A method for evaluating breast cancer screening strategies using screen-preventable loss of life.

The objective of this study is to describe how screen-preventable loss of life (screen-PLL) can be used to analyze the distribution of life savings with mammographic screening. The determination of screen-PLL with mammography is possible using a natural history model of breast cancer that simulates clinical and pathologic events of this disease. This investigation uses a Monte Carlo Markov model with data from the Surveillance, Epidemiology, and End Results Program; American Cancer Society; and National Vital Statistics System. Populations of one million women per screening strategy are simulated over a lifetime with mammographic screening based on current guidelines of the American Cancer Society (ACS), United States Preventive Services Task Force (USPSTF), triennial screening from age 50-70, and no screening. Screen-PLL curves are generated and show guideline performance over a lifetime. The screen-PLL curve with no screening is determined by tumor discovery through clinical awareness and has the highest values of screen-PLL. The ACS and USPSTF strategies demonstrate screen-PLL curves favoring the elderly. The curve for triennial screening is more uniform than the ACS or USPSTF curves but could be improved by adding screen(s) at either end of the 50-70 age range. This study introduces the use of screen-PLL as a tool to improve the understanding of screening guidelines and allowing a more balanced allocation of life savings across an aging population. The method presented shows how screen-PLL can be used to analyze and potentially improve breast cancer screening guidelines.

Introduction

The goal of screening is to detect early disease when treatment is more likely to be beneficial or lifesaving [1, 2]. The opportunity to save life-years through screening is expressed mathematically as screen-preventable loss of life (screen-PLL). If an effective intervention is available, the difference in years between individuals’ premature deaths and the underlying population’s life expectancy is a measure of the potential life savings.

This idea of calculating preventable loss of life for a disease is not new. Examples of this principle are found in analyzing the changing mortality of tuberculosis in the 1940s and more recently in measuring the burden of disease, such as chronic obstructive pulmonary disease [3, 4]. However, application of this concept to mammography screening and breast cancer survival has not been widely described in the medical literature.

Since the introduction of mammography screening in the 1960s, clinical trials have shown mortality reduction, with a few studies showing conflicting results [5-7]. The debate about the use and starting age of mammography screening continues to the present day, as there remains no agreed standard to measure screening performance over a lifetime [8-11]. The United States Preventive Services Task Force (USPSTF) recommends biennial screening for women aged 50 to 74 with average-risk women aged 40–49 encouraged to discuss the advisability of screening with their providers [7]. The American Cancer Society (ACS) recommends annual screening from age 45 to 55, followed by biennial mammography until life expectancy is less than 10 years [12]. In addition, the ACS allows for the opportunity to start screening with annual mammography between the ages of 40 and 44 and to continue yearly mammography at age 55 and older. A strategy of triennial screening from ages 50–70 is similar to the recommendation of The National Health Service in the United Kingdom [13].

In reality, data needed to calculate screen-PLL for breast cancer such as the size and lethality of undetected tumors, time of metastasis, and expected ages of death from breast cancer or alternative causes are not clinically available. However, a computer model of breast cancer’s natural history can evaluate these factors and permits insights not possible in clinical practice. A published model of breast cancer screening [14] is adapted to measure screen-PLL.

The objective of this study is to describe a computer modeling approach to determine the screen-preventable life savings achievable within an aging population of women screened by the ACS and USPSTF guidelines, as well as triennial screening and no screening.

Methods

Screen-PLL is the number of preventable months of life lost to breast cancer from progressive tumors that are mammographically detectable at an early stage and curable. In this study, the screen-PLL value for a simulated individual is calculated at all ages. An assumption is that mammographic detectability starts at a tumor size of 0.2 cm [15]. However, some undetected tumors may lead to distant metastases and premature death from breast cancer. For each woman who dies prematurely, screen-PLL is calculated as the difference between her expected age of death from general mortality and her age of death from breast cancer, as seen in Fig 1. In this figure, screen-PLL is zero until the tumor reaches 0.2 cm. Subsequently, screen-PLL increases to the difference between her expected non-breast cancer death age and her breast cancer death age and remains constant. Screen-PLL returns to zero with metastasis or detection by mammography or clinical signs and symptoms. Detection eliminates the need for further screening. In the model, metastatic disease is assumed incurable since death in most breast cancer cases is associated with metastases [16-18]. Once metastases occur, further screening offers no life-saving benefit. With detection or with presence of metastatic disease, the opportunity to prevent a breast cancer death through screening ends.

Fig 1

Screen-preventable loss of life in relation to disease progression.

FMC, first malignant cell; BCA, breast cancer; PLL, preventable loss of life. Mammographic detectability begins at age t1 with a tumor size of 0.2 cm. The detectable range of a primary tumor is from 0.2 cm (t1) to BCA death (t3). Screen-PLL is the age difference between the expected non-BCA death and BCA death, defined as t4-t3. Screen-PLL is 0 from the FMC to the tumor size of 0.2 cm, the value of t4-t3 from 0.2 cm to detection or metastasis, and 0 thereafter.

As an example, consider a 60-year-old woman who has a localized progressive primary tumor that is 0.2 cm in diameter. Without screening, metastatic disease is predicted by the model to occur at age 65 and death from breast cancer at age 72. This woman’s expected age of death from other causes in the model is 92. This is a case of premature death from breast cancer that could have been prevented by mammography. The screen-PLL is 240 months or 20 years, calculated as (92–72) ×12. The screen-PLL of 240 months exists for this individual starting with detectability at age 60 and extending until metastasis at age 65. Once metastases occur, screening mammography is considered of no value in preventing breast cancer death, and screen-PLL is zero thereafter.

A cohort’s screen-PLL at a given age is the sum of all individuals’ screen-PLLs. The cohort’s screen-PLL curve is constructed by plotting the screen-PLL over time. Screen-PLL curves are created for three groups of one million women each, following the ACS, USPSTF and triennial strategies. The ACS guideline is modeled according to the base recommendation and does not include the options for earlier screening and for continued annual mammography at age 55 and older. A screen-PLL curve is also created for an unscreened cohort of one million women in whom clinical signs and symptoms are the only means of detection.

The characteristics of the ACS, USPSTF, and triennial screen-PLL curves are analyzed. A flat curve indicates that screen-PLL is distributed evenly across ages, resulting in an equitable allotment of life savings to the population.

The study uses a published Monte Carlo Markov natural history model for breast cancer, calibrated using statistics from the Surveillance, Epidemiology, and End Results Program (SEER), the American Cancer Society and the National Vital Statistics System [14, 19–21]. In the model, a woman begins at birth with no cancer [14]. Breast cancer arises with a first malignant cell and follows Gompertzian growth with disease progression dependent on tumor volume [22-24]. Age-specific incidence data for breast cancer among average-risk white women in the United States are used to determine the age of onset of clinical disease [25]. The proportion of non-progressive tumors, which are non-lethal and contribute to overdiagnosis, is assumed in our model to be 30%, based on a finding in an analysis of SEER data [26]. It is assumed that during the study, women are 100 percent compliant with their assigned screening strategy. Sensitivity of mammography increases with tumor diameter, starting at 0.2 cm, and is assumed higher for post-menopausal women [14]. Stopping ages for the ACS guideline rule, 10-year or less life-expectancy, were calculated using life table data [14]. Tumor detection results from either mammography screening or from clinical signs and symptoms. The disease is classified into one of seven clinical stages using the American Joint Committee on Cancer’s criteria [27]. Death may occur from either metastatic breast cancer or from other causes. In the case of a breast cancer death, life years lost to breast cancer are calculated as the difference between the projected death age from general mortality and the death age from breast cancer.

Results

Fig 2 shows the distributions of screen-PLL in months for cohorts of one million women each, following the ACS, USPSTF, and triennial strategies, as well as for an unscreened cohort of one million women labeled as No Screening. At birth, there is no screen-PLL. As the cohort ages and breast cancer incidence rises, the screen-PLL increases. For the unscreened cohort, the screen-PLL at age 30 is approximately 75,000 life-months. The screen-PLL continues to increase until a plateau of about 575,000 life-months from approximately age 45 to 65. After age 65, screen-PLL rapidly decreases and approaches zero. The USPSTF curve follows the No Screening curve until about age 43. After that age, the USPSTF curve decreases before the onset of screening. Shortly after screening stops at age 74, the USPSTF and No Screening curves converge. The ACS curve decreases significantly with the start of annual screening at age 45 and converges with the USPSTF curve with the start of biennial screening at age 55. The curves stay in relative alignment until about age 74, when USPSTF screening ends. The ACS curve subsequently dips and eventually converges again with the USPSTF curve and the No Screening curve at around age 80 when ACS screening ends. The triennial strategy PLL curve follows the USPSTF curve closely until screening begins at age 50. The triennial PLL curve remains slightly above the USPSTF curve throughout and converges with No Screening at age 70.

Fig 2

Breast cancer screen-PLL for the ACS screening guideline, USPSTF screening guideline, triennial screening strategy, and no screening.

One million women in each group. Screen-PLL, screen-preventable loss of life; ACS, American Cancer Society; USPSTF, United States Preventive Services Task Force.

Discussion

Screening mammography remains an important preventive strategy despite some enduring controversies. Since a mammography guideline determines the number of possible screening tests, improving uniformity in the distribution of life savings across age groups ensures equitability. This notion has attracted little attention in the design of screening strategies, as if all participants would benefit equally simply by participating in a screening program. This is clearly not the case. Guideline recommendations determine the distribution of life savings from screening within a population. The concept of screen-PLL highlights the differences between guidelines.

The ACS and USPSTF guidelines continue screening for women into their 70s when, due to limited life expectancy, the screen-PLL is lower than at any other age. This favors elderly women. Screen-PLL for ACS, USPSTF, and triennial strategies are highest for women in their 40s. This non-uniformity in screen-PLL indicates disproportionate allocation of screening benefits. Age-related deviations in uniformity of screen-PLL indicate opportunities to improve equitability of screening benefits. Screening is best implemented in an age range of higher screen-PLL. One might conclude that for a given number of screens and with all other factors held constant, the guideline with the most uniform screen-PLL is the most efficient.

Conceptually, factors that determine screen-PLL include breast cancer incidence and mortality, general life expectancy, clinical symptoms, screen timing, and the sensitivity of mammography. In the model, mammographic resolution is assumed to be 0.2 cm and determines the magnitude of screen-PLL. Improved resolution would increase screen-PLL, while decreased resolution would lower it. However, mammographic resolution does not alter the shape of the screen-PLL curve, but simply shifts the screen-PLL curves up or down.

Screen-PLL accumulates with the onset and growth of cancers. This is due to the natural history of the disease and is not dependent upon the initiation of screening. Although the potential to reduce screen-PLL exists outside the age boundaries of screening guidelines, it is limited by the available screening technologies, the relatively small number of cases in young women, and the reduced life-expectancy in older women.

In an unscreened cohort, screen-PLL increases as age-specific breast cancer incidence increases, and plateaus from about age 45 to 65. This plateau is due to tumor discovery through clinical manifestations along with the offsetting effects of increasing tumor incidence and diminishing life expectancy. After age 65, shortened life expectancy from general mortality causes a rapid decrease in screen-PLL. Accelerating this trend is the fact that fewer elderly women with screen-preventable disease will die of their cancer.

In the screen-PLL curves for ACS, USPSTF, and triennial screening, there is a reduction in screen-PLL each time mammography is performed and an increase during the interim between screens, resulting in the peaks and valleys seen in Fig 2. Each time a screen is performed, tumors are detected, changing the screen-PLL for these individuals to zero and reducing the screen-PLL for the cohort.

There are limitations to this study. First, only women in the United States with average breast cancer risk are modeled. Thus, the results may vary for other populations. Additionally, as with any model, the results are dependent upon the underlying assumptions and model calibration. The existence of non-progressive tumors and occurrence of overdiagnosis are unpredictable for an individual. Estimates of the overdiagnosis level in a population vary greatly [26, 28–33]. In this study, an overdiagnosis rate of 30% is assumed. Regardless of the actual overdiagnosis value, only women at risk of death from breast cancer influence screen-PLL. By definition, women overdiagnosed by screening do not die from breast cancer and thus do not contribute to preventable life years lost to breast cancer. As a result, screen-PLL curves for a given screening strategy with different overdiagnosis rates generate superimposable screen-PLL curves as observed in validation testing.

In this study we have examined screen-PLL as a method for assessing breast cancer screening guidelines within the framework of a natural history model. Crucially, the study simply explores concepts and consequences of screen-PLL in delivering one benefit, preventable loss of life, to a population of aging women. Contrasts of screen-PLL curves among three major guidelines illustrate the principle of equity of health care delivery among young and aging women. Equity across ages in screening guidelines is not directly considered in guideline development, but we suggest that it should be. Perhaps this gap exists because a clinical decision metric such as screen-PLL has not been developed.

This study of screen-PLL has not considered monetary factors for a screening program and provides no cost-effectiveness implications. While screen-PLL makes inequities in a breast cancer screening strategy obvious, it does not evaluate the economics of screening strategies. However, the concept of screen-PLL could be applied in the context of cost-effectiveness in future studies. Such an analysis would need to consider the cost of therapies at various stages of disease as well as the costs and harms of screening. Additionally, the concept of screen-PLL may be used in future guideline development to offer additional life savings for the same number of screens.

Conclusions

This study broadens the concept of screen-PLL as a method to improve the understanding of breast cancer screening guidelines in distributing life savings across an aging population. By analyzing screen-PLL curves, the health benefits of mammography screening may be more equitably allocated.

Screen-PLL curve data.

Excel.xlsx workbook with monthly data used to create screen-PLL curves for Fig 2.

(XLSX)

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.

15 Oct 2020

PONE-D-20-25853

A Method for Evaluating Breast Cancer Screening Strategies Using Screen-Preventable Loss of Life

PLOS ONE

Dear Dr. Carter,

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 Nov 29 2020 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

We look forward to receiving your revised manuscript.

Kind regards,

Eugenio Paci, MD

Academic Editor

PLOS ONE

Additional Editor Comments:

I found the proposal of Screen-Preventable Loss of Life estimate very interesting and the tool potentially useful in order to assess the impact of different screening guidelines. The interesting example of USPSTF and ACS is a significant example. However, as the reviewer discussed, each guideline has many aspects to be considered and areas of uncertainties which must be considered.

I guess that the new version of the paper should answer to some the suggestions , because it is important if the new tool might be considered in the future comparison of screening guidelines. In my view, the authors should assess the impact of the evaluation not only in terms of gain in term of better impact. The question is relevant when two guidelines which are different in minimal aspects in terms of benefit are justified? A guideline could increase minimally the benefit but with increasing costs or harms or , which is also important, practical complexity for the woman and service.

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

https://journals.plos.org/plosone/s/file?id=wjVg/PLOSOne_formatting_sample_main_body.pdf and

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

2. We note that you have stated that you will provide repository information for your data at acceptance. Should your manuscript be accepted for publication, we will hold it until you provide the relevant accession numbers or DOIs necessary to access your data. If you wish to make changes to your Data Availability statement, please describe these changes in your cover letter and we will update your Data Availability statement to reflect the information you provide.

[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: Partly

**********

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

Reviewer #1: No

**********

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: No

**********

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

**********

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: PONE-D-20-25853, A Method for Evaluating Breast Cancer Screening Strategies Using Screen-Preventable Loss of Life

Line 72: The authors should completely describe the American Cancer Society guideline, which also encourages women to make an informed decision about beginning screening at ages 40-44 (but should begin no later than 45), while also having the choice to continue annual screening after age 55, with the same stopping criteria. I appreciate that you can’t practically model that. In the methods section you might note that guidelines are modeled according to the base recommendation and do not accommodate shared decision periods.

Line 89: By this definition of screen-PLL, wouldn’t a breast cancer death at any age be a premature death? Loss of life is the projected remaining years of life among the cohort that is still living and has not died from breast cancer, but is it estimated from the expected date of death if the cancer had not been detected by screening? The example is clear in the context of a 0.2 cm tumor (very small), but the methods should describe how the model works for other occult tumor sizes that are not metastatic, unless all screen detected cancers have these parameters. Consider the ACS guideline, which could be judged to be a little extreme with respect to 10 + year longevity as a criteria for quitting--any estimate of Screen-PLL < 10 years would, by the guideline definition, should not have invoked a referral to screening, and any longevity over 85 is off the table for the USPSTF. Personally, avoiding a premature death with some meaningful number of years less than 10 (allowing for the treatment and recovery period) would likely be judged to be worth it by the patient, but also obviously, we have numerous studies in the literature of women with severe life-limiting comorbidity, who should not have been referred to screening, who die from another cause within a 1-2 years of their diagnosis. So, the question, is there a threshold for a screen-PLL that is not worth pursuing because it may lead to more harm than benefit? Also, I’m guessing that the role of therapy is assumed to be stage-specific and fixed, so that the contribution of screening, and earlier detection, is fixed. That is worth mentioning if that is the case.

Line 92: The way screen-PLL is expressed could use a bit more explanation, I think. The authors state that screen-PLL increases to the difference between her expected non-breast cancer death age and her (expected?) breast cancer death age. In your example, the expected breast cancer death age is set at 12 years if she is detected with a 0.2 cm tumor. But, in this example, it seems that screen-PLL can only decrease as a function of age at diagnosis, specifically younger age at diagnosis, and to a lesser extent, tumor size (with screen-PLL also decreasing with < tumor size) within the detectable range, which runs up to metastasis. Does this model assume all symptomatic breast cancers are metastatic? Is metastatic here synonymous with distant disease? The logic of the PLL being a function of T4-T3 is obvious, but how T3 is estimated could be made clearer.

Line 124: A description of the basis for comparison between the guidelines seems to be missing. Does the model assume that the age to begin and end screening results in 100 percent adherence with screening and 100% screen detection of cancers? If so, please explain, and explain how stopping ages were modeled for the ACS guideline. Is it just modeled on a life-table?

Line 133: Is it reasonable to state that the PLL beings to accumulate as women increase in age, even though guidelines do not endorse screening for average risk women until 40? The authors might acknowledge that the true potential of reducing PLL is within the boundaries of the recommended screening protocols, before and after which we have to accept some breast cancer deaths and life years lost are unavoidable.

Line 177: The first limitation could be judged to be perfunctory and is not worth mentioning since any study done in a single country faces the same global limitation in true generalizability, although you could note that this is a function of the screening recommendations and the burden of disease. I’d explain why it is a limitation, and perhaps just note that the results will vary by population risk and the screening protocol. In some respects, the US is perhaps the largest heterogeneous population in the world. Second, population estimates of overdiagnosis don’t vary greatly because of population differences, they vary greatly due to 1) the limitations of the data to estimate overdiagnosis, 2) the wide range of methodologies applied, which have been judged to range from quite good to quite flawed. Here the authors have chosen a quite flawed estimate (30%), which is many times higher than the estimates from studies that have shown more sophisticated awareness of factors associated with differences in incidence over time in a population exposed to screening vs. not exposed to screening. re are several limitations that the authors have not identified. If the authors have used overdiagnosis rates estimated from SEER because they’re using incidence and mortality from SEER, the model does not require this. I suggest that the authors consider the work of Danish investigators on this issue….. (1) Lynge, E., Beau, A. B., Christiansen, P., von Euler-Chelpin, M., Kroman, N., Njor, S., & Vejborg, I. (2017). Overdiagnosis in breast cancer screening: The impact of study design and calculations. Eur J Cancer, 80, 26-29. doi:10.1016/j.ejca.2017.04.018; (2) Lynge, E., Beau, A. B., von Euler-Chelpin, M., Napolitano, G., Njor, S., Olsen, A. H., . . . Vejborg, I. (2020). Breast cancer mortality and overdiagnosis after implementation of population-based screening in Denmark. Breast Cancer Res Treat. doi:10.1007/s10549-020-05896-9; and (3) Njor, S. H., Paci, E., & Rebolj, M. (2018). As you like it: How the same data can support manifold views of overdiagnosis in breast cancer screening. Int J Cancer, 143(6), 1287-1294. doi:10.1002/ijc.31420. The authors could do a sensitivity analysis (0, 5, 10, 19, and 30)…these estimates are all in the literature.

There are some limitations the authors have not mentioned. At line 120, the authors state, “Age

specific incidence data for breast cancer among average-risk white women in the United States

are used to determine the age of onset of clinical disease.” The authors should acknowledge that only half, and maybe not that many breast cancers are screen detected, or preferred, detected among women attending screening in any give year (screen detected and interval cancers). So, the SEER data are a mix of high risk and average risk women, and a mix of screen-detected and clinically detected disease. This, and the implications for the model, should be acknowledged.

Lines 183-189: I don’t really see these as study limitations. Perhaps shift to another part of the discussion, or move to the methodology.

**********

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: No

[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.

1 Nov 2020

November 1, 2020

Authors’ responses to reviewer and editor comments:

Reviewer #1: PONE-D-20-25853, A Method for Evaluating Breast Cancer Screening Strategies Using Screen-Preventable Loss of Life

1. The authors should completely describe the American Cancer Society guideline, which also encourages women to make an informed decision about beginning screening at ages 40-44 (but should begin no later than 45), while also having the choice to continue annual screening after age 55, with the same stopping criteria. I appreciate that you can’t practically model that. In the methods section you might note that guidelines are modeled according to the base recommendation and do not accommodate shared decision periods.

Authors’ Response: We have added these details to the description of the ACS guideline in the Introduction. We have noted in the Methods that our model considers only the base recommendation.

2. By this definition of screen-PLL, wouldn’t a breast cancer death at any age be a premature death?

Authors’ Response: Yes, death from breast cancer is considered premature in our model, as stated in the Introduction. No additional changes to the manuscript have been made regarding this comment.

3. Loss of life is the projected remaining years of life among the cohort that is still living and has not died from breast cancer, but is it estimated from the expected date of death if the cancer had not been detected by screening?

Authors’ Response: Screen-PLL is evaluated for all women as the number of preventable months of life lost to breast cancer from progressive tumors that are mammographically detectable at an early stage and curable. It is calculated as the difference between her expected age of death from general mortality and her age of death from breast cancer. However, once the cancer is detected by screening or clinical symptoms, screen-PLL is zero, as further screening provides no additional benefit. The term preventable has been added to the Methods in the definition of screen-PLL for clarity.

4. The example is clear in the context of a 0.2 cm tumor (very small), but the methods should describe how the model works for other occult tumor sizes that are not metastatic, unless all screen detected cancers have these parameters.

Authors’ Response: Occult tumors less than 0.2 cm diameter are assumed in the model to be undetectable by mammography, therefore are not considered screen-preventable. Occult tumors greater than 0.2 cm but not metastatic are precisely the tumors that contribute to screen-PLL. No changes have been made to the manuscript regarding this comment.

5. Consider the ACS guideline, which could be judged to be a little extreme with respect to 10 + year longevity as a criteria for quitting--any estimate of Screen-PLL < 10 years would, by the guideline definition, should not have invoked a referral to screening, and any longevity over 85 is off the table for the USPSTF. Personally, avoiding a premature death with some meaningful number of years less than 10 (allowing for the treatment and recovery period) would likely be judged to be worth it by the patient, but also obviously, we have numerous studies in the literature of women with severe life-limiting comorbidity, who should not have been referred to screening, who die from another cause within a 1-2 years of their diagnosis. So, the question, is there a threshold for a screen-PLL that is not worth pursuing because it may lead to more harm than benefit?

Authors’ Response: The intent of the study was to compare screen-PLL curves for major screening guidelines. Developing an ideal strategy with respect to optimal start and stop ages as well as screening frequency is beyond the scope of this study. The concept of screen-PLL may be beneficial in helping develop future breast cancer screening guidelines. The judgment of what amount of screen-PLL is worth pursuing lies in the domain of cost-effectiveness. The Discussion was expanded to include some of these concepts.

6. Also, I’m guessing that the role of therapy is assumed to be stage-specific and fixed, so that the contribution of screening, and earlier detection, is fixed. That is worth mentioning if that is the case.

Authors’ Response: The role of stage-specific therapy is not considered. The model assumption is that metastatic disease is the cause of breast cancer death and that non-metastatic cancers are curable. No changes have been made to the manuscript regarding this comment.

7. The way screen-PLL is expressed could use a bit more explanation, I think. The authors state that screen-PLL increases to the difference between her expected non-breast cancer death age and her (expected?) breast cancer death age. In your example, the expected breast cancer death age is set at 12 years if she is detected with a 0.2 cm tumor. But, in this example, it seems that screen-PLL can only decrease as a function of age at diagnosis, specifically younger age at diagnosis, and to a lesser extent, tumor size (with screen-PLL also decreasing with < tumor size) within the detectable range, which runs up to metastasis. Does this model assume all symptomatic breast cancers are metastatic? Is metastatic here synonymous with distant disease? The logic of the PLL being a function of T4-T3 is obvious, but how T3 is estimated could be made clearer.

Authors’ Response: The model does not assume that all symptomatic breast cancers are metastatic. As discussed in the methods section, tumors may be detected by screening or with clinical symptoms. We define metastatic disease as distant disease, which arises in the model when the woman’s primary tumor reaches pre-determined volume thresholds. This has been clarified in Methods. The age of breast cancer death in the model (t3) for a given woman is based on her metastatic disease. Death occurs when her metastatic tumor load reaches a terminal level, which was explained in Methods. No further additions were made.

We appreciate the reviewer’s comments that screen-PLL decreases with advancing age as life-expectancy shortens and that younger women would tend to have a higher screen-PLL, given longer life-expectancies. These concepts were presented in Discussion and no additional explanations were made.

Screen-PLL is constant during the detectable tumor range and does not vary with primary tumor size. Screen-PLL is zero once metastatic disease occurs. The rationale behind screen-PLL has been expanded throughout the revised manuscript.

8. A description of the basis for comparison between the guidelines seems to be missing. Does the model assume that the age to begin and end screening results in 100 percent adherence with screening and 100% screen detection of cancers? If so, please explain, and explain how stopping ages were modeled for the ACS guideline. Is it just modeled on a life-table?

Authors’ Response: The following was added to Methods to address these items of guideline adherence, screen detection, and ACS stopping age modeling:

It is assumed that during the study, women are 100 percent compliant with their assigned screening strategy. Sensitivity of mammography increases with tumor diameter, starting at 0.2 cm, and is assumed higher for post-menopausal women. Stopping ages for the ACS guideline rule, 10-year or less life-expectancy, were calculated using life table data.

9. Is it reasonable to state that the PLL beings to accumulate as women increase in age, even though guidelines do not endorse screening for average risk women until 40? The authors might acknowledge that the true potential of reducing PLL is within the boundaries of the recommended screening protocols, before and after which we have to accept some breast cancer deaths and life years lost are unavoidable.

Authors’ Response: PLL does begin to accumulate even before the age at which screening is recommended based on the guidelines. The potential to reduce PLL exists outside the age boundaries of screening guidelines, however it is limited by the available screening technologies, the relatively small number of cases in young women, and the limited life-expectancy in older women. These notions have been added to the Discussion.

10. The first limitation could be judged to be perfunctory and is not worth mentioning since any study done in a single country faces the same global limitation in true generalizability, although you could note that this is a function of the screening recommendations and the burden of disease. I’d explain why it is a limitation, and perhaps just note that the results will vary by population risk and the screening protocol. In some respects, the US is perhaps the largest heterogeneous population in the world.

Authors’ Response: We agree with the reviewer that this is a relatively minor limitation which is common to similar studies performed within a single country. We have modified the text to reflect that results may vary with different populations.

11. Second, population estimates of overdiagnosis don’t vary greatly because of population differences, they vary greatly due to 1) the limitations of the data to estimate overdiagnosis, 2) the wide range of methodologies applied, which have been judged to range from quite good to quite flawed. Here the authors have chosen a quite flawed estimate (30%), which is many times higher than the estimates from studies that have shown more sophisticated awareness of factors associated with differences in incidence over time in a population exposed to screening vs. not exposed to screening. re are several limitations that the authors have not identified. If the authors have used overdiagnosis rates estimated from SEER because they’re using incidence and mortality from SEER, the model does not require this. I suggest that the authors consider the work of Danish investigators on this issue….. (1) Lynge, E., Beau, A. B., Christiansen, P., von Euler-Chelpin, M., Kroman, N., Njor, S., & Vejborg, I. (2017). Overdiagnosis in breast cancer screening: The impact of study design and calculations. Eur J Cancer, 80, 26-29. doi:10.1016/j.ejca.2017.04.018; (2) Lynge, E., Beau, A. B., von Euler-Chelpin, M., Napolitano, G., Njor, S., Olsen, A. H., . . . Vejborg, I. (2020). Breast cancer mortality and overdiagnosis after implementation of population-based screening in Denmark. Breast Cancer Res Treat. doi:10.1007/s10549-020-05896-9; and (3) Njor, S. H., Paci, E., & Rebolj, M. (2018). As you like it: How the same data can support manifold views of overdiagnosis in breast cancer screening. Int J Cancer, 143(6), 1287-1294. doi:10.1002/ijc.31420. The authors could do a sensitivity analysis (0, 5, 10, 19, and 30)…these estimates are all in the literature.

Authors’ Response: The authors are aware of the reviewer’s concerns regarding the vagaries of breast cancer overdiagnosis. As noted, the model was standardized on SEER incidence and mortality data. A published study for overdiagnosis of SEER data estimated overdiagnosis at 30 percent.

Importantly, only women at risk of death from breast cancer influence screen-PLL. By the very definition of overdiagnosis, women overdiagnosed by screening do not die from breast cancer and thus do not contribute to preventable life years lost to breast cancer. As a result, screen-PLL curves for a given screening strategy with different overdiagnosis rates generate superimposable screen-PLL curves. This was confirmed with internal validation testing.

The concept of breast cancer overdiagnosis is interesting and contentious. Many reasons contribute to the inconsistencies of overdiagnosis observations/measurements as observed in the articles referenced by the reviewer. The dynamics of overdiagnosis are not the intent of this study, and overdiagnosis does not influence the characteristics of screen-PLL.

The Discussion was modified to include these concepts.

12. There are some limitations the authors have not mentioned. At line 120, the authors state, “Age

specific incidence data for breast cancer among average-risk white women in the United States

are used to determine the age of onset of clinical disease.” The authors should acknowledge that only half, and maybe not that many breast cancers are screen detected, or preferred, detected among women attending screening in any give year (screen detected and interval cancers). So, the SEER data are a mix of high risk and average risk women, and a mix of screen-detected and clinically detected disease. This, and the implications for the model, should be acknowledged.

Authors’ Response: We agree that incidence data reflects a mix of screen detected and interval cancers. The population that we have modeled includes both groups of women. No changes have been made to the manuscript regarding this comment.

13. I don’t really see these as study limitations. Perhaps shift to another part of the discussion, or move to the methodology.

Authors’ Response: This concept has been moved out of limitations and to another part of the discussion.

Additional Editor Comments:

1. I guess that the new version of the paper should answer to some the suggestions, because it is important if the new tool might be considered in the future comparison of screening guidelines. In my view, the authors should assess the impact of the evaluation not only in terms of gain in term of better impact. The question is relevant when two guidelines which are different in minimal aspects in terms of benefit are justified? A guideline could increase minimally the benefit but with increasing costs or harms or, which is also important, practical complexity for the woman and service.

Authors’ Response: The authors have examined screen-PLL as a method for assessing breast cancer screening guidelines within the framework of a natural history model. Crucially, the study simply explores concepts and consequences of screen-PLL in delivering one outcome, preventable loss of life, to a population of aging women. Contrasts of screen-PLL curves among three major guidelines illustrate the notion of equity of health care delivery among young and aging women. Equity across ages in screening guidelines is not a notion considered directly in guideline development, but we suggest that it should be. Perhaps this gap exists because a clinical decision tool handling these screen-PLL concepts is not available.

This study of screen-PLL has not considered cost factors for a screening program such as the cost of screening and subsequent therapy and provides no cost-effectiveness implications. While screen-PLL makes inequities in a breast cancer screening strategy obvious, it does not evaluate the economics of breast cancer screening strategies. The judgment of what amount of screen-PLL is worth pursuing lies in the domain of cost-effectiveness. The notion of screen-PLL could be applied in the context of cost-effectiveness in future studies. A cost-effectiveness screen-PLL analysis would need to consider the cost of therapies at various stages of disease as well as the costs and harms of screening.

The Discussion was modified to include these concepts.

2. We note that you have stated that you will provide repository information for your data at acceptance. Should your manuscript be accepted for publication, we will hold it until you provide the relevant accession numbers or DOIs necessary to access your data. If you wish to make changes to your Data Availability statement, please describe these changes in your cover letter and we will update your Data Availability statement to reflect the information you provide.

Authors’ Response: Relevant data is submitted with the manuscript revision.

16 Nov 2020

A method for evaluating breast cancer screening strategies using screen-preventable loss of life

PONE-D-20-25853R1

Dear Dr. Carter,

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,

Eugenio Paci, MD

Academic Editor

PLOS ONE

Additional Editor Comments (optional):

This is a very interesting and important tool in order to compare different screening policies.

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 #1: 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 #1: Yes

**********

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

Reviewer #1: 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 #1: 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 #1: 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 #1: (No Response)

**********

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 #1: Yes: Robert A. Smith

20 Nov 2020

PONE-D-20-25853R1

A method for evaluating breast cancer screening strategies using screen-preventable loss of life

Dear Dr. Carter:

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. Eugenio Paci

Academic Editor

PLOS ONE