Characteristics of the study population

Overall, 862 women (BCS, 357 (41%); CF, 505 (59%)) participated in this QoL survey, including 283 (33%) Namibian Black (BCS, 136; CF, 147) and 107 (12%) non-Black (58:49, respectively) women, 227 (26%) Nigerian (75:152), and 246 (29%) Ugandan (88:157) women (Table 1). In BCS, mean time since diagnosis was 6.7 years (± 0.7, min–max = 5.0–8.7). BCS had slightly higher prevalence of early-stage diagnosis but similar SEP indicators than ABC-DO women who were potential BCS but lost to follow-up (n = 110, % of full ABC-DO cohort: 6% of Black and 13% of non-Black Namibians, 6% of Nigerians, and 11% of Ugandans) (Additional Table 1). BCS were on average ~ 2 years older than CF women (mean age, 57.5 vs. 55.1 years, driven by older average age in the 65 + age group), were more educated (higher education, 30% vs. 20%), and more likely to have ever drunk alcohol (55% vs. 45%) or to be married (52% vs. 46%). While the prevalence of HIV positivity among BCS was similar to that of CF women (13% vs. 16% in Namibian Black and 13% vs. 12% in Uganda, respectively; ≤ 3% for Namibian non-Black and Nigerian BCS and CF women), the prevalence of obesity and of other comorbidities were lower in BCS than in CF women (28% vs. 36% and 42% vs. 50%, respectively), the latter being driven by lower prevalence of hypertension (35% vs. 42%) and diabetes (5% vs. 13%) in BCS.

Table 1 Characteristics of the ABC-DO breast cancer survivor subset and the group of cancer-free women who completed the QoL surveyPerformance of the WHOQOL-BREF questionnaire in the study population

Internal consistency was acceptable for all QoL domains, with Cronbach’s alpha coefficients ranging from 0.62 for the social relationships domain to 0.86 for the physical health domain (Additional Table 2). All domain scores were positively correlated with both the QOLgh and QOLov scores and with each other (Additional Table 2). QOLgh and QOLov were most strongly correlated with both the psychological (Pearson correlation coefficient (r) = 0.68 and 0.67, respectively) and physical health domains (r = 0.66 and r = 0.59, respectively) and the least with the social relationships domain (r = 0.42 and r = 0.43, respectively). The psychological domain correlated most strongly with all others (physical health, 0.77; environment, 0.67; social relationships, 0.62).

WHOQOL-BREF QoL scores in BCS

Overall, BCS had good QOLov and QOLgh scores (mean score (95% CI) 80% (77, 82) and 77% (75, 79), respectively) (Fig. 1, Additional Table 3). BCS had high QoL scores in the physical health (76 (74, 78)), psychological (76 (74, 77)), and social relationships (i.e., calculated without considering sexual life satisfaction) (77 (75, 79)) domains, while that of the environment domain was 12 p.p. lower (65 (64, 67)). In each domain, less than 10% of BCS reported poor or very poor QoL. However, there were between-settings disparities in QoL, with scores being constantly highest in Namibia (and within Namibia, in non-Black women) and lowest in Uganda.

Fig. 1

Distribution of WHOQOL-BREF item scores by QoL domain, in ABC-DO breast cancer survivors and cancer-free women. BCS, ABC-DO breast cancer survivors; QoL, quality of life. NB: Adjusted mean scores (95% CI) obtained from linear regression models adjusted for interviewer (categorical), country (Namibia, Nigeria, Uganda), race (Black, non-Black), and age (< 45, 45–49, 50–54, 55–59, 60–64, and 65 + years) and predicted for a woman in the age category 55–59 years, except for Namibian non-Black women (the age category 60–64 years was used for these women because they were, on average, older than those in other settings), interviewed by the interviewer whose adjusted mean QoL score was closest to the overall adjusted mean QoL score

Differences in general WHOQOL-BREF items between BCS and CF groups

Overall, BCS had higher QoL scores than their CF counterparts across all general items/domains of the questionnaire (mean BCS-CF absolute difference in scores: QOLov, + 13 p.p.; QOLgh, + 7 p.p.; from + 4 p.p. (physical health) to + 8 p.p. (social relationships) for WHOQOL-BREF domains). Upon adjustment for design variables, BCS-CF differences in QOLov and QOLgh were slightly attenuated (AMD (95% CI) 10.65 p.p. (7.56, 13.74) and 5.84 p.p. (2.71, 8.98), respectively) (Fig. 2). While BCS’ QOLov scores were higher in every setting (min, + 7% in Namibian non-Black women; max, + 15% in Nigerians), there was no evidence of a BCS-CF difference in QOLgh in Namibian non-Black women (range, from + 6% in Namibian urban Black women to + 10% in Nigerians). Results were similar when restricting to HIV-negative women (Additional Fig. 1). Both QOLov and QOLgh increased with increasing educational level and with being married, and, for QOLov, also with increasing BMI. In contrast, both QOLov and QOLgh were lower in women who had comorbidities, had higher number of children at home, and, for QOLgh, also in those who ever used tobacco or resided in rural areas (Additional Table 4). In non-Black Namibians, where BCS were substantially more educated than CF women (technical/university education, 48% vs. 29%, respectively), further adjustment for educational level reduced the magnitude of the BCS-CF AMD in QOLov score by − 2.3 p.p. (i.e., ~ 10% of a change in QoL category) (Table 1, Additional Fig. 2), but did not affect that in QOLgh score (Additional Fig. 3). BCS-CF AMD in QOLov and QOLgh scores were marginally changed when adjusting for the other correlates (Additional Figs. 2 & 3).

Fig. 2

Adjusted mean differences (AMD) in WHOQOL-BREF domain scores between breast cancer survivors (BCS) and cancer-free (CF) women, by country and race. AMD, adjusted mean difference; BCS, breast cancer survivors; CF, cancer-free women; 95% CI, 95% confidence interval; QoL, quality of life. NB1: Absolute BCS-CF adjusted mean difference (AMD) in QoL scores in percentage points (p.p.), with 95% CI, obtained from linear regression models adjusted for interviewer (categorical), country (Namibia, Nigeria, Uganda), race (Black, non-Black), and age (< 45, 45–49, 50–54, 55–59, 60–64, and 65 + years). NB2: Adjusted mean scores (95% CI) predicted for a woman in the age category 55–59 years, except for Namibian non-Black women (the age category 60–64 years was used for these women because they were, on average, older than those in other settings), interviewed by the interviewer whose adjusted mean QoL score was closest to the overall adjusted mean QoL score

Differences in WHOQOL-BREF QoL domains scores between BCS and CF groupsPhysical health domain

There was no evidence of BCS-CF differences in physical health scores in any of the settings before or after adjustment for design variables (AMD (95% CI) 1.52 p.p. (− 0.82, 3.87)) (Fig. 2); however, there was large heterogeneity between domain-specific items (Figs. 1 and 3). BCS reported substantially higher levels of fatigue and similar working capacity and ability to perform daily living activities than CF women, but they scored higher in the other items (Fig. 3). Scores were higher in more educated or married women, but were lower in rural women, in those with non-HIV related comorbidities, or in ever tobacco users (Additional Table 4). In each setting, adjusting for any of these correlates changed marginally the magnitude of the BCS-CF AMD in physical health scores (i.e., all < 10% of a change in QoL category) (Additional Fig. 4). Interestingly, in Nigeria and Uganda (i.e., the two countries where this analysis could be undertaken), while urban BCS had higher physical health scores than urban CF women, in rural women, there was no BCS-CF difference (data not shown).

Fig. 3

Adjusted mean differences in WHOQOL-BREF domain items between breast cancer survivors (BCS) and cancer-free (CF) women. AMD, adjusted mean difference; BCS, breast cancer survivors; CF, cancer-free women; 95% CI, 95% confidence interval; QoL, quality of life. NB1: Absolute BCS-CF adjusted mean difference (AMD) in item scores, expressed on a scale from 1 (lowest score = worst QoL) to 5 (highest score = best QoL), with 95% CI, obtained from linear regression models adjusted for interviewer (categorical), country (Namibia, Nigeria, Uganda), race (Black, non-Black), and age (< 45, 45–49, 50–54, 55–59, 60–64, and 65 + years). NB2: Adjusted mean scores (95% CI) predicted for a woman in the age category 55–59 years, except for Namibian non-Black women (the age category 60–64 years was used for these women because they were, on average, older than those in other settings), interviewed by the interviewer whose adjusted mean QoL score was closest to the overall adjusted mean QoL score

Psychological domain

Psychological scores adjusted for design variables remained 4% to 6% higher in BCS than in CF women in Uganda, Nigeria, and Namibia non-Black population groups (AMD (95% CI) 3.86 p.p. (− 0.05, 7.76), 6.43 (2.20, 10.66), and 4.93 (0.61, 9.25), respectively); however, there was no evidence of a BCS-CF difference in Black Namibians (Fig. 2). While in most settings, BCS reported similar levels of positive feelings as CF women and higher scores in the other items, Namibian Black BCS scored lower for both spirituality and positive feelings and did not differ from CF women for cognitive abilities and self-esteem (data not shown). Bodily acceptance was the only item higher in BCS than in CF women in all settings (Fig. 3). Scores increased with both increasing educational level or BMI, and with being married, but were lower in rural women, in those with comorbidities, or in ever tobacco users (Additional Table 4). However, adjusting for each of these correlates changed marginally the magnitude of the BCS-CF difference in psychological domain scores (Additional Fig. 5).

Social relationship domain

Social scores adjusted for design variables remained 5% to 8% higher in BCS in most settings (min–max AMD (95% CI) 5.11 p.p. (0.27, 9.94) (Black Namibians); 7.76 (3.20, 12.32) (non-Black Namibians)) driven by higher QoL regarding personal relations (Figs. 2 and 3). In contrast, in Nigeria where social support scores were lower than in the other settings (moderate QoL or lower, ~ 60% in both groups), there was no evidence of a BCS-CF difference in both domain and items scores (Fig. 2, data not shown). Scores increased with both increasing educational level or BMI and with being married (Additional Table 4). However, BCS-CF differences in scores changed marginally when adjusting for any of these correlates (Additional Fig. 6). Among respondents, sexual life satisfaction was substantially lower in BCS than CF women in all settings except Uganda, where scores were low in both groups (poor to very poor QoL: BCS, 32%; CF, 44%) (Figs. 1 and 3, data not shown).

Environment domain

Environmental scores adjusted for design variables remained ~ 4% higher in BCS in all settings (min–max AMD (95% CI) 3.65 p.p. (− 0.10, 7.41) (Nigeria); 10.77 (5.71, 15.83) (Namibian non-Black women)), except in Black Namibians, for which the lack of evidence of a BCS-CF difference was driven by rural BCS (urban Black Namibians, 5.40 (− 0.13, 10.94)) (Fig. 2). Both groups scored lowest in financial resources (poor to very poor QoL: BCS 49% vs. CF 64%), opportunities for recreation and leisure activities (30% vs. 42%, respectively), and access to transportation (19% vs. 26%) (Fig. 1). While scores in the latter two items and access to health and social care were similar for BCS and CF women, BCS scored substantially higher in all other domain-specific items (Fig. 3). Environment scores increased with increasing educational level, increasing BMI and being married, but were lower in women who resided in rural areas, in those who were HIV-positive, or in those who ever used tobacco (Additional Table 4). Further adjustment for educational level (i.e., a SEP indicator) reduced the magnitude of the BCS-CF differences in scores among Namibian non-Black women by 2.6 p.p. (~ 10% of a change in QoL category) but only changed marginally the estimates in the other settings (Additional Fig. 7). Adjusting for each one of the other correlates did not affect BCS-CF differences in scores. For all QoL domains, results from adjusted models remained unchanged in sensitivity analysis (Additional Fig. 1).