Obesity is generally accepted as an adverse prognostic factor in breast cancer. A meta-analysis of 82 studies reported an increased risk of breast cancer mortality, hazard ratio 1.35 (95% limits 1.24-1.47) for ‘obese’ women (body mass index (BMI) 30+) compared to those with a ‘normal’ BMI (18.5 to 25) [1], seen in both pre- and post-menopausal women. This meta-analysis showed significant publication bias, suggesting that some small studies with null or inverse results have not been published. Many studies are based on incomplete or selective data: for example, one of the largest studies excluded 65% of otherwise eligible patients as they had no data on BMI recorded [2].

Several mechanisms have been suggested by which obesity could affect breast cancer prognosis; biological mechanisms influencing tumour progression; interactions with therapies; and health care-related issues affecting treatment and diagnosis.

Obesity is associated with elevated levels of serum oestrogen, produced by conversion of androgens by aromatase in adipose fat [3], and lower levels of sex hormone-binding globulin, which lowers oestrogenic activity [4]. Obesity is associated with higher levels of insulin and the adipocyte derived cytokine leptin [5] and could have effects related to markers of inflammation [6]. Effects through these mechanisms would be expected to be greater in post-menopausal women; however, in the meta-analysis noted no difference in effects by menopausal status was seen [1]. Breast cancer patients who are obese have been shown to have greater expression of proliferation genes [7], and faster growing tumours as assessed by Ki-67 [8].

These hormonal-based mechanisms suggest that anti-oestrogenic therapy might be of greater benefit to obese women. This has not been shown for tamoxifen [9], but a greater benefit from raloxifene in women with higher BMI has been suggested [10]. Obese women may have a reduced response to aromatase inhibitors [11, 12]. While the efficacy of full doses of chemotherapy does not appear to be affected by obesity [9, 13], obese women are likely to receive sub-optimal dosages of chemotherapy [14–16]. In one study in a patient population with a high prevalence of obesity, practice standards to avoid under-dosing are suggested as the reason why no effects of BMI on outcomes were seen [17].

Obese women may be disadvantaged at diagnosis; they may have larger primary tumours, more positive lymph nodes, and more advanced stage [18], and they may be less likely to be diagnosed by screening [19]. The association between BMI and breast cancer outcome may vary in women of different ethnic groups [20]. A stronger adverse effect of obesity on breast cancer survival in women of Asian ancestry has been shown in some studies [21, 22].

In this study, we assessed associations between breast cancer-specific and overall survival, and recurrence, with BMI in a large population-based cohort of women with breast cancer in New Zealand (NZ). Patients were diagnosed between January 2000 and June 2014 and followed to last follow up, death, or Dec 2016 or to death; median follow up 4.1 years. We restricted the main analysis to the 1049 patients with stage 1 to 4 breast cancer who received chemotherapy as part of their primary treatment; 98% had data on BMI collected after diagnosis but before systemic treatment. We were able to take into account age, menopausal status, ethnicity, social deprivation, comorbidity, mode of diagnosis, staging, grade, and receptor status, and primary treatment. We also assessed the outcomes in all 2296 patients, irrespective of treatment, who had known BMI data.

This group of breast cancer patients treated with chemotherapy have a high prevalence of obesity (BMI over 30), with 35% having a BMI of 30 or greater, and 14% a BMI of 35 or greater. There was no increase in breast cancer mortality or in total mortality even in very obese women; the hazard ratio (HR) for women with a BMI of 35 or over, compared to those with BMI of 21-24, for breast cancer mortality was 0.96, with 95% confidence limits of 0.56 to 1.67, adjusted for other demographic and clinical factors; and for overall mortality, the adjusted HR was 1.03 (95% limits 0.63 to 1.67). There was no indication of a dose-response trend, either in the main analysis or in subgroup analyses, or when assessing BMI as a continuous variable throughout its range.

These results contrast with many other studies. A meta-analysis of 82 studies showed an increased risk of breast cancer mortality, HR 1.35 (95% limits 1.24-1.47) for ‘obese’ women with a BMI over 30 compared to ’normal weight’ women with a BMI between 18.5 and 25 [1]. There was a slightly greater increase in total mortality, HR 1.41 (95% limits 1.29-1.53), due mainly to a substantial, although non-significant, increase in cardiovascular mortality (HR 1.60, 95% limits 0.66-3.87). However, as noted earlier, the meta-analysis had evidence of publication bias, with Egger’s test being significant (P=0.03) for the studies of breast cancer mortality; the authors suggested that “small studies with inverse results are missing”.

Our results suggest that the disadvantageous prognostic effect of obesity reported elsewhere does not apply to this breast cancer population treated with chemotherapy in New Zealand. There is no clear explanation, apart from chance variation, for the contrast between these results and the results of the meta-analysis. Our data were 98% complete and based on objective clinical records after diagnosis and prior to systemic therapy. In the meta-analysis [1] differences in outcomes between obese and normal-weight women were similar for BMI assessed pre-diagnosis and within 12 months after diagnosis. However, the excess risk in underweight women was greater with post-diagnosis assessment. In the meta-analysis [1], the association of BMI with total mortality was stronger in pre-menopausal women (RR 1.75, limits 1.26-2.41) than in post-menopausal women (RR 1.34, limits 1.26-2.41), but this heterogeneity was not statistically significant. Restriction of the data to invasive, early-stage, or mammographically detected cases made little difference to the results. In our analyses, we found no associations within groups defined by menopausal status, stage, method of detection, or other clinical or demographic variables.

The main analysis presented is based on a continuous population-based clinical registry, but then restricted to patients receiving chemotherapy. For these patients information on height and weight was fully recorded, with 98% completeness. The assessment was after diagnosis and before systemic therapy was started. We are cautious about the interpretation of results for patients who did not receive chemotherapy, as there is substantial missing data, but the results were similar, not showing any association of breast cancer mortality or total mortality with BMI. However, we have shown that the missing data is not random, and is associated with survival outcomes [35]. Another New Zealand study is hard to interpret as it is based on only 27% of eligible patients because BMI data was not available on the others [36].

Some other studies have also shown no association with BMI. A study in Louisiana of 523 patients, not selected on treatment, of whom 55% were obese (BMI > 30), showed no association with overall or disease-specific survival, with a median follow-up of 49 months [17]. The authors suggested that with the high prevalence of obesity in these centres, clinicians would be more expert in dealing with obese patients and less likely, for example, to undertreat with chemotherapy. That would also apply in our population, where obesity is a prominent and familiar issue. A large randomised trial, the NSABP B-14 trial, showed no associations of breast cancer mortality with BMI [9], and under-treatment may be less likely in a trial. This trial of tamoxifen assessed 3385 women with node-negative, ER positive breast cancer, with a median follow up of 166 months. Obese women did have a higher risk of contralateral breast cancer incidence, other cancer incidence, deaths from causes other than breast cancer, and total mortality. In an analysis of 489 patients in three randomised trials of chemotherapy for metastatic breast cancer, there was no association between BMI and progression-free or overall survival [37]. Obese patients had a significantly improved progression-free survival in a study restricted to women receiving upcapped doses of chemotherapy [38]. A recent study showed more advanced staging in obese patients, but no significant effect on survival [39]. The effects of obesity on survival may only apply to certain subgroups. Thus, a study showed no overall effect on survival or recurrence, although an adverse outcome was seen in the subset of luminal A cancers [40]. In another study, obesity was associated with lower survival only in receptor positive tumours with positive lymph nodes, while it was associated with improved survival in receptor negative tumours [41]. A study of nearly 15,000 patients with pre-diagnostic BMI data showed no associations between recurrence or mortality in overweight women, and only a 10% increase in risks in obese or in severely obese women [42]. However, several more recent randomised trials of chemotherapy have shown poorer outcomes in obese women; for example [43–47].

BMI is a convenient and widely used measurement, but other assessments of body size may be relevant. A study of over 90,000 patients, assessed pre-diagnosis, showed that mortality was related to greater triceps skin fold thickness, but not to BMI [48]. In another study of post-menopausal patients, weight, but not BMI, was related to mortality [49]. Central obesity seems to have a stronger impact on African American women than general adiposity as measured by BMI [20].

BMI is related to many other lifestyle and dietary issues. In a study of 9513 breast cancer patients, mortality was related to low physical activity and to comorbidity, but not to BMI [50], although in another study BMI but not physical activity was related to increased breast cancer deaths [51]. In studies using pre-diagnosis dietary data, mortality increased regularly with intake of fat [52] in one study, and in another it increased with dietary saturated fat, and decreased with higher beta-carotene and vitamin A intakes [53].

Our study population has a substantial proportion of women of Maori and Pacific Island ethnicity (2013 census 22% and 4% respectively), the proportions in patients treated with chemotherapy being 19% and 3%; but the outcomes specific to these groups and to the majority non-Maori non-Pacific group did not show associations with BMI. A stronger effect of BMI on survival in breast cancer has been suggested for women with Asian ancestry, while effects in African Americans and Hispanics were similar to non-Hispanic whites [20], based mainly on US data. Our study population had a substantial frequency of comorbidity, including diabetes, and diabetes has been shown to be associated with a lower overall survival in breast cancer patients [54]; in our study, hazard ratios were not changed substantially by controlling for comorbidity, but the numbers were too small to assess effects specifically in those with diabetes.

In summary, in a population based series of women with breast cancer in New Zealand, with 35% having a BMI of 30 or greater, no association between BMI and overall or breast cancer-specific survival, or disease-free survival, was seen in patients receiving chemotherapy, with 98% complete data on BMI; nor in all patients, irrespective of treatment, although for non-chemotherapy patients there was substantial missing data. We have no clear explanation, apart from chance variation, as to why our results differ from some other studies, but it is important to document this difference, particularly as previous meta-analyses have shown some publication bias.