Current Trends in the Utilization of Preoperative Breast Magnetic Resonance Imaging Among Women With Newly Diagnosed Breast Cancer

PURPOSE

The clinical benefit of preoperative breast magnetic resonance imaging (MRI) for early-stage breast cancer (BC) remains controversial. We examined trends and the associated factors of preoperative breast MRI use.

METHODS

This study cohort, constructed from Optum Clinformatics database, included women with early-stage BC who had a cancer surgery between March 1, 2008, and December 31, 2020. Preoperative breast MRI was performed between the date of BC diagnosis and index surgery. Multivariable logistic regressions, one for elderly (65 years and older) and the other for non-elderly patients (younger than 65 years), were performed to examine factors associated with the use of preoperative MRI.

RESULTS

Among 92,077 women with early-stage BC, the crude rate of preoperative breast MRI increased from 48% in 2008 to 60% in 2020 for nonelderly and from 27% to 34% for elderly women. For both age groups, non-Hispanic Blacks were less likely (odds ratio [OR]; 95% CI, younger than 65 years: 0.75, 0.70 to 0.81; 65 years and older: 0.77, 0.72 to 0.83) to receive preoperative MRI than non-Hispanic White patients. Across Census divisions, the highest adjusted rate was observed in Mountain division (OR compared with New England; 95% CI, younger than 65 years: 1.45, 1.27 to 1.65; 65 years and older: 2.42, 2.16 to 2.72). Other factors included younger age, fewer comorbidities, family history of BC, axillary node involvement, and neoadjuvant chemotherapy for both age groups.

CONCLUSION

The use of preoperative breast MRI has steadily increased. Aside from clinical factors, age, race/ethnicity, and geographic location were associated with preoperative MRI use. This information is important for future implementation or deimplementation strategies of preoperative MRI.

Use of preoperative breast MRI increased 25% from 2008 to 2020, despite inconclusive clinical benefit.

INTRODUCTION

The clinical benefit of preoperative breast magnetic resonance imaging (MRI) for early-stage breast cancer (BC) remains controversial. Although the sensitivity of contrast-enhanced breast MRI is superior to that of mammography and ultrasound in detecting other sites of BC in either breast, its specificity is low to moderate and may lead to false-positive results.¹ Studies examining the association between receipt of preoperative breast MRI with more extensive surgery (mastectomy), contralateral prophylactic mastectomy, or reconstruction have been inconclusive.^2-8 Furthermore, preoperative breast MRI had no effect on the rate of re-excision and reoperation as well as locoregional recurrence and survival.^9-12 The current American Society of Breast Surgeons (ASBrS) consensus guidelines do not recommend routine use of breast MRI for local staging in patients with newly diagnosed BC.¹³

CONTEXT

• Key Objective

• Given continuing debate over the clinical benefit of preoperative breast magnetic resonance imaging (MRI), what is the contemporary trend of its use among adult women with early-stage breast cancer (BC) and factors associated with its use?

• Knowledge Generated

• Use of preoperative breast MRI increased from 48% in 2008 to 60% in 2020 among nonelderly women with private insurance and from 27% to 34% for elderly women enrolled in Medicare Advantage. Aside from clinical factors such as family history and receipt of neoadjuvant chemotherapy, variations were observed by race/ethnicity (lowest use among non-Hispanic Blacks) and geographic regions (highest use in Mountain division).

• Relevance

• Growing use of preoperative breast MRI will increase costs of BC care for patients and health care system, calling for more research to identify patients who will truly benefit from this practice to guide interventions that promote appropriate use and disincentivize inappropriate use of this expensive technology.

Previous studies have reported an increasing trend of preoperative breast MRI use among older women in the SEER registry regions (0.8% in 2000, 21% in 2007, and 25.2% in 2009).^3,14-17 Among women between age 25 and 64 years, one study using the TRICARE database reported that the percentage of patients who received MRI after mammogram or ultrasound increased between 2006 and 2015.¹⁸ Aside from that study, no studies have examined the contemporary trend of preoperative breast MRI use and the associated factors.

Using contemporary data, we examined preoperative breast MRI use among nonelderly and elderly women with newly diagnosed BC to document the utilization pattern of preoperative breast MRI over time and investigate factors associated with preoperative breast MRI use and sources of variations.

METHODS

Data Source

We used Optum deidentified Clinformatics Data Mart from 2007 to 2020 to construct the study cohort. The database, comprising enrollees in commercial health plans and Medicare Advantage, includes enrollment records, medical claims, and pharmacy claims for members of a large national managed care company affiliated with Optum. The population coverage is geographically diverse, spanning all 50 states. The University of Texas MD Anderson Cancer Center's Institutional Review Board determined that this study was exempt from full review as it involved deidentified data.

Study Cohort

We used current procedural terminology and international classification of disease 9 and 10 diagnosis/procedure codes to identify the study cohort. The cohort ascertainment algorithm began with women with early-stage BC^19,20 who had an index surgery (breast-conserving surgery or mastectomy) between March 1, 2008, and December 31, 2020, and met the following inclusion/exclusion criteria: (1) full insurance enrollment for at least 14 months before the index surgery, (2) no preoperative radiation therapy, (3) no neoadjuvant chemotherapy more than 8 months before the index surgery, (4) date of BC diagnosis within 3 months of initial treatment (neoadjuvant chemotherapy or surgery), (5) no history of BC, and (6) no distant metastasis. In addition, we excluded patients with missing values of Census division and those younger than 20 years.

Outcomes

Preoperative breast MRI was defined as breast MRI between the date of BC diagnosis and the index surgery.

Other Key Variables

Other variables included age, year of index surgery, race/ethnicity (non-Hispanic White, non-Hispanic Black, Hispanic, Asian, and not specified patients), Census divisions, insurance type (health maintenance organization [HMO] v non-HMO), family history of BC (yes/no), genetic susceptibility to BC (yes/no), axillary node involvement (yes/no), and receipt of neoadjuvant chemotherapy (yes/no). Comorbidity score was calculated using the algorithm proposed by Klabunde et al^21,22 and categorized as 0 (healthiest), 1, and ≥ 2. The codes and the corresponding time period of data extraction are available on request.

Statistical Analysis

Because of considerable variation in the proportion of surgically treated patients with preoperative breast MRI between elderly and nonelderly patients, we performed separate analyses for each. We used the Cochrane-Armitage test²³ to compare the trend in the rate of unadjusted preoperative MRI from 2008 to 2020. The chi-squared test was used to examine the bivariate associations between preoperative breast MRI use and each covariate. Multivariable logistic regression models were used to examine factors associated with the utilization of preoperative breast MRI. Covariates with a P of < .2²⁴ in bivariate analysis were included in the regression model.

We conducted statistical analyses using SAS version 9.4 (SAS Institute, Cary, NC) and STATA 15.0 (StataCorp LLC, College Station, TX). All statistical tests were two-sided, and the statistical significance level was .05.

Sensitivity Analysis

The first sensitivity analysis split the study duration into three time periods, 2008-2011, 2012-2015, and 2016-2020, and conducted separate multivariable logistic regression for each to better understand the diffusion pattern of preoperative MRI, especially the characteristics of early versus late adopters. Another sensitivity analysis excluded patients who underwent surgery in year 2020 to remove the potential influence of COVID-19 pandemic on the use of preoperative breast MRI.

RESULTS

Study Cohort

The study cohort included 92,077 women, of whom 41% underwent preoperative breast MRI. For women younger than 65 years (40,237 patients, 43.7%), 54.3% had preoperative breast MRI compared with 29.9% among patients 65 years and older (51,840 patients, 56.3%; Table 1). For nonelderly and elderly women, 73.8% and 43.5% had no comorbidity, 73.1% and 74.5% were non-Hispanic White patients, 27.8% and 24.9% resided in South Atlantic, and 13.8% and 31.7% had an HMO plan, respectively. Among the nonelderly and elderly cohorts, 27.2% and 14.2% had a family history of BC, 8.4% and 4.8% had axillary node involvement, and 14% and 5.1% received neoadjuvant chemotherapy, respectively. Breast-conserving surgery was performed in 57.9% of nonelderly patients and 70.9% of elderly patients.

TABLE 1.

Patient Characteristics and Treatment by Age Group

The Trend of Preoperative Breast MRI Utilization

The crude rate of preoperative breast MRI increased from 47.9% in 2008 to 59.7% in 2011 for the nonelderly group and from 26.9% to 34.0% for the elderly group (Table 1). The Cochrane-Armitage test demonstrated a statistically significantly increasing trend of preoperative breast MRI use for both age groups.

Factors Associated With Preoperative Breast MRI

Results from multivariable analysis showed that for nonelderly patients (left panel, Table 2), compared with non-Hispanic White women, Asians had significantly higher odds (odds ratio [OR], 1.22; 95% CI, 1.09 to 1.36) of receiving preoperative breast MRI, whereas non-Hispanic Blacks had significantly lower odds (OR, 0.75; 95% CI, 0.70 to 0.81). The adjusted rates of preoperative breast MRI were 59.3%, 54.8%, 54.2%, and 48.2% for Asian, non-Hispanic White, Hispanic, and non-Hispanic Black patients, respectively. Among the elderly group, compared with non-Hispanic White, non-Hispanic Black patients had significantly lower odds of receiving preoperative breast MRI (OR, 0.77; 95% CI, 0.72 to 0.83 [right panel, Table 2]). However, there were no statistically significant differences in the odds of receiving preoperative breast MRI between Asians, or Hispanics and non-Hispanic Whites.

TABLE 2.

Multivariable Logistic Regression Model for Factors Associated With Preoperative Breast MRI by Age Group

There was significant variation in preoperative breast MRI use by geographical location and age group. Nonelderly women residing in the West South (WS) Central division had the lowest odds of receiving preoperative breast MRI (OR, 0.82; 95% CI, 0.72 to 0.93 compared with New England; left panel, Table 2). By contrast, the elderly in the WS Central division had higher odds of receiving preoperative breast MRI (OR, 1.21; 95% CI, 1.07 to 1.36). Overall, women living in these three divisions (New England, East North Central, and WS Central), regardless of age group, had a lower likelihood of receiving preoperative breast MRI than most divisions except East South (ES) Central (for nonelderly only) and West North Central (for elderly only). The adjusted rate of preoperative breast MRI was highest in Mountain (nonelderly and elderly: 60% and 40.4%, respectively), followed by Mid-Atlantic (58%, 31.9%) and South Atlantic (57.3%, 31.5%; Figs 1A and 1B).

FIG 1.

(A) Adjusted preoperative breast MRI rates by Census division, younger than 65 years. (B) Adjusted preoperative breast MRI rates by Census division, 65 years and older. EN, East North; ES, East South; MRI, magnetic resonance imaging; WN, West North; WS, West South.

For the nonelderly group, women who were younger (age, 20-49; OR, 1.45; 95% CI, 1.38 to 1.52), enrolled in non-HMO health plans (OR, 1.16; 95% CI, 1.09 to 1.23), with a family history of BC (OR, 1.60; 95% CI, 1.52 to 1.68) and axillary node involvement (OR, 1.15; 95% CI, 1.06 to 1.26), and received neoadjuvant chemotherapy (OR, 2.52; 95% CI, .35 to 2.71) were more likely to receive preoperative breast MRI (left panel, Table 2). Similar results were found in the elderly group, except that HMO enrollment was not associated with receiving preoperative breast MRI. Compared with nonelderly women with a comorbidity score of two or higher, those with a comorbidity score of 0 or 1 were more likely to receive preoperative breast MRI (comorbidity 0: OR, 1.84; 95% CI, 1.71 to 1.99; comorbidity 1: OR, 1.48; 95% CI, 1.36 to 1.62). Similarly, elderly women with lower comorbidity were more likely to receive preoperative breast MRI. Finally, nonelderly women who had mastectomy (OR, 1.11; 95% CI, 1.07 to 1.16) were more likely to receive preoperative breast MRI than those who had breast-conserving surgery. A similar relationship was observed in the elderly group.

Sensitivity Analysis

Sensitivity analyses stratified by time periods (year 2008-2011, 2012-2015, and 2016-2020) showed that the association between the use of preoperative breast MRI and the sociodemographic and clinical factors examined in the main analysis above was similar across these three time segments (left panel, Appendix Table A1, online only). Among nonelderly patients, our finding suggests that patients who resided in New England were more likely to be early adopters of preoperative breast MRI. For both elderly and nonelderly patients, those with a family history of BC were more likely to be early adopters, whereas those residing in the Pacific were more likely to be late adopters. For the sensitivity analyses that excluded patients who had BC surgery in 2020 (right panel, Appendix Table A1), findings were similar to those from the main analyses.

DISCUSSION

This study examined the use of preoperative breast MRI for women with early-stage BC who underwent BC surgery between 2008 and 2020, focusing on two groups of patients not well understood in the literature: nonelderly with private insurance and elderly in Medicare Advantage. In 2008, the rate of preoperative breast MRI was 48% and 27% among nonelderly and elderly patients, respectively. During this 13-year time span, the rate of preoperative breast MRI had a rapid increase between 2008 and 2011, remained stable between 2011 and 2018, and rose again in 2019. By the end of 2020, the rate of preoperative breast MRI approached 60% (for nonelderly) and 34% (for elderly) of women. Overall, over time, there was a 25% and 26% increase in use of preoperative breast MRI in the nonelderly and elderly cohorts, respectively. Aside from clinical factors known to be associated with preoperative breast MRI, we found variations in the use of preoperative breast MRI by race/ethnicity and geographic region.

Differences in the rate of preoperative breast MRI by race/ethnicity were observed. After controlling for other covariates, non-Hispanic Black women were the least likely to undergo preoperative breast MRI in both the nonelderly and elderly cohorts. Our finding of lower use of preoperative breast MRI among non-Hispanic Blacks in the elderly group is consistent with previous SEER-Medicare studies.^4,15,16 Although a previous study in women age 25-64 years reported no association between race/ethnicity and preoperative breast MRI use, more than 50% of the study cohort did not report race/ethnicity.¹⁸ The reason behind the different rate of preoperative breast MRI use between non-Hispanic Black and non-Hispanic White women is unclear. One possible explanation is more limited access to breast MRI facilities as one study noted longer travel time to MRI facilities.²⁵ However, accessibility to breast MRI is unlikely to be a contributing factor for the observed racial difference in preoperative breast MRI use as a multicenter study reported that non-Hispanic Black women lived closer to an MRI facility than non-Hispanic White women.²⁵ Another possible explanation is that a lower proportion of Black women had dense breasts compared with White women.²⁶ Preoperative breast MRI, with its higher sensitivity in extremely dense breasts, would potentially better define the extent of involvement of the known cancer and also better identify additional sites of cancer in either breast.²⁷ Therefore, we speculated that non-Hispanic Black women may be less likely to benefit from breast MRI to define the extent of disease. Racial differences in breast density would also explain our finding that Asian women had the highest likelihood of receiving preoperative breast MRI than any other race/ethnic group among the nonelderly cohort as Asian women are more likely to have dense breasts.^28,29 However, there was no difference in the use of preoperative breast MRI between Asian and non-Hispanic White patients in the elderly group, likely because of the decrease in breast density in this older cohort,³⁰ irrespective of race/ethnicity. Finally, structural racism in the United States may also attribute to racial disparity in preoperative breast MRI use between non-Hispanic Black and non-Hispanic White patients as previous studies suggested.^31,32 Although these hypotheses are intriguing, more research is warranted to understand this race/ethnic disparity in BC treatment.

To our knowledge, this study is the first to report wide variation in the rate of preoperative breast MRI by Census division. Women with early-stage BC who lived in the Mountain and Mid-Atlantic were more likely to receive preoperative breast MRI than those who resided in any other Census divisions, regardless of age group. For nonelderly patients, the lowest adjusted rate was found among those who lived in the WS Central division, whereas for elderly patients, the lowest adjusted rate was found among those who resided in the New England division. The gap between divisions with the highest and lowest rate of preoperative breast MRI use was more than 10-percentage points for the nonelderly group and more than 15-percentage points in the elderly group. Analyses stratified by time segments indicated early adoption of preoperative breast MRI among nonelderly patients residing in New England and late adoption among patients in both age groups living in the Pacific.

Previous studies using SEER-Medicare data have shown geographic variations in unadjusted rates of preoperative breast MRI across SEER registry areas.^15-17 The difference between the SEER registry location with the highest (New Mexico/Seattle) and the lowest (Hawaii) rate of preoperative breast MRI was more than 10-percentage points. These studies, although informative, were limited to patients with BC who were enrolled in Medicare and resided in one of the SEER registry locations. Thus, estimates reported from our study offer better representation in terms of both geographic coverage and age distribution. Geographic variation in preoperative breast MRI use may be driven by the availability of breast MRI machines²⁵ in the sense that areas with limited capacity of breast MRI likely have lower use of preoperative breast MRI. To explore this, we calculated the number of breast MRI facilities per 100,000 women using American College of Radiology (ACR) breast MRI database.³³ On average, ES Central and WS Central divisions have the lowest number of breast MRI facilities per 100,000 women (87.5 and 63.9, respectively). In our study, these two divisions also had the lowest rates of preoperative breast MRI use than other divisions in the nonelderly cohort, but not in the elderly cohort. The reasons for the geographic variation in preoperative breast MRI use are complex and go beyond the availability of breast MRI facilities. Other contributing factors likely include differences in provider discussion of and recommendation for preoperative breast MRI, patient preference, and other unmeasured factors. Future research using a mixed-methods approach to conduct an in-depth investigation of these factors is needed.

Consistent with previous studies, clinical factors found to be significantly positively associated with preoperative breast MRI use in our study included healthier status (fewer/no comorbidities), family history of BC, axillary nodal disease, receipt of mastectomy, and receipt of neoadjuvant chemotherapy.^3,15-17,34 Two survey studies showed that a strong family history of BC was one of the key determinants that surgeons consider when ordering preoperative breast MRI.^34,35 We found that women with a family history of BC were more likely to be early adopters of preoperative breast MRI. Regarding breast surgery, the plans to proceed with a mastectomy could affect the recommendation for preoperative breast MRI. In addition, mastectomy could be an outcome of preoperative breast MRI on the basis of the MRI experience or MRI results although we are unable to determine a causal relationship in our study. In addition, consistent with ASBrS and CMS guidelines supporting the role of preoperative breast MRI in determining eligibility for neoadjuvant chemotherapy,^13,36 women who had neoadjuvant chemotherapy were more likely to undergo preoperative breast MRI. Finally, we did not find a significant association between genetic susceptibility to BC and preoperative breast MRI. Although somewhat surprising, it is possible that these women underwent a breast MRI as part of high-risk BC screening,³⁷ which would have occurred before BC diagnosis and is therefore not included in our study. It is also possible that women with genetic susceptibility might have elected to undergo bilateral mastectomy, and thus, preoperative MRI might not have been performed.

This study has several limitations. First, the study cohort consisted of women with BC enrolled in a large insurance company, the UnitedHealthcare. As one of the top five insurers in the United States, patients covered in the Optum data are diverse with respect to age, race/ethnicity, and geographic representation. Nevertheless, the utilization pattern of preoperative breast MRI reported in this study may not be generalizable to women with other health insurers, especially those with Medicaid. Second, although claims data provide the opportunity to analyze the real-world utilization pattern of preoperative breast MRI among women with early-stage BC at a scale that better reflects the practice patterns observed in the overall population than those from single-institution studies or clinical trials, the data often do not contain all clinically relevant information that may affect preoperative breast MRI use, such as breast density, subtypes of BC (eg, invasive lobular cancer, occult BC, Paget's disease), indeterminate radiographic findings, or clinical-radiographic discordance.^4,13,15 In addition, over 70% of the study population was non-Hispanic White patients, which may limit the statistical power to make inferences in the use of MRI in the other racial and ethnic groups. Finally, other factors associated with preoperative breast MRI use documented in previous studies were not observable in our study, such as marital status,^14,16,18 urbanicity,^4,15,16 physician's volume,¹⁵ and physician's practice preference.^15,17,35 Information on patient-provider networks that can potentially affect the diffusion of preoperative MRI was also not available.

In conclusion, the use of preoperative breast MRI exhibited an upward trend between 2008 and 2020 for women with early-stage BC, with the rates among nonelderly patients consistently higher than those among elderly women. In addition to age and clinical characteristics, race/ethnicity and geographic location played a significant role in the receipt of preoperative breast MRI.

APPENDIX

TABLE A1.

Sensitivity Analysis: Part I and Part II

AUTHOR CONTRIBUTIONS

Conception and design: I-Wen Pan, Tina W.F. Yen, Ya-Chen Tina Shih

Financial support: Ya-Chen Tina Shih

Administrative support: Ya-Chen Tina Shih

Collection and assembly of data: I-Wen Pan, Ya-Chen Tina Shih

Data analysis and interpretation: All authors

Manuscript writing: All authors

Final approval of manuscript: All authors

Accountable for all aspects of the work: All authors

AUTHORS' DISCLOSURES OF POTENTIAL CONFLICTS OF INTEREST

Current Trends in the Utilization of Preoperative Breast Magnetic Resonance Imaging Among Women With Newly Diagnosed Breast Cancer

The following represents disclosure information provided by authors of this manuscript. All relationships are considered compensated unless otherwise noted. Relationships are self-held unless noted. I = Immediate Family Member, Inst = My Institution. Relationships may not relate to the subject matter of this manuscript. For more information about ASCO's conflict of interest policy, please refer to www.asco.org/rwc or ascopubs.org/op/authors/author-center.

Open Payments is a public database containing information reported by companies about payments made to US-licensed physicians (Open Payments).