The impact of the COVID-19 pandemic on breast cancer screening, treatment, and overall survival.

Introduction
During the COVID-19 pandemic, the health care system faced many unprecedented challenges. These challenges necessitated changes in clinical protocols and patient workflows to help protect patients and health care staff from exposure to COVID-19 and ensure that care could be provided to those that were affected by COVID-19. The cancer care system also implemented changes including the greater use of virtual visits, the temporary suspension or reduction of cancer screening and diagnostic services, triaging patients for treatment based on acuity, and the redeployment of health care staff to process COVID-19 tests and care for acute cases of COVID-19.
One topic that caused anxiety among health care providers, patients, and the public was the potential negative impact of COVID-19 on breast cancer [1]. Studies estimated that delays in diagnosing and treating breast cancer would lead to increases in the number of advanced cancers and decreases in survival [2–4]. As a result, guidelines for managing the care of individuals with breast cancer during the pandemic were introduced that included the increased use of neoadjuvant systemic therapy, oral formulations, and hypofractionation [5, 6]. The implementation of many of these changes had already begun based on evidence from clinical trials. Since the start of the pandemic, studies from several jurisdictions have examined pandemic-initiated changes in breast cancer treatment and their impact on outcomes [7]. Most of these studies are small, not population-based, took place in a single center, were based on self-reported data from surveys, or included only the first year of the pandemic. The objective of this study was to use population-based data to examine the association between the COVID-19 pandemic and breast cancer screening, treatment, and overall survival for individuals diagnosed with breast cancer to December 2022 in Manitoba, Canada.

Methods

Setting
Manitoba is a province located in central Canada with a population of approximately 1.35 million. Two-thirds of the population live in the capital city of Winnipeg. CancerCare Manitoba (CCMB) is the provincial cancer agency that provides cancer treatment services to all Manitobans diagnosed with cancer. CancerCare Manitoba also manages BreastCheck, the population-based breast cancer screening program. BreastCheck invites eligible women for a bilateral, two-view mammogram every 2 years at four fixed sites or mobile mammography units that visit approximately 45 rural, northern, and First Nations communities each year.
Manitoba reported its first COVID-19 case on March 12th, 2020. A state of emergency was declared on March 20th, 2020, and the first lockdown was implemented on April 1st, 2020. By the end of March 2020, patients presenting to CCMB locations were screened for COVID-19 symptoms, travel history, and COVID-19 exposure. When possible, visits shifted from in-person to telephone or virtual appointments. Neoadjuvant chemotherapy strategies were increased for some cancer sites, including breast cancer. Hypo-fractionated and single fraction radiation treatments were recommended to reduce resource radiotherapy consumption and minimize attendance at CCMB [8]. For early-stage estrogen receptor positive breast cancers, adjuvant hormonal therapy alone was employed in elderly patients as a substitute for adjuvant radiotherapy based on evidence from randomized controlled trials [9]. Elective surgeries were postponed and radical cancer surgeries were prioritized within disease sites or delayed at the discretion of treating physicians due to limitations in operating room availability and staffing [10]. In March 2020, BreastCheck operations were suspended as screening program staff were re-deployed to support COVID-19 protocols such as patient and staff entry screening at CCMB [11]. BreastCheck resumed limited operation at its primary location in June 2020 followed by other locations and the mobile screening units in July and August.

Study design and population
We performed a population-based, quasi-experimental cross-sectional study with an interrupted time series (ITS) analysis [12] to examine the number of breast cancer screening mammograms, first treatment rates, and 2-year survival prior to COVID-19 (January 2015 to March 2020) and after the start of the COVID-19 pandemic (April 2020 to December 2022).

Data sources
We used data from the Manitoba Cancer Registry to identify individuals diagnosed with invasive breast cancer (International Classification of Diseases for Oncology 3rd Edition (ICD-O) code C50), stage at diagnosis (American Joint Committee on Cancer (AJCC) TNM staging 7th edition before 2018 and 8th edition from 2018 onward), and type of treatment (Canadian Classification of Health Interventions codes for surgery, radiotherapy (RT), chemotherapy, targeted therapy, and hormone therapy). The Manitoba Cancer Registry is a population-based, high-quality registry that is legally mandated to collect and maintain accurate, comprehensive information about all individuals diagnosed with cancer in Manitoba [13]. Data from Manitoba’s population-based organized breast screening program were used to determine the number of screening mammograms that occurred for women 50 to 74 years of age.

Outcomes
The number of screening mammograms was examined from January 1, 2015, to December 31, 2022. We estimated the cumulative difference and percent cumulative difference between the fitted and counterfactual number of screening mammograms as of December 31, 2022. The estimated cumulative difference in screening mammograms was defined as the difference between the cumulative counterfactual count (the estimated number of mammograms in the absence of COVID-19) and the cumulative fitted count (i.e., smoothed estimates derived from observed data). The estimated percent cumulative difference in the number of screening mammograms was defined as the cumulative difference in the fitted count divided by the cumulative difference in the counterfactual count.
For all individuals diagnosed with stage I–III breast cancer, we examined the proportion of individuals who had surgery and RT and the rate of RT fractions per person-years. For those with stage I–III ER+/HER2- breast cancer, we also examined the proportion of individuals who had chemotherapy and hormone therapy. For those with stage I–III HER2+ breast cancer, we examined the proportion of individuals who had neoadjuvant and adjuvant chemotherapy, neoadjuvant and adjuvant targeted therapy (trastuzumab), and hormone therapy. For those with stage I–III triple-negative breast cancer, we examined the proportion of individuals who had neoadjuvant and adjuvant chemotherapy. For individuals diagnosed with stage IV breast cancer, we examined the proportion of individuals who had chemotherapy, RT, targeted therapy, and hormone therapy. Finally, we examined 2-year overall survival for individuals diagnosed with stage I–III and stage IV breast cancer.

Statistical analyses
We used an ITS analysis which accounts for baseline trends to investigate the association between the COVID-19 period and changes in screening mammogram counts, 1-year treatment proportions post-diagnosis, and 2-year survival post-diagnosis. This approach compared the COVID-19 period (April 2020 to December 2022) outcomes to counterfactual estimates as if the pandemic had not occurred based on pre-COVID-19 trends (i.e., baseline trends). This contrasts with pre-post designs which can overestimate or underestimate the association of the COVID-19 period by ignoring baseline trends.
A linear model was used to analyze monthly screening mammography counts. Logistic regression was used to analyze quarterly percentages of 1-year treatment post-diagnosis. Royston-Parmar models were used to analyze survival data [14]. For the survival analyses, individuals diagnosed prior to March 15, 2020 were censored on that date to avoid pre-pandemic cancer cases having follow-up during the COVID-19 period. Survival analyses for stages I–III were analyzed quarterly. Due to small event numbers, survival analyses for stage IV were analyzed by 6-month intervals. The month of March 2020, the quarter of January–March 2020, and the 6-month period of January–June 2020 were removed from the screening mammogram, treatment, and survival analyses, respectively, because COVID-19 restrictions were implemented throughout March 2020.
Each model included a binary intervention indicating the COVID-19 period status (i.e., zero during the pre-COVID-19 and one during the COVID-19 periods). A time variable indicating the time since the start of the study period was included to account for baseline trends. Seasonal effects were accounted for using a spline function or categorical variable. Fitted values (i.e., smoothed estimates from observed values) were generated by analyzing observed data. Counterfactual values were generated by estimating values with the binary indicator of COVID-19 as zero rather than one. COVID-19-by-time interactions were considered if the plotted fitted values in the COVID-19 period did not fit the observed data well. If linear or non-linear time interactions did not fit the observed data well, multiple COVID-19 dummy variables representing different periods during the pandemic were included. Kaplan–Meier and restricted mean survival time were calculated for survival analyses [15]. Delta restricted mean survival time at 2-year follow-up between fitted and counterfactual estimates represents the mean survival time lost or gained during the first 2 years of follow-up post-diagnosis during the COVID-19 period.
Scaled quantile residuals were used to evaluate model fit for linear and logistic models [16]. Contrast statements or bootstrapping with 1,000 replications were used to generate 95% confidence intervals (CI) included in the plots. Ratios between counterfactual and fitted estimates and 95% CI derived from contrast statements were reported. The estimated cumulative difference in the number of screening mammograms during the pandemic was calculated as of December 31, 2022. The 95% CIs for the cumulative difference estimates were calculated with bootstrapping and 1,000 replications.
Data analyses were performed in SAS version 9.4 (SAS Institute Inc., Cary, NC, USA) and R version 4.4.1 (R Foundation for Statistical Computing, Vienna, Austria). The following R packages were used: haven, splines, Hmisc, lattice, ggplot2, car, DHARMa, multcomp, lmtest, and rstpm2 [16–24]. This study was approved by the University of Manitoba’s Health Research Ethics Board (HS23979; H2020:264), Manitoba Health’s Provincial Health Research Privacy Committee (2020/2021–16), and CCMB’s Research and Resource Impact Committee (2020–14). Informed consent was not required. This study follows STROBE reporting guidelines for observational studies.

Results

Breast cancer screening
A total of 341,646 screening mammograms were provided in Manitoba from January 2015 to December 2022. When BreastCheck resumed limited operations at one location in June 2020 after ceasing operations for 2 months, there was a 52% decrease in the number of screening mammograms (ratio = 0.48, 95% CI 0.32, 0.65) compared to the counterfactual (Fig. 1, Supplemental Table 1). By December 31, 2022, there was a 27% decrease in the number of screening mammograms (ratio = 0.73, 95% CI 0.58, 0.89), with an estimated cumulative deficit (i.e., the estimated backlog) of 27,452 screening mammograms (20.5% fewer screening mammograms, 95% CI 11.8%, 28.3%) (Supplemental Table 2).

Breast cancer treatment
A total of 7,102 individuals were diagnosed with invasive breast cancer in Manitoba from 2015 to 2022. Most were 50 to 79 years of age (71.6%) and were diagnosed at stage I (60.7%) (Table 1). Of the individuals who were stage I–III, 65.4% were diagnosed with ER + /HER2- breast cancer. The number of individuals diagnosed with breast cancer decreased by 6.2% in 2020 compared to 2019 and then increased by 11.4% in 2021 and 2022 compared to 2020.

Stage I–III breast cancer
Fewer individuals diagnosed with stage I–III breast cancer from April to June 2020 had surgery when comparing the COVID-19 period to the counterfactual, but the difference was not significant (ratio = 0.59 (0.33, 1.07) (Fig. 2A; supplemental Table 3). There was also no significant difference in the proportion of individuals diagnosed from July 2020 to December 2022 who had surgery. Significantly fewer individuals diagnosed from April 2020 to December 2022 had RT (ratio = 0.69, 95% CI 0.53, 0.91) compared to the counterfactual (Fig. 2B). In addition, the number of RT factions per person-year was significantly lower for those diagnosed from April to June 2020 (ratio = 0.81, 95% CI 0.71, 0.92) and July 2020 to December 2022 (ratio = 0.60, 95% CI 0.56, 0.66) compared to the counterfactual (Fig. 2C).

Stage I–III ER+/HER2- breast cancer
Figure 3 shows the proportion of individuals diagnosed with stage I–III ER+/HER2- breast cancer who had chemotherapy and hormone therapy. There was no significant difference in the proportion of individuals who had chemotherapy (Fig. 3A, supplemental Table 4). However, the proportion of individuals who had hormone therapy was significantly lower for individuals diagnosed from April 2020 to December 2022 compared to the counterfactual (ratio = 0.55, 95% CI 0.36, 0.84) (Fig. 3B).

Stage I–III HER2+ breast cancer
The proportion of individuals diagnosed with stage I–III HER2+ breast cancer from April 2020 to December 2021 who had neoadjuvant chemotherapy was non-significantly higher than the counterfactual (Fig. 4A, supplemental Table 4). For those diagnosed from January to December 2022, the proportion who had neoadjuvant chemotherapy was significantly higher (ratio = 3.64, 95% CI 1.18, 11.16). The proportion of individuals who had adjuvant chemotherapy was significantly lower for those diagnosed from April to June 2020 (ratio = 0.14, 95% CI 0.04, 0.47) and from January to December 2022 (ratio = 0.19, 95% CI 0.08, 0.48) (Fig. 4B). The proportion of individuals who had neoadjuvant target therapy was significantly higher for those diagnosed from April to June 2020 (ratio = 3.34, 95% CI 1.13, 9.84) (Fig. 4C). The proportion of those who had adjuvant targeted therapy was significantly lower for individuals diagnosed from April to June 2020 (ratio = 0.14, 95% CI 0.04, 0.45) and January to December 2022 (ratio = 0.18, 95% CI 0.07, 0.45) and non-significantly lower from July 2020 to December 2021 (Fig. 4D). The proportion of individuals who had hormone therapy was significantly lower for those diagnosed from April 2020 to December 2022 (ratio = 0.43, 95% CI 0.22, 0.83) (Fig. 4E).

Stage I–III triple-negative breast cancer
The proportion of individuals diagnosed with stage I–III triple-negative breast cancer from April to September 2020 who had neoadjuvant chemotherapy was significantly higher than the counterfactual (ratio = 2.92, 95% CI 1.04, 8.17) (Fig. 5A, Supplemental Table 4). There was no significant difference in the proportion of individuals who had adjuvant chemotherapy (Fig. 5B).

Stage IV breast cancer
The proportion of individuals diagnosed with stage IV breast cancer from April 2020 to December 2022 who had surgery, chemotherapy, RT, targeted therapy, and hormone therapy was not significantly different than the counterfactual (Fig. 6A–E; Supplemental Table 5).

Overall survival

Stage I–III breast cancer
In unadjusted analyses, a non-significant decrease in 2-year overall survival for individuals diagnosed with stage I–III breast cancer was observed from April to June 2020 (all ages delta RMST -7.1 days, 95% CI -24.2, 8.4; 50 to 74 years of age delta RMST -13.0 days, 95% CI -33.6, 5.3); 75 + years of age delta RMST -8.8 days, 95% CI -59.6, 27.6) (Fig. 7A–C; supplemental Table 6). This decrease was reduced after adjusting for age and stage. There was no difference in 2-year overall survival compared to the counterfactual from July 2020 to December 2022.

Stage IV breast cancer
In unadjusted analyses, a non-significant decrease in 2-year overall survival for individuals of all ages diagnosed with stage IV breast cancer was observed from July 2020 to December 2022 (Fig. 8; supplemental Table 7). The decrease was larger after adjusting for age and stage but still not significant.

Discussion
This study examined the impact of the COVID-19 pandemic on screening, treatment, and overall survival for individuals diagnosed with breast cancer. We found that the changes made to the health care system in response to the pandemic did not have a deleterious effect on breast cancer patients in Manitoba. This highlights that the judicious use of available evidence during lockdowns was effective at rationing scarce surgical and clinical resources while effectively managing breast cancer risks to mortality. We were also able to use pandemic-era constraints to change clinical pathways that aligned with ongoing changes in the literature such as shifts toward more neoadjuvant treatments and delivering larger doses of radiation per treatment over fewer days. Remarkably, the COVID-19 pandemic helped to accelerate the implementation of updated clinical trial-supported evidence regarding breast cancer treatment.
Like other Canadian provinces and countries across the world, breast screening in Manitoba was suspended during the first few months of the pandemic. Screening rates then quickly reached pre-pandemic numbers but began to decrease in January 2022 during the peak of the fifth wave of the pandemic. Consequently, the deficit in the number of screening mammograms has continued to grow. The shortage that has accrued since the beginning of 2022 is influenced by mammography technologist shortages. According to the Canadian Association of Medical Radiation Technologists, the vacancy rate for mammography technologists was 12% in 2023 compared to an overall health care industry vacancy rate of 5.8% and an average labor force vacancy rate of 3.4% [25]. As the population ages and the age at which someone is eligible for breast screening in Manitoba is lowered to include individuals in their 40 s [26], the demand for breast mammogram screening will increase. This may exacerbate the existing deficit unless mammography technologist staffing shortages are addressed. This situation is not unique to Manitoba; mammography technologist shortages have been reported across Canada as well as in other countries including the United States and the United Kingdom [25, 27, 28].
Many jurisdictions reported a reduction in the number of breast cancer surgeries during the first wave of the pandemic [29–31]. The decrease in breast cancer surgery from April to June 2020 in Manitoba reflects changes in clinical decision-making in which breast cancer patients saw their oncologist before their surgeon supporting the redirection of resources to address hospital-based human resource needs associated with COVID-19 patients. The decrease was brief; by July 2020, breast cancer surgery rates were no longer below pre-pandemic levels. The proportion of individuals diagnosed with breast cancer who had RT has increased over time but remains lower than the pre-pandemic trend, reflecting the permanent adoption of the omission of RT among elderly patients with ER+ breast cancer who can be safely managed with hormonal therapy alone [32]. The decrease in the number of RT fractions was due to the rapid implementation of hypofractionation regimens for breast cancer in accordance with the publication of clinical trial evidence showing the non-inferiority of five-fraction adjuvant radiotherapy when compared to longer fractionation schedules commonly in use prior to 2020 (i.e., the UK FAST-Forward trial) leading to updated guidelines and recommendations [33, 34]. Typically, the adoption of clinical trial results into routine clinical practice (known as the research to evidence-based practice time gap) is 15 years [35]. However, given the pressures associated with the pandemic, the implementation of the UK Fast-Forward hypofractionated schedules into routine clinical practice in Manitoba took approximately six weeks to implement post-publication. This is likely the fastest time to implement any novel radiation technique in the history of CancerCare Manitoba.
Rates of neoadjuvant chemotherapy and targeted therapy for individuals diagnosed with stage I–III HER2 + and triple-negative breast cancer began to increase before the pandemic based on the results of clinical trials (i.e., KATHERINE and CREATE-X trials) that found a survival benefit when systemic therapy is given neoadjuvantly, with changing or adding systemic therapy post-operatively if a complete response is not obtained [36–38]. The use of neoadjuvant systemic therapy then increased more rapidly after the pandemic because of the treatment protocols aimed at postponing surgery [7, 32]. In Manitoba, the adoption of KATHERINE trial results into clinical practice required multidisciplinary collaboration and the revision of patient workup requirements which took time. For example, work with pathology to introduce HER2 reporting on biopsies was required as well as the increased use of axillary ultrasound at time of diagnosis. The increase in neoadjuvant targeted therapy in 2021 observed in this study is the result of this complex health system process change and education. This contrasts with the ease of radiation hypofractionation uptake discussed above, as that protocol was limited to a single specialty and logistically easier to implement. Neoadjuvant chemotherapy and targeted therapy are now considered the standard of care for HER2+ and triple-negative breast cancer including individuals with stage II and III operable disease [38].
A high proportion of individuals with stage I–III ER+/HER2− breast cancer received hormone therapy before and after the pandemic. However, the proportion of individuals who had hormone therapy was significantly lower after the pandemic compared to the counterfactual. To examine if the decrease was due to delays in the start of hormone therapy, we extended the study’s follow-up from 1 to 2 years. We found no increase in the number of individuals who started hormone therapy more than 1 year after diagnosis. Hence, the decrease in hormone therapy is not because of delays in starting treatment. There were no mandatory or prescriptive initiatives by the breast disease site group at CancerCare Manitoba during the pandemic aimed at reducing hormonal therapy use. However, two clinical trends are hypothesized to have contributed to the observed drop in hormonal therapy. First, neoadjuvant hormonal therapy was used as a temporizing measure early in the pandemic by patients with ER+ breast cancer who required breast surgery, but due to pandemic-constrained operating room resources, could not get surgery in a timely manner. This practice was discontinued after operating room resources became more available as the pandemic progressed. Second, adjuvant hormonal therapy was used in lieu of adjuvant RT for ER+ breast cancer patients early in the pandemic prior to the implementation of UK Fast-Forward 5 fraction whole breast RT. This practice was in keeping with the findings of two previously published clinical trials (PRIME II and CALGB 9343) [9, 39, 40]. Now that the UK Fast-Forward has been widely implemented, some patients with an ER+ breast cancer with a low risk of recurrence opt for five fractions of adjuvant RT rather than undergo 5 years of adjuvant hormonal therapy. These trends may be responsible, in part, for the drop in hormonal therapy use in non-metastatic patients as observed in our data.
Delays in breast cancer treatment can significantly increase the risk of death [41]. For this reason, at the start of the pandemic, several studies simulated the impact of COVID-19 on cancer mortality and survival with results ranging from a small impact on breast cancer mortality to a substantial excess in mortality [3, 4, 42–46]. We did not find significant decreases in 2-year overall survival for stage I–III or stage IV breast cancer using real-world data for any age group. While this is reassuring, because the number of events is small, analysis with longer follow-up (i.e., 5- and 10-year survival) must also be completed.
The results of this study should be interpreted in the context of its strengths and limitations. For this study, we used a quasi-experimental study design with an ITS analysis that included a long pre-intervention period. This permitted the evaluation of outcomes before the start of the COVID-19 pandemic as well as the inclusion of seasonality and interactions between COVID-19 and time in the analysis. We also had timely access to population-based, system-level cancer data and were able to include data to December 2022 which encompasses multiple COVID-19 pandemic waves. However, because the association between COVID-19 and breast cancer screening, treatment, and survival may differ between jurisdictions, our results must be interpreted within the Manitoba context. Population-based studies should be carried out elsewhere, including pre-COVID-19 trends and sufficient follow-up time.

Conclusions
In Manitoba, Canada, there was a significant decrease in the number of screening mammograms 2 years after the start of the pandemic. Breast cancer treatment changed with a short-term reduction in surgery, a decrease in the number of RT fractions, and an increase in the use of neoadjuvant systemic therapy in accordance with evidence from clinical trials and updated treatment guidelines. Despite these changes to screening and treatment, we found no impact on 2-year overall survival.

Supplementary Information
Below is the link to the electronic supplementary material.