Physician Sex and Other Factors Associated With Type of Breast Cancer Surgery in Older Women

Hypothesis  Physician-related factors as well as patient characteristics may explain why women aged 65 years or older with early-stage breast cancer undergo lumpectomy less often than younger women, despite National Institutes of Health recommendations favoring lumpectomy over mastectomy.

Design  A descriptive and analytical retrospective computer-assisted telephone survey.

Setting  A population-based random sample of breast cancer survivors in Colorado, identified from the Colorado Central Cancer Registry.

Patients  Women aged 65 to 84 years when diagnosed as having stage I or II breast cancer, treated 1 to 6 years previously with mastectomy or lumpectomy, and without recurrence or second primary cancers. Among women contacted, 58% participated. Results of 198 interviews are reported.

Methods  Survey questions included patient decision-making participation and physician recommendations, sources and amount of treatment information provided by physicians, physician characteristics, and patient surgery preferences and demographic characteristics. A multivariate logistic regression model identified factors independently associated with lumpectomy.

Results  Lumpectomy was strongly associated with higher patient education, female physician sex, patient age 75 years or older, and amount of physician-provided information. The number of physician-provided information sources was associated with surgery explanations, and female physicians provided more sources of information. A physician decision or recommendation for surgery type was reported by 61% of women, of whom 93% underwent the recommended procedure. A subset of patients (13%) reported deferring the surgery decision to someone else.

Conclusions  These results suggest that better-educated and better-informed older women are more likely to undergo lumpectomy, and that physicians may influence breast cancer patients' decisions about surgery type.

MORE THAN 150 000 women develop breast cancer each year. Almost half (45%) are aged 65 years or older, because of a progressive rise in incidence with age.¹ Seventy-five percent to 80% have cancers that are stage I or II ("early stage") according to American Joint Committee on Cancer staging criteria.² Clinical trials indicate that they can be effectively treated with either breast-conserving surgery (lumpectomy) with radiation or modified radical mastectomy.³ In 1990, the National Institutes of Health Consensus Development Conference recommended lumpectomy over mastectomy as preferable treatment for stage I and II disease, because of equivalent overall survival and less negative body-image perceptions.^4-9 Most women are candidates for lumpectomy, the only contraindications being a large tumor-to-breast size ratio leading to poor cosmetic results, or multicentric disease.¹⁰

Observed rates of operations, however, suggest that lumpectomy is underused. In particular, older women are less likely than younger women to receive a lumpectomy.^11-16 The reasons for this are unclear. Although fewer breast cancer procedures per surgeon and smaller hospital size have been associated with lumpectomy underuse, these factors do not explain the lower rates in older patients,¹¹ nor are they explained by patient comorbidity; greater age remains an independent risk factor after adjustment for comorbidity.^17,18 Possible explanations include differences in older women's involvement in decision making,^19,20 in physician recommendations,^21,22 and in treatment information.²³ In addition, older women may place different importance on quality-of-life outcomes, such as body image or treatment convenience.^24,25 Little is known about these issues in older women²⁶; therefore, the purpose of this study is to describe factors associated with the type of surgical procedure performed in older women with early-stage breast cancer.

The study was designed as a descriptive and analytical retrospective survey, using a computer-assisted telephone interview of 200 older Colorado women who were surgically treated for early-stage breast cancer. The report is part of a larger study evaluating breast cancer treatment satisfaction among Colorado Medicare beneficiaries.

Women with stage I or II breast cancer aged 65 to 84 years at diagnosis, treated with either mastectomy or lumpectomy between January 1, 1991, and December 31, 1996, and without known recurrence or second primary cancer, were identified from the records of the Colorado Central Cancer Registry, the statewide tumor registry maintained by the Colorado Department of Public Health and Environment. Women aged 85 or older at diagnosis were excluded because of the increased likelihood that memory and/or hearing loss would impair their ability to participate in the retrospective telephone survey.²⁷ Before sampling, the population was stratified by race or ethnicity, procedure, and year of diagnosis. One specific study objective was to assess experiences of minority women; therefore, all eligible Hispanic and African American women were included, whereas non-Hispanic white women were randomly sampled. Equal numbers of women with each surgery type were selected. The sample total was 681 women.

In accordance with Colorado Central Cancer Registry guidelines, a letter was mailed to patients' physicians of record explaining the study, verifying approval by the Colorado Multiple Institutional Review Board (which governs research involving human subjects at the University of Colorado Health Sciences Center, Denver), and asking for permission to invite the patient to participate. Nonresponding physicians were contacted by a second mailing and telephone call. Once permission was obtained, a research assistant recruited patients with a precontact letter and follow-up telephone call. Standard survey procedures were used, including 8 attempts to reach each woman and varying calling times. Aggressive survey methods, such as trying to convince hesitant patients to participate and calling back those who initially refused, were not used. All interviews were conducted between May and November 1997.

A survey instrument containing both closed- and open-ended questions was designed for use in a structured, computer-assisted telephone interview. Development was based on focus group discussions with non-Hispanic white, African American, and Hispanic women previously treated for breast cancer; literature review; expert consultation; and pilot testing. Survey questions asked women to identify the specialty and sex of the main physician most involved in helping them decide on or plan surgery. Questions asked women whether they recalled having a choice of surgery in their particular situations, and if not, why. Women were asked whether the main physician made the decision about the type of surgery, made a recommendation but left the final decision up to them, or made no recommendation at all. Survey respondents who reported receiving a physician recommendation were asked about the nature of the recommendation. Women who recalled a surgical choice were asked their reasons for procedure type, with the use of an open-ended question allowing multiple answers.

Six questions to assess patients' desire for participation in decision making, modified from the Autonomy Preference Index (API), were included.¹⁹ Preferences for decision making were determined by adding the scores of these 6 items, then linearly adjusting totals to range from 0 to 100, where 0 corresponds to a complete lack of desire for decision making and 100 corresponds to the strongest possible desire. In the initial report, there was excellent internal consistency of the API, with a Cronbach α coefficient of 0.82.¹⁹ In our study, the Cronbach α coefficient was 0.59.

To explore the issue of information, patients were asked whether they were offered 6 types of information: discussions with the physician, with office staff, or with other women; written materials; videos; or second opinions. Answers to each item were added, for total scores ranging from 0 to 6. Also evaluated was the number of information sources independently sought by patients (literature, video, second opinion, or consultation with other women), with total scores ranging from 0 to 4. Finally, demographic information on participants was obtained.

All data were analyzed with SAS version 6.12 (SAS Institute Inc, Cary, NC). Bivariate analyses were performed by means of χ² tests to compare proportions. Continuous data were analyzed with a standard t test or analysis of variance. Finally, multivariate logistic regression was performed to determine factors independently associated with the outcome of lumpectomy, which are expressed as adjusted odds ratios with 95% confidence intervals. Factors examined for association with procedure type with P values of .10 or less were entered into the model, which was adjusted for age, race or ethnicity, and year of diagnosis.

Of 681 cases identified from Colorado Central Cancer Registry records, no permission was given to contact 170 (25%). After duplicate names were removed, 488 women were available for recruitment. Of these, 57 (12%) were unreachable, 4 (0.8%) were non–English speaking, 27 (6%) had illness precluding interview, 2 (0.4%) were deceased, and 3 (0.6%) had cancer recurrence. Of the remaining 395 women, 167 (42%) declined and 228 (58%) agreed to be interviewed. Twenty-eight interviews, not reported, were conducted to pilot test the questionnaire. Of 200 interviews completed with the final instrument, 198 were in women without recurrence and eligible for analysis.

The mean age of respondents at diagnosis was 72 years (range, 65-84 years), and the mean number of years from diagnosis was 3.2. Their composition was 81% non-Hispanic white, 13% Hispanic, and 6% African American. Response rates were similar across these groups. Nearly equal numbers of women with each procedure were interviewed (96 with lumpectomy and 102 with mastectomy). Nonrespondents did not differ from respondents in race or ethnicity, mean age at or year of diagnosis, or procedure type. Women not contacted because of physician refusal were somewhat older than interviewed women (74 vs 72 years). The reasons given for nonapproval (eg, nursing home occupancy and impaired cognitive function) may explain this.

Table 1 shows women's recall of the treatment decision-making process. Overall, 74% of women recalled they had a choice of procedure, and 24% recalled they did not. Of the latter, 22 women (11% of total women) reported that their tumor was too large or advanced. All had a mastectomy. They did not differ in age, race or ethnicity, marital status, income, education, health status, or urban vs rural residence from respondents who recalled a treatment choice. Since clinical reasons explained their surgery, they were excluded from the multivariate analysis.

The remaining 26 women (13% of total) reported a lack of choice for other reasons, most commonly that someone else instructed them on the type of surgery to have (20/26). Many reported that the physician made the procedure decision; less commonly a family member made it. Twelve had a lumpectomy and 14 had a mastectomy. These 26 women were different from those who reported a procedure choice; they were more likely to be 75 years or older at diagnosis (52% vs 24% among those who had a choice; P = .001) and more likely to be Hispanic (27% vs 10% among those who had a choice; P = .02). They were retained in the multivariate analysis because there was no apparent clinical reason for lack of procedure option.

Finally, Table 1 shows question results regarding physician recommendations. Of 107 women who reported a physician decision or recommendation, 93% had the procedure recommended. Of those not following the recommendation, 4 of 5 chose lumpectomy.

Among all women who chose a mastectomy, the most common reason was fear of recurrence (61%). Physician recommendation was second (41%), and "just wanting to get treatment over with" was third (37%). Among women who chose a lumpectomy, keeping the breast was the most frequent reason (43%); equivalent survival was mentioned by 41% and physician recommendation by 37%. Women with female physicians were more likely than women with male physicians to mention concern about appearance (25% vs 7%; P = .004) or keeping their breast (67% vs 38%; P = .05) as reasons for lumpectomy, and less likely to mention equivalent survival (0% vs 50%; P = .001). Overall, 29% with mastectomy and 20% with lumpectomy reported a current fear of recurrence (P = .16).

Table 2 shows the modified API results for all women, assessing preference for decision-making involvement. The mean score was 62.6 (SD, 19.7). There were no differences in mean scores by procedure type, age, or physician sex. However, Hispanic women and those with less education scored significantly lower in autonomy preferences. Women who recalled a lack of procedure choice (for clinical or nonclinical reasons) also had lower scores.

The number of information sources received was evaluated in the 176 women without clinical lack of choice. The mean number of sources was 2.3. Higher scores were associated with having a lumpectomy vs mastectomy (2.5 vs 2.0; P = .01), age less than 75 years vs 75 years or more (2.45 vs 1.9; P = .01), decision participation vs deferral (2.4 vs 1.6; P = .003), and having a female vs a male physician (3.0 vs 2.2; P = .006). The number of sources offered was not related to race or ethnicity, or educational level. A higher mean number of sources was strongly associated with receiving an explanation of lumpectomy compared with not (2.5 vs 1.3; P<.001) and trended toward association with receiving an explanation of mastectomy compared with not (2.4 vs 1.8; P = .05).

The mean score for the number of information sources sought independently by patients was 1.7. The mean did not differ by surgery type, race or ethnicity, age, physician sex, or education level; however, patients who deferred decision making on surgery type sought less information on their own than did patients who made decisions (1.5 vs 1.8; P = .04).

Table 3 displays the patient demographics and other factors examined for association with treatment type by means of bivariate analyses. There were no significant differences between the 2 surgery groups in age at diagnosis, race or ethnicity, mean number of years since diagnosis, marital status, urban vs rural residence, self-reported health status, or health maintenance organization insurance. Women with a lumpectomy had a higher educational level (P = .001). There was a nonsignificant association between higher income and lumpectomy (P = .08), perhaps because of an association between education and income (P = .001; data not shown). There was no significant difference in the specialty of the main physician involved in treatment planning, but women with a lumpectomy were more likely to have had a female physician (P = .002). Most of these female physicians were surgeons; the breakdown of their specialties was identical to that of male physicians. There was no association between treatment barriers (financial, transportation, distance, and home obligations) and procedure type.

Table 4 shows the results of the multivariate logistic regression performed to identify factors independently predictive of lumpectomy. Higher level of education, female sex of the physician planning surgery, age 75 years or older at diagnosis, and the number of sources of physician-provided information were positively associated with receiving a lumpectomy.

Further analyses to explain results of this multiple logistic regression were reviewed. Women with more than a high school education were significantly less likely to report that they received a recommendation of mastectomy (30% vs 60% with less education; P = .002). Higher patient educational level was associated with female physician sex (P = .04); however, physician sex remained independently associated with lumpectomy when education was controlled for. Female physicians tended to recommend lumpectomy more frequently than did male physicians (67% vs 52%), although this was not statistically significant (P = .30).

This study was designed to provide insight into how surgical treatment decisions are made in older women with early-stage breast cancer, given evidence that lumpectomy with radiation offers survival equivalent to that with mastectomy but is underused. Of 198 women interviewed, only 22 (11%) reported having clinical reasons for mastectomy, including multicentric disease and tumor size. This is consistent with previous observations¹⁰ that most women with early-stage disease are candidates for breast-conserving surgery (89% in our study). This latter group of women was the main focus of our study.

We found that 13% of all women recalled they did not have a choice of surgery because a decision was made for them, usually by a physician. These women were more likely to be older and Hispanic. They also scored lower in preferences for autonomy in decision making. This indicates that their lack of choice may have been due to "paternalism with permission."¹⁹ In these situations, patients transfer their right to decision making to a physician or family member, and greater age has been reported as the most important predictor of this type of decision-making style.¹⁹ We cannot be certain that this lack of choice was due to "paternalism with permission" rather than other factors, such as less physician facilitation of involvement in decision making. However, women who did not participate in decision making sought less information on their own, which suggests they may have been more passive. In a Canadian study, patients with breast cancer older than 70 years preferred a more passive role in decision making.²⁸ Ethnic differences in API results, such as the lower desire for autonomy found in Hispanic women, have not been previously reported. However, our results are consistent with those of others suggesting that minority patients have less participatory decision-making interactions with physicians.²⁹

The multivariate logistic regression analysis identified factors predictive of lumpectomy. Higher education had the strongest association, which has been reported previously.^12,13 The reasons for this are unknown. We found no association between higher education and more information offered by physicians or sought by patients. Educated women may be better able to understand and choose among treatment alternatives, although we found no differences by education in reasons given for choice of procedure (even those related to knowledge of outcomes such as recurrence and overall survival). We did find that women with less education scored lower in autonomy preferences, similar to another report that women of lower education preferred a more passive role in decision making.²⁸ Women with less education were more likely to recall a physician recommendation for mastectomy, but whether this resulted from differences in physician approach or from preferential recall of information supporting treatment choices is not clear. However, literature suggests that there may be relationships between patient sociologic factors and physician clinical decision making.³⁰

Female physician sex was also independently associated with lumpectomy in the multivariate model. Other than education, there were no associations of patient demographic or personal characteristics (including race or ethnicity, age, marital status, urban vs rural residence, income, or autonomy preferences) with female physicians. Women with female physicians may differ in other characteristics, such as beliefs and attitudes, which could influence both physician selection and surgery preference. We did find differences in their reasons for choosing lumpectomies. However, this could also indicate that physicians vary by sex in how they present surgical options to patients with breast cancer. Female physicians may emphasize more patient-centered outcomes, such as body appearance (favoring lumpectomy), while male physicians may emphasize more disease-centered outcomes, such as recurrence (favoring mastectomy).

Interestingly, study participants with female physicians recalled being offered more sources of information about treatment than those with male physicians. This finding has been suggested in other studies, including an analysis of audiotaped conversations between patients and physicians in a variety of primary care settings, which concluded that female physicians engaged in more information giving and partnership building than did male physicians.^31-33

Although year of diagnosis was not related to procedure, more patients were treated by female physicians during 1994 to 1996 than during 1991 to 1993, perhaps reflecting the recent trend of training more female physicians, including surgeons. This suggests that other factors, such as the number of years since training, younger age, or practice at larger or teaching hospitals, might confound the relationship between physician sex and lumpectomy. We cannot test for this, as individual data on physicians were not collected. In addition, the number of patients who reported having female physicians, on whom these results are based, is modest. Therefore, our results regarding physician sex must be interpreted with caution.

Receiving more sources of physician-provided information was independently related to having a lumpectomy. The relationship shown between information and receipt of a lumpectomy explanation could suggest that more physician-provided information influences treatment type by including an explanation of lumpectomy. Some states (not including Colorado) have passed legislative mandates requiring disclosure of both types of breast cancer treatment options. However, these mandates have had limited effects on increasing lumpectomy rates, perhaps because they cannot alter a physician's behavior regarding information disclosure.^34,35 Given the retrospective nature of our study, we cannot conclude that physician disclosure and explanation of options influenced surgery type; instead, women may have preferentially remembered receiving only information about their procedure (a recall bias). However, it is less likely that procedure type would bias recall of the overall amount of physician-provided information received (which was positively associated with lumpectomy).

The remaining predictor in the logistic model was age at diagnosis. Our results suggest that women older than 75 years may undergo lumpectomy more often than those aged 65 to 74 years. This finding is similar to other literature suggesting that "older-old" women undergo lumpectomy to avoid more extensive surgery.¹⁸ A physician recommendation for lumpectomy was reported slightly more frequently by women aged 75 years or older than by younger women. Since women aged 75 years or older are more likely to defer to physicians for decisions, these recommendations may explain our results.

The influence of physician recommendations on procedure received was demonstrated in several other findings, including the high association between procedure type recommended and procedure undergone (93% concordance). Physicians recommended mastectomy 45% of the time to women in our study who (according to our knowledge) had no clinical contraindications to breast-conserving surgery, despite National Institutes of Health recommendations favoring lumpectomy for most women. Previously, it has been suggested that physicians may have personal opinions about procedure options that influence their recommendations more than do data from scientific trials.^21,22 Liberati et al²¹ found that some physicians favored radical breast surgery for prevention of local disease recurrence, and some physicians favored limited breast surgery, for reasons of equal survival and better cosmetic results. These physician beliefs resemble those stated by women in our study, who listed fear of recurrence and physician preference as reasons for mastectomy, and equal survival and physician preference as reasons for lumpectomy. The high frequency with which women reported reasons of "physician preference" is also consistent with findings by Ward et al,²⁵ who noted that physician recommendation was the third most common factor taken into account by women when choosing breast cancer surgery.

Last, we found that women who had a mastectomy tended to worry more currently about recurrence. This may explain their initial selection of mastectomy as treatment, but it also suggests that having a mastectomy may not adequately reduce their worries. It has been shown that women overestimate their risk of cancer recurrence.³⁶ To increase the use of lumpectomy, therefore, fear of recurrence may need to be addressed with improved education and counseling.

There are some important limitations of our study to address. The number of respondents was slightly more than half of those surveyed. Aggressive methods, such as trying to convince hesitant potential respondents to participate, and calling back those who initially refused, were not used, given respect for patients' privacy. A higher response rate may have altered our results. Although there were no obvious differences in nonrespondents in terms of age, race, or procedure, other characteristics such as educational level may have differed. The amount of time elapsed since diagnosis and treatment (chosen to increase the number of eligible minority participants) may have compromised detailed memory and introduced the possibility of recall bias. The questionnaire was worded to achieve general rather than specific detail, and only 6% of women reported having a very difficult time remembering answers. Another limitation was a lack of access to the original pathology reports. We were thus unable to verify the appropriate use of mastectomy in patients with stage I or II disease who reported a lack of procedure choice because of large or advanced tumors, or to evaluate whether there may have been other anatomic considerations affecting treatment type in some women. Last, since this study focused on Medicare-eligible women, there was no younger group of women for comparison.

Our results suggest that better informed and educated older women are likely to choose lumpectomy over mastectomy, and that physicians have an important role in patients' decisions about surgery, such as providing information, making recommendations, and making decisions for patients who defer to them. Physician-related factors such as sex may modify the effect of the physician's role in treatment decisions. To understand these results, further aspects of the physician-patient decision-making process could be explored. Examination of physician interactions with less educated women, including how information is provided, may be revealing. Less educated, older, and/or minority patients may require more physician facilitation of information gathering and decision-making participation. How physicians make recommendations for patients who defer decisions would also be of interest. Further studies of the relationship between physician sex and treatment type could be undertaken to see if current findings are reproducible. If so, determining whether this is prompted by differences in patient behavior, communication interactions, or physician behavior would have important implications for medical education and for women's health care. As the demographics of the US population changes, with more minorities and older adults, insight into their health care needs and experiences will become increasingly important to health care providers.

Dr Cyran was supported in part during fellowship training at the University of Colorado Health Sciences Center by National Research Service Award grant 5T32-PE10006. The analyses on which this publication is based were performed under contract 500-96-P611, entitled "Utilization and Quality Control Peer Review Organization for the State of Colorado," sponsored by the Health Care Financing Administration, Department of Health and Human Services, Washington, DC.

The content of this publication does not necessarily reflect the views or policies of the Department of Health and Human Services, nor does mention of trade names, commercial products, or organizations imply endorsement by the US government.

The authors assume full responsibility for the accuracy and completeness of the ideas presented. This article is a direct result of the Health Care Quality Improvement Program initiated by the Health Care Financing Administration, which has encouraged identification of quality improvement projects derived from analysis of patterns of care, and therefore required no special funding on the part of this contractor. Ideas and contributions to the authors concerning experience in engaging with issues presented are welcomed.

We thank Elizabeth Hopewell, BS, MSPH, for her research assistance, and Jack Finch, MS, Robin Bott, and Wei Lin from the Colorado Central Cancer Registry and Tim Byers, MD, and Lari Wenzel, PhD, for their dedicated help to this research study. We also acknowledge the contributions of Robin DiMatteo, PhD, Patricia Ganz, MD, and Celia Kaplan, PhD, who reviewed the questionnaire during its design.

Corresponding author and reprints: Elizabeth M. Cyran, MD, MSPH, Division of General Internal Medicine, Campus Box B-212, University of Colorado Health Sciences Center, 4200 E Ninth Ave, Denver, CO 80262 (e-mail: Lisa.Cyran@UCHSC.edu).