infra-renal repairs) [6]. Furthermore, nosocomial transmission of COVID-19 (and associated excess mortality) has been reported throughout the UK [7] even after introduction of COVID-minimised pathways (so called GREEN pathways) to reduce the risk of this happening. This has led to a backlog of men awaiting elective surgery, a backlog of unscreened 65-year-old men, and men with known AAA in surveillance who may have gone over the referral threshold. As further waves of COVID-19 infection grip the UK there are uncertainties as to how to manage this situation with a delicate balance between risk of COVID-19 transmission during screening and treatment and competing demands on NHS resources balanced against the risk of untreated AAA rupturing. This is particularly stark when the large number of new cases thought to be due to a more transmissible variant is considered, especially as those with AAA are likely to be at high risk of a poor outcome following COVID-19 infection as there appear to be shared risk factors for both [8].

Here, we use a discrete event simulation model for AAA screening, previously developed and validated to provide evidence for the national screening programme for men in England, to explore different approaches to post-lockdown service resumption.

Methods

Model

NAAASP invites men at age 65 for an ultrasound scan [3]. Those with an aortic diameter <3cm are considered at acceptably low future risk of growth and rupture and are discharged. Those with small AAAs (3.0–4.4 cm diameter) are recalled for ultrasound-based surveillance on an annual basis, and those with medium AAAs (4.5–5.4 cm diameter) every three months. Whilst under surveillance there is a risk that individuals may suffer a ruptured AAA, with the risk of rupture increasing as a function of the size of the aneurysm [9]. Once an AAA reaches 5.5 cm in size, the risk is deemed too great and the individual is referred for a consultation to consider elective repair. These features define the screening programme policy, though at any stage an individual may choose to decline screening or surveillance. It is the policy of the programme to not further follow-up those who decline at any stage, though such individuals may subsequently have incidental detection (or re-detection) of their AAA following investigations outside of the screening programme.

A discrete event simulation (DES) model has previously been developed in R version 3.6.3 [10] to represent the life course of individuals invited to AAA screening and to allow investigation of different screening policies without the need for large clinical trials [11]. The full pathway reflecting both the natural history and screening programme is described in detail elsewhere [10, 11]. In brief, transitions between each state are modelled using rates and probabilities informed by NAAASP, and large randomised trials and studies of AAA screening and surgery (S1 Table). These parameters are used to simulate future event times for key screening and clinical events such as AAA rupture alongside different screening and intervention policies. Here, we recommission this model to investigate the potential impact of changes to NAAASP in the light of COVID-19.

The original DES model simulated events for a new cohort at a given age (e.g. 65) from the time of invitation to screening up to their date of death or age 95 (the time horizon). The repurposed DES model is extended here to allow events to be simulated from a cohort of individuals already under surveillance in the NAAASP, through simulation of key characteristics (age and aortic diameter) at the inception of the model (“time zero”), which is taken to be March 2020 when the initial UK national “lockdown” was imposed. The model is further extended to allow temporary suspension of a) invitation to primary screening, b) surveillance appointments, c) elective surgery, and to allow transitory increases in a) the AAA diameter threshold for elective surgery and b) non-attendance at both primary screening and surveillance scans (which may arise due to individuals engaging in more cautious behaviours as a result of the pandemic [12]). The full model and data underpinning the population initialisation can be found on a GitHub repository at https://github.com/mikesweeting/AAA_DES_model.

Data

We are interested in outcomes for two different cohorts affected by changes to NAAASP services: (1) men turning age 65 who are eligible for the primary screening scan from March 2020 onwards, and (2) men with known AAAs under surveillance within the national screening programme in March 2020. For the surveillance cohort, the characteristics of the modelled population (joint distribution of age and AAA diameter) are informed from data supplied by NAAASP about the population of men under surveillance in May 2020 (personal communication).

Post-COVID-19 policy scenarios

Public Health England and the UK National Joint Vascular Implementation Board published guidelines for the resumption of AAA screening and intervention services respectively in June 2020 [5, 13]. In the first instance, we explore the impact of the proposed post-COVID-19 changes to services in a series of analyses altering each aspect of the screening and intervention process in turn, over a range of potential transition period lengths. The model is broken up into three periods, with the second and third periods differing in length over the different scenarios modelled:

Full lockdown period (months 1–3; corresponds to Apr-Jun 2020 in the UK)

Transition period (subsequent period of reduced service, ranging from 0-60m in length depending on scenario; corresponds to Jul 2020 onwards in the UK)

Return to pre-COVID-19 service (post-transition period)

The full set of scenarios modelled is detailed in Table 1. We investigate the impact of changes relating to the invited group (varying suspension of invitation and attendance rates) and to the surveillance cohort (varying suspension of surveillance scans, drop-out rates, and time at increased size threshold for elective surgery) in a univariate manner. In addition, the cumulative impact of these changes in the surveillance cohort is also assessed.

Table 1

Modelled COVID policy scenarios for AAA services.

Cohort	Model	Status quo model parameters*
Invited 65-year-olds	I0 (status quo)	Attendance: 75%
		Drop-out rate/annum: 6%
		Threshold for surgery: 5.5 cm
Surveillance	S0 (status quo)	Drop-out rate/annum: 6%
		Threshold for surgery: 5.5 cm
		Changes from status quo model**
Invited 65-year-olds	I1	Delay to invitation: ranges from 3 months– 5 years
	I2	Delay to invitation: 6 months
		Attendance: 45–75%
Surveillance	S1	Suspension of surveillance scans: ranges from 3 months– 2 years
	S2.1	Suspension of elective surgery: 3 months
		Suspension of surveillance scans: 6 months
		Drop-out rate/annum: ranges from 6–15% for 1 year
	S2.2	Suspension of elective surgery: 3 months
		Suspension of surveillance scans: 6 months
		Drop-out rate/annum: ranges from 6–15% for 2 years
	S3	Suspension of elective surgery: 3 months
		Suspension of surveillance scans: 6 months
		New threshold for surgery: 7.0 cm for 6 months up to 5 years

* Full details regarding parameter inputs and sources provided in S1 Table.

** Unlisted parameters remain unchanged from status quo. All time periods listed relate to the period following March 2020.

Each scenario model is run for 10 million hypothetical individuals randomly drawn with replacement from the distribution of the relevant population. The models are run for a period of 30 years, with screening policies reverting to pre-COVID-19 norms for the whole of the post-transition period. Model convergence is summarised using cumulative results from consecutive sub-runs each of 1 million individuals (S1 and S2 Figs). Total numbers of AAA-related deaths, operations (both elective and emergency) and ruptures over the whole follow-up period are recorded for each model. These clinical results are reported as percentage change from the status quo as well as expected increase in number of events when scaled to the population of England [14].

Results

Invited 65-year-old cohort

Fig 1 shows the impact on outcomes following different delays to invitation to primary screening amongst current 65-year-olds. Changes to outcomes are small in the first two years of delay, though it is estimated there could be a 4% increase in AAA-related deaths over the lifetime of this cohort if delays in invitation continued for five-years, resulting in 120 additional AAA-related deaths (Table 2). Changes in attendance rates have a more marked impact on outcomes, where reduced attendance could result in 124 (4.5%) more AAA-related deaths and 66 (6.0%) additional emergency operations if attendance dropped from 75% to 55% (Fig 2, Table 2).

Fig 1

65-year-old cohort: Change in key outcomes over varying delay to primary invitation (model I1).

Table 2

Predicted excess AAA deaths and emergency operations in the national invited 65-year-old cohort over 30y period.

Length of delay to invitation	Excess AAA deaths (excess emergency operations) in Model I1*	Attendance rate at primary scan	Excess AAA deaths (excess emergency operations) in Model I2*
6m	0 (0)	65%	56 (29)
12m	0 (0)	55%	117 (62)
24m	0 (0)	45%	175 (94)
36m	18 (9)
48m	55 (29)
60m	112 (59)

* Model I1 = delay to invitation; Model I2 = reduced attendance at primary scan (with 6m delay to invitation)

Notes: National male 65-year-old cohort for England: n = 279,798; expected AAA deaths over 30y in status quo = 2564; expected emergency operations over 30y in status quo = 1041.

Fig 2

65-year-old cohort: Change in key outcomes over varying attendance at primary scan (model I2).

Surveillance cohort: Scan suspension

Suspending ultrasound scans in the surveillance cohort could result in 9 (0.4% increase) additional AAA-related deaths if scans were suspended for one year (Table 3, Fig 3). Of these, 2 (1% increase) are in the sub-group measuring 4.5–4.9 cm at the start of the pandemic and 7 (8% increase) in the sub-group measuring 5.0–5.4 cm; <0.1 are in the 3.0–4.4 cm sub-group. More pronounced effects are evident for suspension for two years and beyond. Suspending surveillance for two years could result in 40 excess AAA-related deaths overall; a 1.9% increase over the lifetime of the surveillance cohort. Of these, 1 is in the 3.0–4.4 cm sub-group and 17 (7% increase) in the 4.5–4.9cm sub-group. However, the remaining 22 excess deaths are in the 5.0–5.4cm range, corresponding to a 24% increase in AAA-related deaths in this sub-group.

Table 3

Predicted excess AAA deaths and emergency operations in the national surveillance cohort over 30y period.

Length of scan suspension	Excess AAA deaths (excess emergency operations) in Model S1*	Length of time at 7cm threshold	Excess AAA deaths (excess emergency operations) in Model S3*	Dropout rate/ annum	Excess AAA deaths (excess emergency operations)
					Model S2.1*	Model S2.2*
6m	2 (1)	6m	2 (1)	8%	46 (24)	84 (44)
12m	9 (5)	12m	9 (5)	10%	84 (44)	152 (81)
24m	40 (22)	24m	41 (23)	12%	120 (63)	219 (116)
36m	110 (61)	36m	99 (55)	15%	174 (92)	314 (167)
48m	230 (126)	48m	176 (98)
60m	408 (222)	60m	260 (146)

* Model S1 = scan suspension; Model S3 = 7cm threshold for elective surgery Model; Model S2.1 = increased dropout for 1y; Model S2.2 = increased dropout for 2y

Notes: National surveillance cohort in March 2020: n = 15,376; expected AAA deaths over 30y in NAAASP status quo = 2152; expected emergency operations over 30y in NAAASP status quo = 745.

Fig 3

Surveillance cohort: Change in key outcomes over varying suspension of surveillance scans (model S1).

Surveillance cohort: Drop-out

In scenarios where the drop-out rate (incorporating declined appointment, out of cohort and non-attendance due to medical reasons) is increased, there is a considerable impact on clinical outcomes (Fig 4). Currently in NAAASP the drop-out rate among those under surveillance is around 6%/annum, though recent research has suggested as few as 59% of invitees may be willing to attend a follow-up appointment during the pandemic [12]. AAA-related deaths could be expected to increase by 84 deaths (3.9%) over the lifetime of the cohort if this drop-out rate increased to 10%/annum for the first year of the pandemic, or 152 (7.1%) excess AAA-related deaths if this increased drop-out continued for two years (Table 3). At 15%/annum drop-out over the course of a year there was an estimated 8.1% increase in AAA-related deaths, resulting in 174 additional AAA-related deaths, and a 14.6% increase if it continued for two years. There is a greater estimated impact on ruptures and emergency operations, which are estimated to increase by around 6% with a 10%/annum drop-out rate for one year (11% for two years) and by 12% with a 15%/annum drop-out rate for one year (22% for two years).

Fig 4

Surveillance cohort: Change in key outcomes over varying dropout rates, applied for (i) 1y (model I2.1) and (ii) 2y (model I2.2).

Surveillance cohort: Threshold for surgery

Fig 5 and Table 3 show the impact of increasing the size threshold for elective surgery to 7cm and varying how long this threshold is applied. There is a modest increase of around 9 (0.4%) AAA-related deaths over the lifetime of this cohort if the increased threshold is applied for one year. However, if the threshold is increased for two years, AAA-related deaths are estimated to increase by 41 (1.9%). Ruptures (and consequently emergency operations) follow a similar pattern, with around a 0.7% increase for one year at the increased threshold, rising to 3.1% for two years.

Fig 5

Surveillance cohort: Change in key outcomes over varying time at increased (7cm) threshold (model I3).

Surveillance cohort: Cumulative impact of changes

S3 Fig shows the results of sequentially adding service changes in order to assess the cumulative impact of these scenarios on AAA-related deaths in the surveillance cohort. S2 Table and Table 2 the corresponding numbers of events for AAA deaths and emergency operations. The results suggest that the impact is additive, with a 10% increase in AAA-related deaths in a scenario with a one year suspension of surveillance scans combined with an increase to 10% drop-out/annum applied for two years and use of a 7 cm threshold for two years. These changes in combination have a similar impact on AAA-related deaths as a four year suspension of surveillance scans alone.

Discussion

The results presented here suggest that short-term pauses in service provision following the COVID-19 outbreak in March 2020 are unlikely to have a large impact on AAA-related deaths. Specifically, scans in those with medium AAAs (4.5–5.4 cm) under surveillance should be resumed within one year, within two years in those with small AAAs (3.0–4.4 cm) under surveillance, and before two years in the invited group, and provided backlogs are cleared quickly. It might be expected that increasing the threshold size for surgery might have a more immediate and dramatic effect on AAA-related deaths. However, in March 2020, any men with AAAs ≥5.5cm had already been referred for elective surgery and are not included in this modelling of the surveillance cohort. Thus, there are no very large AAAs at the outset and it takes some time for AAAs in the 3.0–5.4 cm range to grow to a size where the risk of rupture (and AAA-related death) is very high.

The modelling further shows that any large drop-off in attendance could have a substantial impact on future AAA-related events. The results indicate that a drop to 65% attendance at the primary screening is approximately equivalent to a four year delay in the primary invitation for the 65-year-old cohort in terms of increases in AAA-related deaths. Thus the more significant impact on clinical outcomes may not arise from service change policy, but from reluctance amongst these older men to attend scans during the transition period and being lost to follow-up thereafter. This suggests that there needs to be a concerted effort to ensure men who are due to be screened this year are strongly encouraged to attend screening following the COVID-19 outbreak and on resumption of services. As the national vaccination programme is rolled out, some 65-year old men may wish to postpone their primary screening invitation until after they have been vaccinated. Given the limited impact of delayed invitation in the short term, services can confidently defer the invitation, with re-invitation or encouraged self-referral of those men that have not attended during the lockdown and transition periods. Men on surveillance should be prioritised for vaccination, particularly those with medium AAAs who require an urgent re-scan [15].

Little other work has been carried out on the impact of COVID-19 on AAA services. A recent modelling study in the United States explored the trade-off between COVID-19 mortality and AAA-related mortality. The work focussed on the delay versus immediate repair decision for those with large AAAs, akin to considering an increased threshold for elective surgery [16]. The study uses a decision tree approach rather than the event simulation, rate-based approach used here. This necessarily simplifies the underlying processes; for example, the decision tree model does not include any modelling of AAA growth. The study concludes that the individual patient decision to delay or operate depends on surgical method, age, COVID-19 infection risk, length of delay (3-9m) and diameter. In contrast, we employ a long-term perspective and report results in terms of the impact on AAA-related mortality and surgical burden, scaled to the national population of England.

Modelling assumptions

In addition to assumptions relating to the necessary simplification of the natural history of AAAs and to the estimation of transition rates and probabilities, this modelling work employs a number of additional assumptions related to its use in this context. Specifically, the model does not incorporate any reductions in capacity for screening/surveillance or intervention that may arise due to restrictions on staffing, social distancing or additional cleaning. It is assumed that delayed scans and interventions from the initial lockdown and transition periods can be accommodated immediately in later periods, as soon as services resume; longer modelled delays serve here as a proxy indicator for a slower resumption of services.

We have not modelled the risk of COVID-19 infection and related mortality. Estimates of nosocomial COVID-19 acquisition are not widely available, though published reports provide estimates ranging from 0.1% to 6% of admissions [17, 18]. However, these studies do not differentiate by length of stay, which is typically small for elective AAA repairs (2 and 7 days respectively for endovascular and open repair [19]). The National Vascular Registry in the UK reported 2.2% of those undergoing a vascular procedure tested positive for COVID-19 post-operatively, although this figure does not account for possible pre-operative community-acquired infection based on length of stay. Furthermore, these estimates of nosocomial transmission may not differ greatly from the risk of community-acquisition of COVID-19; in the period July-November 2020, the community incidence of COVID-19 in the UK ranged from 0.5 to 9.5 new cases per 10,000 per day (0.005% to 0.095% of the population per day) [20].

Strengths and limitations

The discrete event simulation model used in this work is underpinned by detailed statistical modelling of AAA growth and rupture rates based on high quality data from a systematic review and meta-analysis of growth and rupture rates of small AAA [21]. Furthermore, the DES has been well validated against data from the Multicentre Aneurysm Screening Study [2] and from a previous Markov model [22], producing reliable estimates of events over the trial follow-up. Data informing the diameter and age distribution of the national surveillance cohort were obtained directly from NAAASP and as such provide a realistic representation of this group for the modelling process.

There are a number of simplifications relating to model structure that were necessary for carrying out this COVID-19-related modelling work. The guidelines are intended for a short-term transition period, and the modelling results here assume that after this time, services will return to pre-COVID-19 levels for the remainder of the 30-year follow-up period tracked by the models. Additionally, the models also assume that any backlog in scans and elective interventions can be rapidly caught up. In practice, both of these aspects are likely to be more complex. The former can and has been explored in modelling, with delays to services well beyond the suggested period explored. However, the latter relates to capacity, which cannot readily be explored in this modelling setup. A backlog in surveillance scans, an increased waiting list for elective surgery together with reduced capacity arising from changes to cleaning protocols and/or staffing would necessarily result in further delays to scans and operations, and thus worsened clinical outcomes. Furthermore, the exclusion of men over 5.5 cm and already referred for surgery in March 2020 from the modelling means that any impact of service changes in this sub-group has not been included in the results presented here.

In addition to these structural assumptions, there are also challenges associated with extrapolating the underlying models of AAA growth and rupture rates to this setting. Specifically, these models are based on data acquired from studies of men with known AAAs under surveillance, where those with AAAs over 5.5cm in diameter generally undergo elective intervention unless they are considered unfit for this procedure. Thus estimates of growth and rupture amongst men with larger AAAs >5.5cm in diameter are based on extrapolations from model projections of men with small AAA (<5.5cm). In the modelling of post-COVID-19 policies here, more men reach these larger sizes, creating a greater influence of these more uncertain parameters in the model.

Conclusions and implications for clinical practice

The relatively large impact of reduced attendance on clinical outcomes points to a careful consideration of the re-invitation policy relating to non-attenders in the surveillance programme during the pandemic. In the context of post-COVID-19 services, introducing future opportunities for scans for non-attenders at both primary and surveillance scans may help counter-act some of these effects. It may also be worthwhile to investigate strategies that may reassure and encourage attendance at both primary and follow-up appointments.

Input parameters used in the discrete event simulation model.

(DOCX)

Click here for additional data file.

(DOCX)

Click here for additional data file.

Convergence in the status quo model in the 65-year-old cohort.

Cumulative proportions for iteration numbers 100,000 to 10m: (a) AAA deaths, (b) ruptures, (c) elective operations.

(TIF)

Click here for additional data file.

Convergence in the status quo model in the surveillance cohort.

Cumulative proportions for iteration numbers 100,000 to 10m: (a) AAA deaths, (b) ruptures, (c) elective operations.

(TIF)

Click here for additional data file.

Cumulative impact of scenarios on surveillance cohort.

S1 = surveillance scan suspension; S2.1 = 10% dropout/annum for 1 year; S2.2 = 10% dropout/annum for 2 years; S3 = 7cm threshold for 2 years before reverting to 5.5cm.

(TIF)

Click here for additional data file.

Transfer Alert

This paper was transferred from another journal. As a result, its full editorial history (including decision letters, peer reviews and author responses) may not be present.

25 Mar 2021

Submitted filename: Kim_PLOSOne_response to reviewers.docx

Click here for additional data file.

30 Apr 2021

PONE-D-21-09840

Modelling the impact of changes to Abdominal Aortic Aneurysm screening and treatment services in England during the COVID-19 pandemic

PLOS ONE

Dear Lois,

Thank you for submitting your manuscript to PLOS ONE. After careful consideration, we feel that it has merit but does not fully meet PLOS ONE’s publication criteria as it currently stands. Therefore, we invite you to submit a revised version of the manuscript that addresses the points raised during the review process.

Please submit your revised manuscript by Jun 14 2021 11:59PM. If you will need more time than this to complete your revisions, please reply to this message or contact the journal office at plosone@plos.org. When you're ready to submit your revision, log on to https://www.editorialmanager.com/pone/ and select the 'Submissions Needing Revision' folder to locate your manuscript file.

Please include the following items when submitting your revised manuscript:

A rebuttal letter that responds to each point raised by the academic editor and reviewer(s). You should upload this letter as a separate file labeled 'Response to Reviewers'.

A marked-up copy of your manuscript that highlights changes made to the original version. You should upload this as a separate file labeled 'Revised Manuscript with Track Changes'.

An unmarked version of your revised paper without tracked changes. You should upload this as a separate file labeled 'Manuscript'.

If you would like to make changes to your financial disclosure, please include your updated statement in your cover letter. Guidelines for resubmitting your figure files are available below the reviewer comments at the end of this letter.

If applicable, we recommend that you deposit your laboratory protocols in protocols.io to enhance the reproducibility of your results. Protocols.io assigns your protocol its own identifier (DOI) so that it can be cited independently in the future. For instructions see: http://journals.plos.org/plosone/s/submission-guidelines#loc-laboratory-protocols. Additionally, PLOS ONE offers an option for publishing peer-reviewed Lab Protocol articles, which describe protocols hosted on protocols.io. Read more information on sharing protocols at https://plos.org/protocols?utm_medium=editorial-email&utm_source=authorletters&utm_campaign=protocols.

We look forward to receiving your revised manuscript.

Kind regards,

Janet Powell

Academic Editor

PLOS ONE

Additional Editor Comments:

Thank you for this interesting submission. It has been reviewed by a modeller and two screening clinicians, one from the UK and one from Sweden. To make your paper more readily understood by screening and other clinicians, please address the comments of reviewers 2 and 3.

Journal Requirements:

When submitting your revision, we need you to address these additional requirements.

1. Please ensure that your manuscript meets PLOS ONE's style requirements, including those for file naming. The PLOS ONE style templates can be found at

and

https://journals.plos.org/plosone/s/file?id=ba62/PLOSOne_formatting_sample_title_authors_affiliations.pdf

2. Please include captions for your Supporting Information files at the end of your manuscript, and update any in-text citations to match accordingly. Please see our Supporting Information guidelines for more information: http://journals.plos.org/plosone/s/supporting-information.

[Note: HTML markup is below. Please do not edit.]

Reviewers' comments:

Reviewer's Responses to Questions

Comments to the Author

1. Is the manuscript technically sound, and do the data support the conclusions?

The manuscript must describe a technically sound piece of scientific research with data that supports the conclusions. Experiments must have been conducted rigorously, with appropriate controls, replication, and sample sizes. The conclusions must be drawn appropriately based on the data presented.

Reviewer #1: Yes

Reviewer #2: Yes

Reviewer #3: Yes

**********

2. Has the statistical analysis been performed appropriately and rigorously?

Reviewer #1: Yes

Reviewer #2: Yes

Reviewer #3: Yes

**********

3. Have the authors made all data underlying the findings in their manuscript fully available?

The PLOS Data policy requires authors to make all data underlying the findings described in their manuscript fully available without restriction, with rare exception (please refer to the Data Availability Statement in the manuscript PDF file). The data should be provided as part of the manuscript or its supporting information, or deposited to a public repository. For example, in addition to summary statistics, the data points behind means, medians and variance measures should be available. If there are restrictions on publicly sharing data—e.g. participant privacy or use of data from a third party—those must be specified.

Reviewer #1: Yes

Reviewer #2: Yes

Reviewer #3: Yes

**********

4. Is the manuscript presented in an intelligible fashion and written in standard English?

PLOS ONE does not copyedit accepted manuscripts, so the language in submitted articles must be clear, correct, and unambiguous. Any typographical or grammatical errors should be corrected at revision, so please note any specific errors here.

Reviewer #1: Yes

Reviewer #2: Yes

Reviewer #3: Yes

**********

5. Review Comments to the Author

Please use the space provided to explain your answers to the questions above. You may also include additional comments for the author, including concerns about dual publication, research ethics, or publication ethics. (Please upload your review as an attachment if it exceeds 20,000 characters)

Reviewer #1: Congratulations on an excellent piece of work. I downloaded the AAA_DEX_model-master folder. I ran the R code. I examined the functions, models and inputs. I appreciated your coding, which is much better than my juvenile efforts. Everything worked as you described in your paper. I didn't have the time to go through all coding lines, so my caveat is that there might be errors.

I have comments that you might or might not find helpful. I think that you need not action any of these for your current manuscript, but they might change some of your thinking for future work. I will take the liberty of emailing you separately, after I have submitted this review, so you feel under no duress to implement any of the following.

a) Outcome The primary metric of any healthcare intervention is quality-adjusted life years. AAA repair and screening have no value if QALYs are unaffected. I assume you primarily report AAA-related mortality because of the apparent failure of RCTs of early vs later repair of AAA < 55 mm to increase survival (UK SAT etc), and similarly a limited effect of screening on AAA rupture, rather than overall survival i.e. perhaps you think that the readers of their paper would dismiss QALYs if they were the primary outcome. My email to you will propose that you use your expertise to address this gaping hole in the logic of scheduled AAA repair: until it is shown to increase QALYs all that scheduled AAA repair can achieve is the replacement of one mode of death (rupture) with another (commonly dementia, cancer, stroke, heart failure, pneumonia). This substitution cannot be assumed to be favourable. I would like to see all papers on AAA repair to consider survival and QALYs as the primary metric. Number of deaths from rupture does not suffice.

b) Other cause mortality The main determinant of QALY with and without scheduled AAA repair is survival with and without AAA repair, which I think is best represented by median life expectancy (although the optimist might consider time to 10% survivorship a better metric for them and the pessimist might consider time to 90% survivorship a better metric). The effect of AAA repair on median survival is most sensitive to variation in ‘death from other causes’, rather than variation in AAA mortality (mostly determined by diameter). You use annual mortality rates in sheet ‘Other cause mortality’ for men aged 65 years to 95 years: I am unsure what the source of these data were. The ONS triennial rates for men in England (or UK) do not quite correspond with the rates you used, but with minimal disparity (I think that you used the ‘qx’ rate rather than the ‘mx’ rate, with which I agree). At least, I did not spot a column in the last six ONS triennial releases that exactly corresponded with your rates. Mortality rates in all age groups and in both men and women have been decreasing over the past century, although rates have increased over the past 12 months, for instance 7917 excess deaths in the age group 65-74 years in the UK. This represents an absolute excess of 1 per 1000 per annum in this age range, whilst age ranges 75-84 and 84+ have increased by 6/1000 and 18/1000 per annum respectively. The conundrum with modelling future survival with and without AAA repair (including screening) is estimating ‘other cause’ mortality rates in the next 30 years. COVID will have killed more of the ‘sick and frail’ men aged 65 years: it is likely that there will be an abnormal reduction in mortality for the next 1-2 years, possibly longer, assuming the national effects of the pandemic resolve. After perhaps five years from the end of the pandemic one might assume that the average annual relative reduction in mortality of 1% to 1.5% might resume. I think that the effect of COVID on patients during hospital admission has much less effect on your model than the effect of COVID outside the hospital (although I haven’t modelled it). If one were going to be precise one could model the probability of COVID infection (possibly asymptomatic) before scheduled AAA repair and the suggestion that 7 weeks’ delay might reduce postoperative deaths to ‘normal’ (e.g. https://associationofanaesthetists-publications.onlinelibrary.wiley.com/doi/10.1111/anae.15464).

c) Heterogeneity of treatment effect My email to you concerns heterogeneity of treatment effect, in the main, which you did not model, but which is crucial for individual decisions and if done well would increase cohort benefit (and reduce harm). As I alluded to above, the effect of scheduled AAA repair on survival is most sensitive to rate of death from other causes: a man or woman with a median life expectancy of two years won’t benefit from AAA repair irrespective of risk of rupture (ie. any diameter); conversely, a man or woman with a much longer life expectancy has longer to accumulate risk of rupture, but more importantly the few deaths from rupture that occur whilst the aneurysm is ‘small’ result in a much greater loss of life and QALY than rupture of larger aneurysm in patients with otherwise short life expectancies. This heterogeneity in ‘other cause mortality’ makes any ‘risk of rupture’ threshold (mainly AAA diameter threshold) nonsense. I cannot avoid the logical conclusion that the decision to operate on intact AAA should depend upon the patient’s rate of ‘other causes of death’ as well as (or actually more than) the risk of AAA rupture. The offer of surgery should be triggered by a modelled increase in median life expectancy in excess of whatever threshold is worthwhile and affordable (maybe a year). I am not arguing against 55mm as a threshold, I am arguing against any threshold that uses AAA diameter alone. Incidentally, were a diameter threshold the correct metric for anyone, it wouldn’t be 55mm. I have modelled the UK SAT: the observed median life expectancy with earlier repair was 1.5 years longer than with later repair was replicated in simulation using the reported rates and times of repair in the two groups. The 55mm threshold is a consequence of UK SAT being underpowered for the particular outcome metric that they specified and the implementation of the protocol making management of the two groups making them more similar than one might suppose from the Methods.

As you know, the survival curves in the UK SAT are specific to that population in 1993. You can imagine what the result was when I simulated the UK SAT in 2020.

d) A general equation for aortic expansion If I interpret your input correctly you modelled an annual increase in aortic diameter of 6%. I appreciate that you are primarily concerned with aortic expansion for a few years after measurement at 40-55mm. Although aneurysmal aortas are reasonably considered ‘different’ to smaller aortas, I think that it is possible to generate a general equation for aortic expansion that would be consistent with measurement of abdominal aortic diameter from birth to the age of 100 years. I spent about one year developing a simulation for a male population born in 1945 and ‘ran it forwards’ to screening at the age of 65 years, using a single equation to determine aortic expansion with age, combined with five equations for rupture (to determine sensitivity). I used ONS general population survival (back to the 1980-2 triennium and then inferred back to 1945) to use with these equations, and I used the distribution of aortic diameter determined on US screening (the same as your sheet ‘AAA Size Distribution’). The following equation for annual expansion was consistent with the number of males surviving to the age of 65 and the distribution of aortic diameters insonated: ((0.000003*power(mm,3))+((0.0017*power(mm,2))-(0.05*mm)+0.45).

e) Equations for rupture Out of interest I compared with your equation for rupture one of the various equations that I’ve developed, and it gives similar results: (0.0000000000003*power(mm,6.2))+(0.0000003*power(mm,2.4))-(0.000029*mm)+0.00024

Reviewer #2: This is a very interesting modelling study examining the impact of COVID on the AAA screening services. It’s timely and relevant, with the only other study like this I know of being a much lower quality publication (reference 13).

However, the paper is complicated and at times can be difficult to follow. It needs simplifying and during revision the methods needs to explain: What data went into creating the model and does it reflect recent publications of NAASP data from post 2018; where do your parameters/presumptions come from and reference them clearly please; and I would seriously consider reducing the number of scenarios as we do have data on how things changed during the worst of the pandemic. The results need to clearly (and simplistically) present how you get to the cumulative impact for each model.

Introduction

1. Line 54 - 55. This should be referenced.

Methods

1. The relevant equator network quality checklist should be added. This may be STRESS for this type of study.

2. Line 81. This needs to be referenced.

3. Line 91. The references in 8 and 9 are from 2018. There are more recent publications on rupture rates from NAAASP. Are these reflected in the model? Please could this information be added. (I note a reviewer has brought this up before but your explanation from line 99 does not make it explicit that you have updated your model).

4. Table 1. Services have largely resumed. Why do you have so many changes from status quo model? The presumptions in these models are confusing and need explaining more clearly.

Results

1. Table 2. Please add column headings which explain the numbers, rather than the term ‘Model” as a heading.

2. Again, I’m confused here having read the preceding text in detail. What is the ‘Period change applied for’? Services have largely resumed and we know how long they were suspended or reduced for.

3. Line 184. Do we have information on drop out rate, or how much we expect it to go up? I’m not sure where these presumptions have come from.

4. Line 120. I’m not getting a good feel for the ruptures, deaths and operations required then cumulative impact as a result of your models. I wonder if this information for each model could be put in a summary table as it is really the crux of the whole study. The figures are of less use and could be appendixes.

Discussion

1. Line 299. You haven’t really looked at the implications for clinical practice. You could model resource requirement, cost etc but that may be beyond the remit of this study. I would limit your conclusions to the results you present in terms of a large excess mortality unless there is careful consideration on how to catch up and encourage men to attend.

Reviewer #3: Thank you for inviting me to review this manuscript. The aim of this paper is to combine knowledge on the NAAASP and the ongoing pandemic on the actual outcome for the invited 65 year old men.

General comments

There is a tsunami of COVID papers, but so far this is a new arena. It is very important for screening-oriented clinicians as well as healthcare providers in general, and the data are scarce.

Since the implementation of screening does carry a large impact on the rupture frequency in the society as well as control activities and number of elective vascular procedures, this data set really does support screening services and vaccination priority services with fresh thoughts!

Struck by the very strong case built through the model, now in April 2021, one could be tickled by the use of including the true story; some parts of the simulation model has actually now passed by and could be reported as “real life data”, did the IRL outcome mirror the model?

Overall really interesting paper, with strong methodology within the limitations of the model as always.

Specific comments:

Overall the authors use “we” too much. Please cut down.

Short title. Uncertain that “modelling changes to AAA screening” ; can be interpreted in different ways?

Abstract

Always prefer a clearly defined objective.

Methods. 3 “ we”…

Would suggest that the registrydata used are mentioned, and the modelling method.

More of aim than method here.

Used thresholds and rationals (5 cm?) is lacking.

Findings.

Why do you use 5 cm here. In the UK in general your used threshold of 5.5 in UL corresponds to almost 6 cm in CT. In most European countries the threshold is 5.5 on CT for treatment. So the choice 5 ? please motivate in text.

Please define Safety.

Introduction

The introduction leans a bit too much on references, which the reader should not have to look up in order to understand the paper.

P3 l 51. Not all repairs ? please define what was restricted for AAA; also was this not regional?

P3 line 63-66. Please present crude numbers since they are published.

Either in table or in text.

Startled by the use “stark”. Is this an English word ?

The aim is understood but could be formulated better (such as the methods text in the abstract) in order to stimulate readers with little experience to read the paper!

Method.

P 4 L 83.

Recalled ? Not the normal vocabulary for screening or surveillance.

Table 1. In the model you use: 7 cm.

Was this really the true threshold during the period; no 6 cm ? 6.5 ?

There is a vast difference probably in rupture risk.

Uncertain of which underlying rupture data you put into the model. Please present.

Results.

It would be fantastic, and interesting to see the Real world data; how was it then ; but understand if this not fits with the paper. It would be nice as a final on the result section.

The dataset is very interesting.

It is of course always interesting to wonder on a modeling of a “not successful screening man “ and a high achiever; meaning; a man that doesn’t come on the first screen invite. Comes on the second. Missed the checkups, turns-up after 5 years with a rupture; vs the “good guy” that comes at all invited occasions. It is highly presumable that the missing outs in the first cohort due to the combined effect of covid- non compliants then will be the dropouts afterwards; it not “new persons”. Does this effect the model ?

Discussion.

Very nice discussion to read and reflect on.

The text on backlog; should fit into the discussion; not only in limitations, since this really is the core critic on modelling; that you cant bring in all aspects of care.

**********

6. PLOS authors have the option to publish the peer review history of their article (what does this mean?). If published, this will include your full peer review and any attached files.

If you choose “no”, your identity will remain anonymous but your review may still be made public.

Do you want your identity to be public for this peer review? For information about this choice, including consent withdrawal, please see our Privacy Policy.

Reviewer #1: Yes: John Bernard Carlisle

Reviewer #2: Yes: Chris Twine

Reviewer #3: Yes: Rebecka Hultgren

[NOTE: If reviewer comments were submitted as an attachment file, they will be attached to this email and accessible via the submission site. Please log into your account, locate the manuscript record, and check for the action link "View Attachments". If this link does not appear, there are no attachment files.]

While revising your submission, please upload your figure files to the Preflight Analysis and Conversion Engine (PACE) digital diagnostic tool, https://pacev2.apexcovantage.com/. PACE helps ensure that figures meet PLOS requirements. To use PACE, you must first register as a user. Registration is free. Then, login and navigate to the UPLOAD tab, where you will find detailed instructions on how to use the tool. If you encounter any issues or have any questions when using PACE, please email PLOS at figures@plos.org. Please note that Supporting Information files do not need this step.

27 May 2021

Reviewer #1:

Congratulations on an excellent piece of work. I downloaded the AAA_DEX_model-master folder. I ran the R code. I examined the functions, models and inputs. I appreciated your coding, which is much better than my juvenile efforts. Everything worked as you described in your paper. I didn't have the time to go through all coding lines, so my caveat is that there might be errors.

I have comments that you might or might not find helpful. I think that you need not action any of these for your current manuscript, but they might change some of your thinking for future work. I will take the liberty of emailing you separately, after I have submitted this review, so you feel under no duress to implement any of the following.

We thank the reviewer for their supportive comments and for taking the time to explore the underlying modelling. We note the comment that this review has provided the comments below for future interest rather than requesting or recommending changes to the current manuscript.

a) Outcome The primary metric of any healthcare intervention is quality-adjusted life years. AAA repair and screening have no value if QALYs are unaffected. I assume you primarily report AAA-related mortality because of the apparent failure of RCTs of early vs later repair of AAA < 55 mm to increase survival (UK SAT etc), and similarly a limited effect of screening on AAA rupture, rather than overall survival i.e. perhaps you think that the readers of their paper would dismiss QALYs if they were the primary outcome. My email to you will propose that you use your expertise to address this gaping hole in the logic of scheduled AAA repair: until it is shown to increase QALYs all that scheduled AAA repair can achieve is the replacement of one mode of death (rupture) with another (commonly dementia, cancer, stroke, heart failure, pneumonia). This substitution cannot be assumed to be favourable. I would like to see all papers on AAA repair to consider survival and QALYs as the primary metric. Number of deaths from rupture does not suffice.

The reviewer is right in taking the view that using AAA deaths and ruptures as outcomes here, this allows the paper to be more readily accessible and quickly understood. This paper focuses on the impact that the pandemic has had on AAA-related outcomes as this is the primary concern of the screening programme. QALYs are a useful output, particularly for cost-effectiveness analyses, though we are not carrying out such an analysis here (see responses to other reviewers below). Of course, when considering policy for AAA screening and intervention, we can only explore how to optimise these services – the question of how to then improve outcomes for other conditions subsequently incurred in a patient in whom AAA death is averted is another matter entirely.

With regard to the 65-year-old cohort, QALYs are averaged over a very large population invited to screening, the majority of whom will not have an AAA; this means any average gains are necessarily very small. Furthermore, the majority of scenarios examined here consider only short-term changes to policy. This means that any additional events will occur at the start of the follow-up period, providing some intuitive sense of the impact on survival for the cohort in question.

b) Other cause mortality The main determinant of QALY with and without scheduled AAA repair is survival with and without AAA repair, which I think is best represented by median life expectancy (although the optimist might consider time to 10% survivorship a better metric for them and the pessimist might consider time to 90% survivorship a better metric). The effect of AAA repair on median survival is most sensitive to variation in ‘death from other causes’, rather than variation in AAA mortality (mostly determined by diameter). You use annual mortality rates in sheet ‘Other cause mortality’ for men aged 65 years to 95 years: I am unsure what the source of these data were. The ONS triennial rates for men in England (or UK) do not quite correspond with the rates you used, but with minimal disparity (I think that you used the ‘qx’ rate rather than the ‘mx’ rate, with which I agree). At least, I did not spot a column in the last six ONS triennial releases that exactly corresponded with your rates. Mortality rates in all age groups and in both men and women have been decreasing over the past century, although rates have increased over the past 12 months, for instance 7917 excess deaths in the age group 65-74 years in the UK. This represents an absolute excess of 1 per 1000 per annum in this age range, whilst age ranges 75-84 and 84+ have increased by 6/1000 and 18/1000 per annum respectively. The conundrum with modelling future survival with and without AAA repair (including screening) is estimating ‘other cause’ mortality rates in the next 30 years. COVID will have killed more of the ‘sick and frail’ men aged 65 years: it is likely that there will be an abnormal reduction in mortality for the next 1-2 years, possibly longer, assuming the national effects of the pandemic resolve. After perhaps five years from the end of the pandemic one might assume that the average annual relative reduction in mortality of 1% to 1.5% might resume. I think that the effect of COVID on patients during hospital admission has much less effect on your model than the effect of COVID outside the hospital (although I haven’t modelled it). If one were going to be precise one could model the probability of COVID infection (possibly asymptomatic) before scheduled AAA repair and the suggestion that 7 weeks’ delay might reduce postoperative deaths to ‘normal’ (e.g. https://associationofanaesthetists-publications.onlinelibrary.wiley.com/doi/10.1111/anae.15464).

Our main interest in this work is a comparison between the “status quo” (pre-covid AAA screening and intervention policy) and short-term changes to services implemented during covid, rather than the repair versus no repair analysis described here. Furthermore, since we hold other assumptions constant (including other cause mortality) across all scenarios, differences in key clinical outcomes should largely reflect differences in policy. In doing so, one assumption we make is that there is no additional risk of covid transmission during a hospital admission over and above the risk in the community. It does not seem unreasonable to expect this additional risk to be very low, and furthermore, there are very little data or high quality published studies that could inform such a parameter.

c) Heterogeneity of treatment effect My email to you concerns heterogeneity of treatment effect, in the main, which you did not model, but which is crucial for individual decisions and if done well would increase cohort benefit (and reduce harm). As I alluded to above, the effect of scheduled AAA repair on survival is most sensitive to rate of death from other causes: a man or woman with a median life expectancy of two years won’t benefit from AAA repair irrespective of risk of rupture (ie. any diameter); conversely, a man or woman with a much longer life expectancy has longer to accumulate risk of rupture, but more importantly the few deaths from rupture that occur whilst the aneurysm is ‘small’ result in a much greater loss of life and QALY than rupture of larger aneurysm in patients with otherwise short life expectancies. This heterogeneity in ‘other cause mortality’ makes any ‘risk of rupture’ threshold (mainly AAA diameter threshold) nonsense. I cannot avoid the logical conclusion that the decision to operate on intact AAA should depend upon the patient’s rate of ‘other causes of death’ as well as (or actually more than) the risk of AAA rupture. The offer of surgery should be triggered by a modelled increase in median life expectancy in excess of whatever threshold is worthwhile and affordable (maybe a year). I am not arguing against 55mm as a threshold, I am arguing against any threshold that uses AAA diameter alone. Incidentally, were a diameter threshold the correct metric for anyone, it wouldn’t be 55mm. I have modelled the UK SAT: the observed median life expectancy with earlier repair was 1.5 years longer than with later repair was replicated in simulation using the reported rates and times of repair in the two groups. The 55mm threshold is a consequence of UK SAT being underpowered for the particular outcome metric that they specified and the implementation of the protocol making management of the two groups making them more similar than one might suppose from the Methods.

As you know, the survival curves in the UK SAT are specific to that population in 1993. You can imagine what the result was when I simulated the UK SAT in 2020.

We agree that this is an important area. However, this is a separate question regarding optimisation of AAA services (in the absence of COVID) by taking a more personalised approach to intervention decisions. The scope of this paper is rather different and investigates the impact of COVID-related changes to services as they are currently implemented.

d) A general equation for aortic expansion If I interpret your input correctly you modelled an annual increase in aortic diameter of 6%. I appreciate that you are primarily concerned with aortic expansion for a few years after measurement at 40-55mm. Although aneurysmal aortas are reasonably considered ‘different’ to smaller aortas, I think that it is possible to generate a general equation for aortic expansion that would be consistent with measurement of abdominal aortic diameter from birth to the age of 100 years. I spent about one year developing a simulation for a male population born in 1945 and ‘ran it forwards’ to screening at the age of 65 years, using a single equation to determine aortic expansion with age, combined with five equations for rupture (to determine sensitivity). I used ONS general population survival (back to the 1980-2 triennium and then inferred back to 1945) to use with these equations, and I used the distribution of aortic diameter determined on US screening (the same as your sheet ‘AAA Size Distribution’). The following equation for annual expansion was consistent with the number of males surviving to the age of 65 and the distribution of aortic diameters insonated: ((0.000003*power(mm,3))+((0.0017*power(mm,2))-(0.05*mm)+0.45).

e) Equations for rupture Out of interest I compared with your equation for rupture one of the various equations that I’ve developed, and it gives similar results: (0.0000000000003*power(mm,6.2))+(0.0000003*power(mm,2.4))-(0.000029*mm)+0.00024

We are grateful that this reviewer has taken time to review the underlying model that we have referenced on github (line 111). The 6% growth rate is taken from one of these files, but refers to the mean percentage increase in diameter per annum; the growth model itself incorporates changes in growth rate according to diameter (full details available in the relevant file and accompanying references).

Reviewer #2:

This is a very interesting modelling study examining the impact of COVID on the AAA screening services. It’s timely and relevant, with the only other study like this I know of being a much lower quality publication (reference 13).

However, the paper is complicated and at times can be difficult to follow. It needs simplifying and during revision the methods needs to explain: What data went into creating the model and does it reflect recent publications of NAASP data from post 2018; where do your parameters/presumptions come from and reference them clearly please; and I would seriously consider reducing the number of scenarios as we do have data on how things changed during the worst of the pandemic. The results need to clearly (and simplistically) present how you get to the cumulative impact for each model.

The full detail of the model has been published elsewhere, which we reference in line 94 of the text. We have now clarified in the text that this reference refers to the model itself (line 92). All of the parameter inputs and sources are also given in the supplementary information accompanying this paper (Supp Table 1; line 97), with full references; details confirming dates have also been added. We have added a footnote to Table 1 in the main paper to highlight this. Age and AAA size distribution at baseline for the surveillance cohort were obtained directly from NAAASP in May 2020 (Suppl Table 1, line 118). AAA size distribution for the invited cohort comes from the first 700,000 men screened in NAAASP (2009-2014) giving an overall prevalence of 1.34%.

There are currently two scenarios relating to the invited cohort and four to the surveillance cohort. The majority of these report on the impact of short-term up to long-term (2-5y post-March 2020) changes in key parameters relating to service delivery. Even if the true covid-related parameter value (e.g. drop-out rate during March 2020 to March 2021) is known, the long-term impact of this change on clinical outcomes is not known, and that is what is investigated here. Furthermore, changes to underlying parameters such as drop-out rates may differ locally, so it is of interest to present results for a range of possibilities, as we do here.

Introduction

1. Line 54 - 55. This should be referenced.

References added to line 54 & 56.

Methods

1. The relevant equator network quality checklist should be added. This may be STRESS for this type of study.

The STRESS checklist for simulation studies has now been completed.

2. Line 81. This needs to be referenced.

Reference added to line 82.

3. Line 91. The references in 8 and 9 are from 2018. There are more recent publications on rupture rates from NAAASP. Are these reflected in the model? Please could this information be added. (I note a reviewer has brought this up before but your explanation from line 99 does not make it explicit that you have updated your model).

There was a publication regarding rupture rates in AAAs <5.5cm in NAAASP published in 2019 (Oliver-Williams et al, Circulation 2019) and refers to follow-up for the years 2009-2017. However, this paper presents results for relatively wide categories of AAA diameter and furthermore the number of ruptures presented in the paper is low (n=31). Here, we employ a more sophisticated joint mixed modelling approach accounting for exact AAA diameter and growth. The model is fitted to data from 11 studies from the RESCAN consortium and uses multivariate meta-analysis to obtain pooled estimates of model parameters. Further information on this is provided in the amended footnote to Supplementary Table 1.

4. Table 1. Services have largely resumed. Why do you have so many changes from status quo model? The presumptions in these models are confusing and need explaining more clearly.

See response to earlier comment regarding reducing the number of scenarios.

A footnote has been added to Table 1 to clarify that parameters not listed in the description of alternative scenarios remain unchanged from the status quo, together with a few notes of clarification in the body of the table.

Results

1. Table 2. Please add column headings which explain the numbers, rather than the term ‘Model” as a heading.

This has been amended as suggested for both Table 2 and Table 3.

2. Again, I’m confused here having read the preceding text in detail. What is the ‘Period change applied for’? Services have largely resumed and we know how long they were suspended or reduced for.

Column headings in Table 2 and 3 have been amended to improve clarity here.

3. Line 184. Do we have information on drop out rate, or how much we expect it to go up? I’m not sure where these presumptions have come from.

A relevant reference has now been added (lines 109, 194-6).

4. Line 120. I’m not getting a good feel for the ruptures, deaths and operations required then cumulative impact as a result of your models. I wonder if this information for each model could be put in a summary table as it is really the crux of the whole study. The figures are of less use and could be appendixes.

This comment refers to line 210, which refers to the start of the section on results of cumulative impact of changes. We agree that the numbers of events are also important: the results in Tables 2 and 3 provide the key figures (AAA deaths and emergency operations) corresponding to Figures 1-5. A similar table has now been added to the supplementary material relating to the cumulative impact results (Suppl Table 2).

Discussion

1. Line 299. You haven’t really looked at the implications for clinical practice. You could model resource requirement, cost etc but that may be beyond the remit of this study. I would limit your conclusions to the results you present in terms of a large excess mortality unless there is careful consideration on how to catch up and encourage men to attend.

There is already strong evidence supporting the cost-effectiveness of aneurysm screening pre-covid (see, for example, Glover et al, Br J Surg 2014). Since all of the scenarios modelled here represent a sub-optimal implementation of the NAAASP screening programme, the goals here are (i) to understand the potential long-term impact of changes that had to be made in light of the pandemic, and (ii) to highlight aspects of the programme which should be prioritised in the return to normal services in order to minimise adverse clinical events. We have therefore focussed our conclusions on these elements, with recommendations to reassure invitees at all stages of screening, and to ensure non-attenders are offered opportunities to return to surveillance (lines 310-315).

An analysis of cost here would not provide the same insight as in a situation where policy decisions can be made based on cost and clinical effectiveness alone. A cost-effectiveness analysis here would likely show that reducing the surgical threshold back to 5.5cm is cost-effective, but cannot account for pandemic-related guidelines that may prevent this. We agree that modelling of limited resource use would be a useful addition, but – as the reviewer suggests – this is beyond the scope of this paper; we have discussed this as a limitation of this work in the manuscript (lines 291-297).

Reviewer #3:

Thank you for inviting me to review this manuscript. The aim of this paper is to combine knowledge on the NAAASP and the ongoing pandemic on the actual outcome for the invited 65 year old men.

General comments

There is a tsunami of COVID papers, but so far this is a new arena. It is very important for screening-oriented clinicians as well as healthcare providers in general, and the data are scarce.

Since the implementation of screening does carry a large impact on the rupture frequency in the society as well as control activities and number of elective vascular procedures, this data set really does support screening services and vaccination priority services with fresh thoughts!

Struck by the very strong case built through the model, now in April 2021, one could be tickled by the use of including the true story; some parts of the simulation model has actually now passed by and could be reported as “real life data”, did the IRL outcome mirror the model?

Overall really interesting paper, with strong methodology within the limitations of the model as always.

We are pleased the reviewer recognises the strengths and value of this work, and the importance of prompt publication. In response to the issue of some time-points in the modelling now being past, please see the reply to Reviewer 2’s first comment.

Specific comments:

Overall the authors use “we” too much. Please cut down.

Some occurrences of “we” have been removed (lines 32, 117, 255, 267)

Short title. Uncertain that “modelling changes to AAA screening” ; can be interpreted in different ways?

Short title has been amended accordingly (line 22)

Abstract

Always prefer a clearly defined objective.

Methods. 3 “ we”…

Would suggest that the registry data used are mentioned, and the modelling method.

More of aim than method here.

Used thresholds and rationals (5 cm?) is lacking.

Use of NAAASP population data now added (line 35). Description of the modelling method (discrete event simulation) is in line 31. Unfortunately it is not possible to provide the rationale for all the scenarios within the constraints of the word limit for the abstract; the details of this are contained within the main text of the paper.

Findings.

Why do you use 5 cm here. In the UK in general your used threshold of 5.5 in UL corresponds to almost 6 cm in CT. In most European countries the threshold is 5.5 on CT for treatment. So the choice 5 ? please motivate in text.

Since this section is describing results in those under surveillance at the start of the pandemic, they are necessarily under 5.5cm. This statement provides results in terms of excess deaths for the very largest of those under surveillance at that time, namely those 5.0-5.4cm. This has now been briefly clarified in the abstract (line 40); full details are in the main text (lines 173-176).

Please define Safety.

We now use the word “outcomes” here (line 41).

Introduction

The introduction leans a bit too much on references, which the reader should not have to look up in order to understand the paper.

P3 l 51. Not all repairs ? please define what was restricted for AAA; also was this not regional?

Reference 1 provides an indication of change in overall admissions for aortic aneurysm (53% reduction in lockdown), though the reduction is likely to be more pronounced within the sub-group of elective rather than emergency repairs. It is also likely there are regional variations reflecting local descision-making, though these data (and our model) present results nationally.

P3 line 63-66. Please present crude numbers since they are published.

Either in table or in text.

This information from reference 6 has now been added to the introduction (lines 64-66).

Startled by the use “stark”. Is this an English word ?

The aim is understood but could be formulated better (such as the methods text in the abstract) in order to stimulate readers with little experience to read the paper!

Please see above for changes made to the abstract.

Method.

P 4 L 83.

Recalled ? Not the normal vocabulary for screening or surveillance.

We have used this terminology here to help distinguish invitations that are part of the surveillance programme from those that relate to the primary screening invitation.

Table 1. In the model you use: 7 cm.

Was this really the true threshold during the period; no 6 cm ? 6.5 ?

There is a vast difference probably in rupture risk.

It is true that the rupture risk for 6cm v 6.5cm v 7cm AAAs differs significantly. However, because we are modelling only those men under surveillance in March 2020 (i.e. those <5.5cm), for scenarios with relatively short-term changes to the threshold, very few men will grow as large as 7cm in this period. In effect, this means that for short-term changes to the threshold, there will be little difference in the results from models applying 6cm, 6.5cm or 7cm thresholds. This is described in the discussion in lines 256-260. Here we use a 7cm threshold to reflect the NJVIB publication summarised in the Introduction, though there is also some evidence relating to the use of 7cm thresholds (see reference 6).

Uncertain of which underlying rupture data you put into the model. Please present.

Details have been added to the footnote of Suppl Table 1.

Results.

It would be fantastic, and interesting to see the Real world data; how was it then ; but understand if this not fits with the paper. It would be nice as a final on the result section.

As discussed in the response to Reviewer 2, although it may be possible to acquire information about what did happen in practice in terms of these policies for the initial period following lockdown (and noting that relevant large-scale data are not yet published), we are interested here in the long-term impact of these policies on clinical outcomes, which of course is not available for comparison.

The dataset is very interesting.

It is of course always interesting to wonder on a modeling of a “not successful screening man “ and a high achiever; meaning; a man that doesn’t come on the first screen invite. Comes on the second. Missed the checkups, turns-up after 5 years with a rupture; vs the “good guy” that comes at all invited occasions. It is highly presumable that the missing outs in the first cohort due to the combined effect of covid- non compliants then will be the dropouts afterwards; it not “new persons”. Does this effect the model ?

I think the question here is about double-counting in the surveillance cohort: those individuals who would have dropped out during follow-up are the same individuals who would now not attend due to COVID. Since we are still modelling drop-out that would usually occur in addition to non-attendance due to COVID, we may over-estimate non-attendance. It is an interesting point, though of course not readily explored or supported by data. The approach we take here is to explore a range of short-term increase to dropout rates, including no increase at all (which could be considered to model a scenario whereby those not attending due to COVID are the same individuals who would not have attended anyway).

Discussion.

Very nice discussion to read and reflect on.

The text on backlog; should fit into the discussion; not only in limitations, since this really is the core critic on modelling; that you cant bring in all aspects of care.

We agree that the discussion around potential backlog of scans and operations is a critical one. However, since this is beyond the scope of the modelling carried out here, we have included this under the subheading of limitations, within the discussion section.

________________________________________

Submitted filename: Kim_PLOSOne_STRESS_checklist.docx

Click here for additional data file.

3 Jun 2021

Modelling the impact of changes to Abdominal Aortic Aneurysm screening and treatment services in England during the COVID-19 pandemic

PONE-D-21-09840R1

Dear Lois,

We’re pleased to inform you that your manuscript has been judged scientifically suitable for publication and will be formally accepted for publication once it meets all outstanding technical requirements.

Within one week, you’ll receive an e-mail detailing the required amendments. When these have been addressed, you’ll receive a formal acceptance letter and your manuscript will be scheduled for publication.

An invoice for payment will follow shortly after the formal acceptance. To ensure an efficient process, please log into Editorial Manager at http://www.editorialmanager.com/pone/, click the 'Update My Information' link at the top of the page, and double check that your user information is up-to-date. If you have any billing related questions, please contact our Author Billing department directly at authorbilling@plos.org.

If your institution or institutions have a press office, please notify them about your upcoming paper to help maximize its impact. If they’ll be preparing press materials, please inform our press team as soon as possible -- no later than 48 hours after receiving the formal acceptance. Your manuscript will remain under strict press embargo until 2 pm Eastern Time on the date of publication. For more information, please contact onepress@plos.org.

Kind regards,

Janet Powell

Academic Editor

PLOS ONE

Additional Editor Comments (optional):

Thanks for responding carefully to the reviewer comments

Reviewers' comments:

Reviewer's Responses to Questions

Comments to the Author

1. If the authors have adequately addressed your comments raised in a previous round of review and you feel that this manuscript is now acceptable for publication, you may indicate that here to bypass the “Comments to the Author” section, enter your conflict of interest statement in the “Confidential to Editor” section, and submit your "Accept" recommendation.

Reviewer #2: All comments have been addressed

**********

2. Is the manuscript technically sound, and do the data support the conclusions?

The manuscript must describe a technically sound piece of scientific research with data that supports the conclusions. Experiments must have been conducted rigorously, with appropriate controls, replication, and sample sizes. The conclusions must be drawn appropriately based on the data presented.

Reviewer #2: Yes

**********

3. Has the statistical analysis been performed appropriately and rigorously?

Reviewer #2: Yes

**********

4. Have the authors made all data underlying the findings in their manuscript fully available?

The PLOS Data policy requires authors to make all data underlying the findings described in their manuscript fully available without restriction, with rare exception (please refer to the Data Availability Statement in the manuscript PDF file). The data should be provided as part of the manuscript or its supporting information, or deposited to a public repository. For example, in addition to summary statistics, the data points behind means, medians and variance measures should be available. If there are restrictions on publicly sharing data—e.g. participant privacy or use of data from a third party—those must be specified.

Reviewer #2: Yes

**********

5. Is the manuscript presented in an intelligible fashion and written in standard English?

PLOS ONE does not copyedit accepted manuscripts, so the language in submitted articles must be clear, correct, and unambiguous. Any typographical or grammatical errors should be corrected at revision, so please note any specific errors here.

Reviewer #2: Yes

**********

6. Review Comments to the Author

Please use the space provided to explain your answers to the questions above. You may also include additional comments for the author, including concerns about dual publication, research ethics, or publication ethics. (Please upload your review as an attachment if it exceeds 20,000 characters)

Reviewer #2: Great changes.

Only comment is that putting the number of patients as well as % in the abstract woful help give context.

**********

7. PLOS authors have the option to publish the peer review history of their article (what does this mean?). If published, this will include your full peer review and any attached files.

If you choose “no”, your identity will remain anonymous but your review may still be made public.

Do you want your identity to be public for this peer review? For information about this choice, including consent withdrawal, please see our Privacy Policy.

Reviewer #2: Yes: Chris Twine

7 Jun 2021

PONE-D-21-09840R1

Modelling the impact of changes to abdominal aortic aneurysm screening and treatment services in England during the COVID-19 pandemic

Dear Dr. Kim:

I'm pleased to inform you that your manuscript has been deemed suitable for publication in PLOS ONE. Congratulations! Your manuscript is now with our production department.

If your institution or institutions have a press office, please let them know about your upcoming paper now to help maximize its impact. If they'll be preparing press materials, please inform our press team within the next 48 hours. Your manuscript will remain under strict press embargo until 2 pm Eastern Time on the date of publication. For more information please contact onepress@plos.org.

If we can help with anything else, please email us at plosone@plos.org.

Thank you for submitting your work to PLOS ONE and supporting open access.

Kind regards,

PLOS ONE Editorial Office Staff

on behalf of

Dr. Janet Powell

Academic Editor

PLOS ONE