Incidence and risk factors for breast cancer subtypes in three distinct South-East Asian ethnic groups: Chinese, Malay and natives of Sarawak, Malaysia

Authors


Abstract

We determined the incidences of the estrogen receptor (ER), progesterone receptor (PR) and human epidermal growth factor receptor 2 (HER2) subtypes among breast cancer cases in Sarawak, Malaysia and their correlation with various risk factors in the three ethnic groups: Chinese, Malay and native. Subtype status was ascertained for 1,034 cases of female breast cancer (93% of all cases diagnosed since 2003), and the age-standardized incidence rates (ASRs) of each subtype were inferred. Case–case comparisons across subtypes were performed for reproductive risk factors. We found 48% luminal A (ER+/PR+/HER2−), 29% triple-negative (ER−/PR−/HER2−), 12% triple-positive (ER+/PR+/HER2+) and 11% HER2-overexpressing (ER−/PR−/HER2+) subtypes, with ASRs of 10.6, 6.0, 2.8 and 2.8 per 100,000, respectively. The proportions of subtypes and ASRs differed significantly by ethnic groups: HER2-positive cases were more frequent in Malays (29%; 95% CI [23;35]) than Chinese (22%; [19;26] and natives (21%; [16;26]); triple-negative cases were less frequent among Chinese (23%; [20;27]) than Malays (33%; [27;39]) and natives (37%; [31;43]). The results of the case–case comparison were in accordance with those observed in western case series. Some uncommon associations, such as between triple-negative subtype and older age at menopause (OR, 1.59; p < 0.05), were found. The triple-negative and HER2+ subtypes predominate in our region, with significant differences among ethnic groups. Our results support the idea that the risk factors for different subtypes vary markedly. Westernized populations are more likely to have factors that increase the risk for the luminal A type, while risk factors for the triple-negative type are more frequent in local populations.

In Asia, breast cancer is the commonest cancer in the two genders combined,1 and its incidence is increasing rapidly.2–4 In 2000, the American Society of Clinical Oncology recommended characterization of three protein biomarkers, estrogen receptor (ER), progesterone receptor (PR) and human epidermal growth factor receptor 2 (HER2), as part of the standard of care for breast cancer.5 Tumors with the most favorable prognosis are ER+, PR+ and HER2− and triple-positive, while ER−, PR− and HER2+ and triple-negative tumors have been associated with the worst prognosis.6, 7

Associations between common risk factors for breast cancer, ethnicity and breast cancer subtypes have been examined in many studies7–9; however, few studies have been conducted in Asian populations.10, 11 We therefore studied the association between breast cancer subtype and common risk factors in breast cancer cases in Sarawak, Malaysia. Sarawak has an ethnic mix of people: half of the 2.5 million inhabitants are natives of Borneo, 26% are Chinese and 23% are Malay. Intermarriage among the three communities is not common. The natives comprise 27 ethnic subgroups, Iban (30%), Bidayuh (8%) and Melanau (6%) being the most numerous.12

According to the Sarawak Cancer Registry, breast cancer is the commonest cancer among women in Sarawak.13 Since 1992, the number of cases treated at the Department of Radiotherapy, Oncology and Palliative Care has increased 2.9-fold, reaching 253 cases in 2010. The Department is the only cancer center in Sarawak, and a comparison of our records against that from the Registry indicates that 93% of all cases of breast cancer reported to the Registry are treated at our center. We were therefore able to estimate the age-standardized incidence rates (ASRs) of the breast cancer subtypes in the three ethnic groups of Sarawak.

Material and Methods

Patients and tumor samples

Between 1998 and 2009, 1,963 cases of breast cancer were treated at our Department. The immunohistochemical test for HER2 became available in our State in 2003, and tests for ER, PR and HER2 were conducted on 1,034 cases. The study protocol was approved by the National Ethics Committee, Ministry of Health of Malaysia.

Evaluation of ER, PR and HER2 status

The status of ER, PR and HER2 in all tumors was determined by immunohistochemistry with the Envision Kit with clones 1D5 on formalin-fixed, paraffin-embedded tissue sections in the central pathology laboratory of Sarawak General Hospital. The DAKO classification system was used to interpret HER2 expression. Scores of 0 and 1+ were considered negative; the result was considered positive if the immunohistochemistry score was 3+ or 2+ with gene amplification identified by fluorescence in situ hybridization with the US Food and Drug Administration-approved PathVysion assay (Vysis) and a ratio of HER2 gene: chromosome of 17 = 2.0. All the results were confirmed in the laboratory in Kuala Lumpur where all such tests are performed for the country. The discrepancy rate was <5%.

Although the presence of basal markers can significantly improve the prognostic value of the triple-negative phenotype,14 immunohistochemical tests for CK5/6 and EGFR expression were not available. Hence, we were unable to classify further the triple-negative cases into basal-like and nonbasal-like breast tumors. Therefore, the tumor subtypes analyzed were:

  • Luminal A: ER+, PR+ and HER2−.

  • Triple-positive: ER+, PR+ and HER2+.

  • HER2-overexpressing: ER−, PR− and HER2+.

  • Triple-negative: ER−, PR− and HER2−.

Information on ethnicity, age at diagnosis, age at menarche, age at menopause, parity, duration of breastfeeding, body mass index (BMI), histopathological details, biomarkers (ER, PR and HER2 status), stage at diagnosis and type of treatment received were extracted from case records; data on use of oral contraceptives and hormonal replacement therapy were not obtained. Data were entered with Epidata software15 and submitted to a series of consistency checks.

Statistical analyses

The ASR for each subtype was calculated by dividing the numbers of patients by subtype and 5-year age group by 93% (the proportion seen in our Department) to estimate the number in the full Sarawak population, according to the classical formula,16 standardized against the World standard population.17 The 95% confidence intervals (CIs) were calculated according to the classical method.16 To evaluate associations between risk factors and breast cancer subtypes, chi-square tests and logistic regression were performed with SAS statistical software. The consistency of results across ethnic groups was checked systematically. The case-only odds ratios (ORs) were adjusted for age at diagnosis (as a continuous variable) and ethnic group (coded in two dummy variables). Luminal A, the most frequent subtype, was chosen as the reference.

Results

Incidence of the four breast cancer subtypes in the three ethnic groups

The age-standardized incidence of all breast cancer subtypes in our case series was 22.0/100,000 (95% CI [20.2;23.8]), which is eight times lower than the rate reported among US whites during a comparable period.18 Within our population, the breast cancer rates varied significantly by ethnic group, being higher in Chinese (ASR, 36.7; 95% CI [32.5;40.8]) than in Malays (26.5; [22.1;30.9]) or natives (10.6; [8.8;12.3]).

In Figure 1, the age-standardized incidences of the four breast cancer subtypes in Sarawak are compared with those reported by Lund et al. for whites in the USA,18 showing that the rates for all breast cancer subtypes were higher in US whites than in Sarawak. Interestingly, the USA:Sarawak incidence ratio was highest (9.8:1) for luminal A subtype and lower for the other three breast cancer subtypes (3.7:1 for triple-positive, 2.4:1 for HER2-overexpressing and 2.9:1 for triple-negative cancers).

Figure 1.

Age-standardized incidence rates of the four breast cancer subtypes in Sarawak, Malaysia (this study) and in US whites (Lund et al., 2010).

The age-standardized incidences of the four breast cancer subtypes in the three ethnic groups in Sarawak are presented in Figure 2. The incidence of luminal A was significantly higher in Chinese (ASR, 19.3) and significantly lower in natives (4.8) than in Malays (9.8). The incidences of triple-positive and HER2-overexpressing subtypes were comparable for Chinese (ASR, 4.3 and 4.9) and Malays (4.8 and 3.6) but significantly lower for natives (1.1 and 1). The incidence of triple-negative was significantly lower in natives (ASR, 3.6) than in Malays or Chinese, who had the same incidence (8.1).

Figure 2.

Age-standardized incidence rates of the four breast cancer subtypes in the three ethnic groups of Sarawak. Error bars correspond to 95% CIs. lumA, luminal A; Tr-pos, triple-positive; her2-over, HER2-overexpressing; Tr-neg, triple-negative.

Proportions of the four breast cancer subtypes in the three ethnic groups

The relative distribution of the four breast cancer subtypes differed significantly by ethnic group (p < 0.0001; Fig. 3). The luminal A subtype represented nearly half the total case series and was significantly more frequent in Chinese (55%) than in Malays (38%) or natives (42%; p < 0.0001 and p < 0.002, respectively, when compared with Chinese). The percentage of triple-positive cases was significantly higher in Malay (16%) than in Chinese (11%) or native (10%) patients (p < 0.05 for both when compared with Malays). The percentage of HER2-overexpressing tumors did not vary significantly by ethnic group, while the triple-negative subtype represented nearly one third of the total case series and was significantly more frequent among natives (37%) and Malays (33%) than among Chinese (23%; p < 0.0001 and p < 0.006, respectively).

Figure 3.

Relative distribution of the four breast cancer subtypes in the three ethnic groups of Sarawak. Error bars correspond to 95% CIs. lumA, luminal A; Tr-pos, triple-positive; her2-over, HER2-overexpressing; Tr-neg, triple-negative.

Distribution of risk factors in the three ethnic groups

The distribution of breast cancer risk factors in the three ethnic groups is presented in Table 1. The factors varied significantly by age, BMI, familial history and most reproductive factors. Chinese patients had higher risks for breast cancer than the other two ethnic groups: significantly fewer children (mean number for parous women, 3.36 [3.19;3.52] for Chinese, 4.19 [3.84;4.54] for Malay and 3.64 [3.39;3.89] for native women; p < 0.0001); a shorter duration of breastfeeding (mean duration per child, 5.3 months [4.5;6.2] for Chinese, 10.1 months [8.5;11.7] for Malays and 8.9 months [7.6;10.3] for natives, p < 0.0001); significantly more frequent familial history; and significantly higher age at diagnosis (mean age, 53.1 [52.1;54.0] than Malays: 49.7 [48.4;51.1] and natives: 48.0 [46.6;49.3] (p < 0.0001), so that they were significantly more likely to be postmenopausal (65.5% compared with 51.2% of Malays and 47.3% of natives). The risk profile of native women appeared to be more protective against breast cancer than that of Malays; however, the only difference that reached significance was for parity, native women being significantly less often nulliparous than Malays (p < 0.02).

Table 1. Age, reproductive factors, body mass index (BMI) and tumor characteristics in the three ethnic groups of Sarawak
inline image

The BMI of Chinese patients was significantly lower (mean, 23.9 [23.5;24.2]) than that of Malays (mean, 25.5 [24.6;26.1]) and natives (mean, 25.3 [24.6;25.9]; both p < 0.0001). This difference remained significant for pre- and postmenopausal women. The grade of tumor at presentation was lower for Chinese than native and Malay patients (p < 0.009), and tumors were diagnosed at an earlier stage in Chinese than in natives and Malays (p < 0.0001; see Table 1).

Risk factors and breast cancer subtypes

Cases of each subtype were compared with cases of the luminal A subtype, with adjustment for age and ethnic group, except in models in which age or ethnicity was the main predictor (Table 2). In comparison with women with the luminal A subtype, those with triple-positive tumors were significantly more likely to be Malay, premenopausal and overweight. The multivariate analysis showed that ethnicity, menopausal status and BMI were independently associated with the triple-positive subtype (p < 0.009, < 0.02 and < 0.04, respectively). The OR associated with Malay ethnicity was 1.96 (p < 0.005) after adjustment for BMI and menopausal status.

Table 2. Risks for luminal A subtype by population characteristics and tumor biomarkers
inline image

In comparison with women with the luminal A subtype, those with HER2-overexpressing tumors were significantly more likely to be Malay and to have children. The number of children did not significantly influence the association. An association of borderline significance (p = 0.053) was observed with menopausal status (Table 2). In the multivariate analysis, only parity and menopausal status remained significantly associated with the HER2-overexpressing subtype (p < 0.03 and p < 0.05, respectively), suggesting that some of the difference associated with ethnicity was attributable to these two factors. The OR associated with Malay ethnicity was 1.62 (p = 0.11) after adjustment for parity and menopausal status.

In comparison with women with the luminal A subtype, those with triple-negative tumors were significantly more likely to be Malay or native, to have been older at menopause and less likely to have breastfed for 0–6 months. These results were consistent across ethnic groups. Triple-negative cases were less likely to have a familial history of breast cancer, and this was relatively consistent across ethnic groups (OR for Chinese, 0.40; p < 0.00; OR for Malay and natives, 0.64; p = 0.17).

The Malay and native women with triple-negative tumors were significantly more likely to be overweight before menopause (OR for BMI 25–29 vs. BMI < 25 = 2.02; p < 0.05); the Chinese patients showed less variation in terms of BMI. Chinese women with triple-negative tumors were older than those with luminal A tumors (OR, 1.03 per increasing year; p < 0.008), but no such association was seen for Malay and native cases.

In the multivariate analysis, only ethnicity, age at menopause and family history remained significantly associated with the triple-negative subtype (p < 0.01, < 0.03 and < 0.02, respectively). The associations described above remained statistically significant after adjustment for grade or stage at diagnosis.

Discussion

The ASRs for the three ethnic groups of Sarawak (10.6–36.7/100 000 for all breast cancer) were 5–17 times lower than those reported for blacks and whites in the USA (171.2–179.3/100,000).18 The ASRs in both our study and that of Lund et al.18 were consistently higher for the luminal A subtype, suggesting that a large proportion of the difference in incidence between high and low risk areas and ethnic groups is due to differences in this subtype of breast cancer. While differences in the rates of the other breast cancer subtypes were observed between the USA and Sarawak and between the ethnic groups in Sarawak, the rates were typically lower than those observed for the luminal A subtype, suggesting the rates of the latter subtype are more stable across regions and ethnic groups.

The incidence of breast cancer in Sarawak (22.0/100 000) is lower than that in the USA (176.0/100,000)1 but within the range observed elsewhere in South-East Asia.1 The incidence of breast cancer has been found to increase with a country's level of “westernization,” while women in developing countries have reproductive patterns that are considered protective against breast cancer, such as young age of birth of the first child, numerous pregnancies, long duration of breastfeeding, late age at menarche and early age at menopause.19, 20 In our study, the incidence of breast cancer in Sarawak was low among natives (10.6/100,000), intermediate in Malays (26.5/100,000) and highest in Chinese (36.7/100,000). This pattern is consistent with what is known about these three groups: Chinese tend to be highly educated, live in urban areas and have a high socioeconomic level (a high level of westernization), while natives tend to live in rural areas with a lower socioeconomic level and Malays are intermediate.21

We consider that most of the differences in incidence that we observed are due to differences in risk factor profile; as such, differences should be smaller among breast cancer cases than in the general population. The percentage of triple-negative cases in our series (29%; range, 23–37%) was higher than that reported in other Asian case series: 23% of 1,417 cases in northeastern China,10 18% of 1820 cases in Tian Jin, China22 and 18% in a case series in Kuala Lumpur, Malaysia.23 High percentages of triple-negative cases were, however, reported in two studies in other developing countries where the breast cancer incidence is low: Huo et al.24 reported 55% triple-negative tumors in a study in Senegal and Nigeria, where the breast cancer incidence is estimated to be 37.6/100,000.1 In Saudi Arabia, where the estimated breast cancer incidence is 22.4/100,000,1 Al-Tamimi et al.25 reported 39% triple-negative cases. These results and ours suggest that high proportions of triple-negative cases are associated with a low level of westernization and low breast cancer incidence, the two being correlated.

Our results on incidence, with the results of Lund et al.,18 strengthen this view. Figure 1 illustrates the expected trend for increasing breast cancer incidence with degree of westernization but also shows that the increase is more marked for the luminal A subtype than for triple-negative tumors. This suggests that the increase in breast cancer incidence correlated with westernization is attributable mainly to an increase in the luminal A subtype, with only a small contribution of the triple-negative subtype.

In our case series, the percentage of HER2-positive cases (triple-positive plus HER2-overexpressing) was high (23%) and was significantly higher in Malays than in Chinese or natives. The difference is due to the high percentage of triple-positive cases among Malays, as the percentage of the HER2-overexpressing subtype was similar in all ethnic groups. A high percentage (25%) of HER2-positive cases is considered characteristic of Asian case series.10, 22, 23, 26 In a population-based study of Asian patients in the USA, Telli et al.27 found that four of the seven distinct Asian–American populations had significantly higher proportions of HER2-positive tumors than white Americans, with 19% among Japanese–Americans, 31% for Filipina–Americans and 36% for Korean–Americans. As Telli et al.27 did not calculate ASRs, we cannot compare the incidences in the three Sarawak ethnic groups with those of Asian–Americans. Nevertheless, the ASR of 5.6/100,000 indicates that the incidence of the HER2-positive subtype in our population is much lower than those in white and black US populations (20.1/100,000 and 22.1/100,000, respectively).18

Studies performed to date do not indicate that reproductive factors explain the high proportion of HER2-positive tumors in Asian populations, perhaps because of lack of information on time since last childbirth. In the largest case–case comparisons to date, with 1,657 triple-positive and 953 HER2-overexpressing cases from 34 studies, no association was found between HER2-positive subtypes and age at menarche, parity, age at first full-term birth, family history or BMI.28 Telli et al.27 concluded that in Asian populations “presumably ethnic specific genetic factors exist that influence the expression of the HER2-positive phenotype.” Our finding of significant differences in the proportions of HER2-positive cases within Asian ethnic subgroups (Malays having significantly more HER2-positive tumors than Chinese or native women) is an interesting aspect for further research. We are collecting information on time since last childbirth for this purpose.

A potential limitation of the study was lack of information on some well-established risk factors for breast cancer, i.e., age at first pregnancy, alcohol consumption, use of oral contraceptives and hormone replacement therapy. Information on these factors is being collected in our ongoing prospective study, and the preliminary results show that alcohol consumption, use of oral contraceptives and hormone replacement therapy are low in Sarawak. Thus, the main shortcoming of the study reported here is lack of information on age at first pregnancy. Another potential limitation is that the study is based on cases from a single institution. While cases seen at our institution represent 93% of all breast cancer cases reported to the Sarawak Cancer Registry during the study period, it is possible that not all breast cancer cases that occurred in Sarawak are reported to the Registry.

Case–case comparisons of subtypes are useful tool for exploring etiological heterogeneity among breast cancer subtypes.29 The results of these comparisons in our series were consistent with those reported for larger case series in western countries,28, 30–33 with some differences. In studies of western case series, women with triple-negative subtype tended to be younger than those with luminal A subtype.30, 31 This trend was not observed in our Malay and native cases, and an opposite trend was observed for Chinese women, as also seen in 1,028 cases in Taiwan, where women under 50 years of age had significantly more luminal A tumors and fewer triple-negative tumors than women over 50 years of age.34 In agreement with those authors, we consider that this trend is attributable to a cohort effect: as reproductive patterns in Asia are changing rapidly, young patients are expected to have a greater risk for the luminal A subtype (fewer pregnancies, late age at first birth and shorter duration of breastfeeding) than older patients. It would be interesting to test this hypothesis in case series elsewhere than in Asia.

In our series, patients with triple-negative tumors were significantly older at menopause than those with the luminal A subtype, and this was consistent across ethnic groups. To our knowledge, the association between subtype and age at menopause has been studied in only two samples in the USA, where no association was found.32, 33 Our patients were therefore significantly more likely to be premenopausal than those with luminal A tumors. This association has been addressed in only two studies,30, 31 neither of which reported a significant association, even though their samples were larger than ours (294 and 137 triple-negative tumors, respectively, compared with 125 in our case series). Further studies in Asian populations are needed to confirm this association.

Our patients with triple-negative tumors were less likely to have breastfed than those with luminal A tumors, and the association was consistent across ethnic groups. This finding is consistent with other reports; absence of breastfeeding is a recognized risk factor for triple-negative breast cancer.7 Women of all ethnic groups with triple-negative tumors were less likely than those with the luminal A subtype to have a family history of breast cancer. This result differs from those published to date, as the triple-negative subtype has been associated with a more frequent22, 35 or equivalent familial history.30, 31 Reporting of family history might have been biased in our series owing to the stigma, especially in lower socioeconomic classes, associated with a diagnosis of breast cancer. Thus, socioeconomic level might be a confounding factor, as suggested previously.36

Our patients with triple-positive tumors were significantly more likely to be overweight than those with luminal A tumors, especially when they were premenopausal. The case–case association between BMI and triple-positive subtype has been explored in three studies with high statistical power,28, 30, 31 none of which showed a significant correlation, although some trends were found. Further studies are needed to confirm or refute this association.

Malays had twice the risk of having a triple-positive tumor as Chinese or native women, and this was not explained by a difference in the distribution of the risk factors studied. This result supports the hypothesis of a specific genetic susceptibility to the triple-positive subtype.27

Our sample of HER2-overexpressing tumors was as large as those studied to date (N = 117, vs. N = 135 in the study of Yang et al.28 and N = 116 in that of Milikan et al.31). In our series, women in all three ethnic groups with HER2-overexpressing tumors were significantly more likely to be parous than those with luminal A tumors. This association has not been observed in previous studies. In the largest sample to date (6,661 cases of luminal A and 135 of HER2-overexpressing tumors in 15 studies), Yang et al. did not find an association.28 Kwan et al.,30 however, found a nonsignificant trend similar to ours, which reached significance in the subgroup of women who had never breastfed. As few women in our sample had never breastfed, we did not have the statistical power to perform a similar test. Women with HER2-overexpressing tumors were more likely to be premenopausal than those with luminal A tumors, especially after adjustment for parity. This was not observed in the two large studies of the relation between menopausal status and HER2-overexpressing subtype. In the USA, Milikan et al.31 found that patients with HER2-overexpressing tumors were more likely to be postmenopausal, while Kwan et al.30 found no association. No adjustment for parity was made in either study. The association between parity, menopausal status and HER2-overexpression is complex and should be studied further in larger samples.

Overall, our results and those in the literature regarding the risk factors for breast cancer subtypes support the idea that the risks for the different subtypes vary markedly, and that differences in the distribution of risk factors may explain much of the difference in the proportions of subtypes in different populations. Factors that increase the risk for luminal A breast cancer include low parity, older age at birth of first child, absence of breastfeeding and early menarche,28, 31 which are more prevalent in westernized populations. In contrast, factors that increase the risk for triple-negative breast cancer include some that are characteristic of developing countries (high parity and young age at birth of first child) and some that are not (absence of breastfeeding, high BMI).28, 31, 35 These findings are consistent with our hypothesis that westernization results in an increased incidence that is more marked for the luminal A than for the triple-negative subtype. More studies are needed to investigate the effect on HER2-positive cancers.

Acknowledgements

The authors thank Ms Danielle Surita and Mr. Fariz for assistance in obtaining full texts of articles. For data extraction and data entry, they thank Mr. KC Yeoh, Ms. Patricia Ng and Ms. Zurina Wahab.

Ancillary