Age, tumour size, axillary node status, histological tumour type and standardized pathological grade are accepted as well-defined prognostic factors in breast cancer 1 . Various studies have also reported striking associations between overweight or obesity at breast cancer diagnosis and poorer prognosis with higher distant recurrence and mortality (for review 2 ). As emphasized by Goodwin et al. 3 , the risk of recurrence and death was respectively 1.78 (95% CI: 1.50-2.11) and 1.36 (95% CI: 1.19-1.55) times greater for obese patients over a 10 years follow-up period.

Moreover, numerous studies reported a weight gain after breast cancer development that might be attributable to the effects of some treatment regimens 4 5 . Weight gain in breast cancer patients has been associated with anti-neoplastic chemotherapy in the majority of studies. Previous studies suggest that weight gain is more pronounced among premenopausal women and among those who were treated with a multiagent regimen 6 7 . However, few reports have not observed increased weight gain during chemotherapy 5 8 9 , in particular with anthracycline-containing regimens 10 11 widely recognized as the gold standard treatment of women with breast cancer.

There is also substantial evidence that weight change during chemotherapy may be associated with a worse prognosis for the cancer patient, both with weight gain 12 and weight loss 13 . The findings reported by the few studies which have explored the prognostic value of weight gain after a diagnosis of breast cancer are mixed: four studies reported that weight gain was associated with a decreased overall survival and increased recurrence risk 8 14 15 16 whereas five others failed to report such associations 10 9 17 . Only one recent study reported some evidence that women with early breast cancer, who had a weight loss during treatment, were at higher risk of recurrence and death compared to women with no weight variation 18 . These discrepancies may be attributable to the heterogeneity of the methods implemented in the different studies, including the duration of post-diagnosis weight assessment, the definition of the prognostic outcomes with a short median of follow-up and treatment (chemotherapy and/or hormonotherapy...). The majority of the previous studies were conducted before anthracycline-base chemotherapy was commonly used. Additionally all the aforementioned studies focused on north American populations with a mean Body Mass Index (BMI) at breast cancer diagnosis classified as overweight 14 19 . However no data is available on the prognostic impact of weight change during chemotherapy treatment in European breast cancer patients who presented a notably lower BMI 20 .

Our study thus investigated the prognostic value (death and recurrence) of weight variation during anthracycline-based chemotherapy treatment of breast cancer in a French population with a long-term follow-up. We also verified the association of weight at breast cancer diagnosis with survival.

In this study, we demonstrate a relationship between weight variation during polychemotherapy treatment and both poorer disease-free survival and overall survival after diagnosis of breast cancer.

The present study is in agreement with previous studies which have found that overweight at the time of diagnosis increased both breast cancer recurrence and mortality. This result has been largely demonstrated in U.S. populations 25 26 27 . The poorer prognosis of obesity at diagnosis could be explained by the fact that overweight women tend to be diagnosed with later stage cancer and therefore more adverse tumour characteristics than normal weight women 28 . For some authors, this association is restricted to women who detected their own cancer and not spread amongst cases detected by either screening mammography or clinical breast examination 29 . Additionally, lower screening rates may partly explain the higher breast cancer mortality in obese women 30 . In agreement with this, in our population who were diagnosed 20 years ago, when no screening mammography was systematically carried out, we observed a striking association between BMI and tumour size.

In our series however, with a median WV equal to zero during chemotherapy treatment of breast cancer patients, 31% of our population presented a significant weight variation (> 5%) whereas 68% had maintained their weight. Our findings of no significant median WV during chemotherapy treatment are in contrast with the vast majority of studies conducted in North America, which have generally reported an average weight gain ranging from 1.7 to 4.4 kg during the years that follow diagnosis in women treated by chemotherapy 5 31 32 . On the other hand, one study carried out in a Korean breast cancer population has already observed a lack of overall weight gain, with 10.4% of the population gaining more than 5% of baseline body weight at 1 year 31 . Regarding the baseline mean BMI, we can observe that populations who did not display a significant weight gain during chemotherapy treatment, including ours, are leaner (mean BMI of 24.4 kg/m2 in our study, 23.5 kg/m2 in Korean study) than the ones used in the U.S. studies 5 in which the mean BMI varied from to 26.3 to 27.4 kg/m2 18 32 . Moreover, one possible explanation is that the chemotherapy regimen administered to our patients incorporated anthracycline-based therapy as reported by Han and al. 33 , whereas the majority of the earlier previous studies that observed a weight gain, involved non-anthracycline-based regimen. Other retrospective and prospective reports have not demonstrated increased weight gain with anthracycline-containing regimens compared with other regimens 10 11 . Fisher et al. 34 noted that 14.4% of patients receiving treatment with the AC regimen gained ≥ 5% over pretreatment weight compared with 42.2% of patient receiving CMF. This result was comparable to our result (14% gained weight) indeed the fact that patient accrual took place prior to the widespread use of 5-HT₃ receptor antagonists in the two studies (cancer-related treatment currently used to reduce the impact of nausea and emesis associated with anthracycline use).

The long-term follow-up of patients who received an anthracycline-based chemotherapy in this study demonstrates that weight variation may not only influence recurrence, but also patient outcome. In multivariate analysis, clinical node involvement was still significant, thus, weight change was the strongest parameter associated with OS and DFS in our series. Literature on prognostic value of weight variation is mixed and not easily comparable. These studies that generally evaluated post-diagnosis weight variation on different periods which varied from a few months to a few years after diagnosis, including different treatments (chemotherapy and or hormonotherapy, radiation only...), did not use the same prognostic outcomes, and sometimes with a short median of follow-up. Among the few studies which evaluated the prognostic value of weight change after breast cancer diagnosis, four studies have shown a poor prognosis 8 14 15 16 whereas five reported no relationship 9 17 18 19 .

The majority of studies observed an impact of weight gain on patient outcome. The largest study to date included 5,204 Nurses' Health Study participants diagnosed with non-metastatic breast cancer between 1976 and 2000 treated with chemotherapy and/or hormonal therapy 14 . This study reported an increased risk of recurrence, breast cancer death and total mortality in patients who gained more than 2 kg/m² by comparison to patients who maintained their weight. However this relationship was found only among women who never smoked and the definition of recurrence included reported lung, bone or brain cancer, but excluded any local recurrences in the ipsilateral breast or new primaries in the contralateral breast. Camoriano et al. reported weight gain having a significant effect on overall survival during treatment with cyclophosphamide, fluorouracile and prednisolone (CFD) or CFD plus tamoxifen, but not on recurrence and this only for premenopausal women 8 . Two other studies conducted prior the common use of anthracycline-based chemotherapy reported a correlation between weight gain and overall survival and/or disease-free survival (without defining which events were included) 15 16 .

Only one recent study reported some evidence that women with early stage breast cancer treated with chemotherapy and/or radiation and tamoxifen who had large weight loss (> 10%) were at higher risk of recurrence and death compared to women with no weight variation. This elevated risk was more pronounced among women who were obese before diagnosis or who had ER- or PR- tumours 19 . One obvious explanation for significant weight loss being related to an increased risk of death could be that the breast cancer disease process itself caused weight loss.

Literature on the prognostic value of weight variation reported some evidence that women who had gained or lost weight have a higher risk of recurrence and death compared to women with no weight variation. So, we chose to group women who gained weight with those who lost weight as a weight changing group. We hypothesized that weight change reflected a metabolic disorder by comparison to women who maintained their weight with an energy balance in equilibrium (Figure 3). Chemotherapy induced a decrease in energy expenditure (lowered basal metabolic rate, thermogenesis, and physical activity 4 ) and different modifications in dietary intake (increase in appetite 14 or decreased ingestion of food due to chemotherapy related nausea and emesis) that can lead to weight gain or loss according to dietary behaviour of patient. Moreover, women with breast cancer receiving adjuvant chemotherapy underwent unfavourable changes in body composition with lean body loss due to a negative nitrogen balance 35 even in the absence of an overall weight change 4 . A lot of data has demonstrated that weight gain during chemotherapy was indicative of sarcopenic obesity 33 . Indeed, chemotherapy for breast cancer like taxane and anthracycline can increase inflammation 36 37 which played a central role on different modifications induced by chemotherapy. Inflammatory cytokines interfered with the satiety centre 38 and catabolism of skeletal muscle protein responsive of sarcopenia but not independently of any of the considered obesity indexes 39 .

Several mechanisms have been proposed to explain the adverse effect of weight gain on risk of recurrence and mortality. First, weight gain and all associated metabolic disorders may predispose women to diabetes or heart disease, thereby predisposing them to morbidity and mortality 6 . However, in our study, only a few patients died from causes other than breast cancer (14%). One possible mechanism was a greater aromatase activity in the excess adipose tissue 40 and an inhibition of synthesis of sex hormone-binding globuline associated with an increased in free estradiol level which stimulates neoplasic cells 41 .

Insulin resistance may be a common mechanism to explain the poor prognosis of patients who experienced a weight loss or a weight gain (Figure 3). Indeed, insulin resistance has been shown in variety of cancer patients with body-weight loss 42 , but was also seen in overweight women 43 . Fasting serum insulin concentration has been directly associated with an increase in both distant recurrence and death in women previously treated for breast cancer 42 43 . There is a strong biological rational for an adverse prognostic effect of insulin. Insulin, a member of a family of growth factors that includes IGF-I and IGF-II, exerts a mitogenic effect on malignant breast cancer cells though IGF-I receptor. It is also hypothesized that visceral obesity increases both insulin-like growth factors (IGF-I, IGF-II) which stimulates the synthesis of sex steroid hormones 31 that are involved in the regulation of normal and malignant growth of epithelial breast cells. Several studies have reported a reduction in circulating concentration of IGF-I in malignant disease, which may also have been associated with nutritional decline and systemic inflammation 44 . Yoshikawa et al. hypothesized that inflammatory reactions might be involved in the development of insulin resistance 42 . Moreover, few studies reported an association between elevated inflammatory cytokines and a worse prognosis in breast cancer patients 45 46 .

Further research is needed to understand the biological mechanisms underlying the relationship between weight variation and breast cancer growth with exploration of insulin resistance in association with body composition, measurement of energy expenditures, calorie intake and inflammatory reaction.

The current study had several limitations including its study design (retrospective chart review) and a relatively small sample size to draw a conclusion on the independent effect of weight variation. Some interesting covariates like "normal " weight prior diagnosis, smoking status, physical activity or sociodemographic aspects (education) have not been explored because of the lack of this data in patients' medical records. Additionally, more detailed measurement of body shape and fat content are lacking. Currently a long-term measurement of weight during patient follow-up is ongoing and could offer the possibility to explore weight variation after treatment which could also affect prognostic outcomes.

Our results suggest that weight change during anthracycline-based treatment of early stage breast cancer is associated with increased risk of recurrence and poorer survival, though they may require additional confirmation. Furthermore, while we have speculated on potential biological targets, more research is needed to understand the biological mechanisms underlying the relationship between weight variation and breast cancer growth.

WV: weight variation; BMI: Body Mass Index; TNM: tumor-node-metastasis; DFS: disease free survival; OS: overall survival; CI: confidence Interval.

The authors declare that they have no competing interests.

ET contributed to the data collection, analysis and wrote the paper. ST and OL contributed to the data collection and helped in in writing the final draft. FK contributed to the data analysis. EG, CA, EP helped in writing the final draft. PG, MAMR, PC and XD contributed to patient recruitment. PC and XD contributed to the study design and analysis and have been involved in drafting the manuscript. All authors read and approved the final manuscript.