In this large cohort study, we observed a positive association between more pro-inflammatory diets and an increased risk of breast cancer, more pronounced in premenopausal women. Overall, women with the highest pro-inflammatory diet (fifth quintile of the ISD) had a significant increased risk of 12% compared with those with the most anti-inflammatory diet (first quintile). Each increase in 1 Sd of the index had a significant increased risk of 4%; rising to 8% among premenopausal women. This finding is particularly relevant for BC prevention since diet together with physical activity and weight control are key modifiable lifestyle factors, and BC is the most common cancer in women, with over 2 million new cases in 2018, and the leading cause of cancer death worldwide [22]. It is also worth noting that so far, no single dietary component apart from alcohol has been found to be a cause BC with convincing degree of evidence [4]. On the contrary, looking at the totality of diet, as it is done by means of dietary patterns, it is likely to reflect an interactive, synergistic and combined effect of dietary components [23]. Moreover, examination of diet as a whole can be more readily translated into dietary guidelines. In our population, a more anti-inflammatory diet is defined by a high consumption of legumes, vegetables, fruits (all kinds), and to a lesser extent, fruit and vegetable juices, coffee, and tea, as reflected by strong inverse correlation of these food group with ISD (Table S3). On the contrary, a more pro-inflammatory diet is characterized by high consumption of meat and meat products (including red and processed meat), foods rich in fats and oils, and sugar and confectionery.
To our knowledge the association between the inflammatory potential of the diet and BC risk has been assessed in six prospective studies. Our results are in line with those from the Swedish Women’s Lifestyle Health [9], in which there was significant increase of 4% of risk for each increase of one unit of the DII, as well as in the Iowa Women’s Health study [13], with a marginally significant increased risk of BC of 11% for women in the highest tertile of the DII. The latter reported a significant interaction with BMI; the only significant increase in risk was observed among obese women. No association between DII and BC risk in postmenopausal women was found in the Women’s Health Initiative [10], but an extended follow-up of the same study [11] reported a significant increased risk for women in the highest quintile of the DII limited to cases ER + PR + HER2+. The authors acknowledged that it is no clear why a diet with high inflammatory potential would be associated with this specific subtype of BC. No association were found in small cohorts in France [12] and Spain [15]; but with regard to the small sample size (158 and 100 BC cases respectively) both studies had little statistical power. The French study [12] reported a significant interaction with alcohol intake: DII was associated with increased BC risk in low-moderate drinkers but had a protective effect among heavier drinkers. According to authors the latter is unlikely to be causal. In this study the DII included alcohol intake and it is unclear how this may have affected the results. Finally, no association between DII and BC risk was observed in the Sister Study cohort [14] but in subgroup analyses significant associations were reported for postmenopausal women, obese women, and ER-PR- BC cases. Unfortunately, further details are not available since results of this study have been published only as an abstract.
All of the above-mentioned studies assessed the inflammatory potential of the diet by means of the DII, whereas we used the ISD. However, the results are comparable as the two indices are quite similar. Actually, for the calculation of the ISD we used the set of weights (inflammatory scores) proposed to calculate the DII. The major difference with respect to the DII was that the intake of each food items was standardized using the mean and standard deviation of the EPIC population instead of those from a regional worldwide database [20]. Furthermore, the Pearson’s correlation coefficient between the ISD and the DII in the EPIC population was 0.91 (P-value < 0.001) [19].
In this work we used a version of the ISD excluding alcohol based on two main considerations. First, although ethanol is considered to be anti-inflammatory in the original DII [20] it seems that alcohol has a dose-dependent effect. The negative relationship with inflammatory markers has been observed only among moderate alcohol consumers suggesting that the presence of other bioactive components in alcoholic beverages rather than ethanol itself may provide anti-inflammatory properties [24, 25]. Second, and even more relevant, is that alcohol is a well-established cause of breast cancer [3, 4]. If a negative association of an anti-inflammatory diet is found, recommendations for BC prevention based on our results would never include the consumption of alcohol. We used the same approach when we assessed the association of BC with the adherence to a Mediterranean diet [26]. Anyway, it is also reassuring that a significant association between the ISD and BC risk was independent of the level of alcohol consumption (Table 3).
Hormones play an important role in BC risk and progression. There is a consistent link between postmenopausal concentrations of endogenous hormones (mainly oestradiol and testosterone) and increased BC risk. There seems to be a similar pattern in premenopausal women, but data are sparser [22]. On the other hand, adiposity and physical activity are both associated with chronic inflammation, which could partially explain the association of these factors with BC. While a state of low-grade chronic inflammation is induced by changes in the pathophysiology of adipokines of obese subjects [27], physical activity may reduce the macrophage production of inflammatory cytokines [28]. We have observed that the association of ISD with BC risk was particularly marked among premenopausal women and showed a consistent (and significant) association among inactive women and those with normal weight. Our results are compatible with the hypothesis that the potential effects of a pro- or anti-inflammatory diet are stronger, or at least more evident, among women for which hormonal pathways are less relevant and those without other strong determinants of systemic chronic inflammation.
A limitation of the present study is that we have only one assessment of diet carried out at recruitment, derived from self-reported information relying on subjects’ memory. Therefore, measurement error cannot be discarded. However, since information was gathered before the identification of cases, such error would be non-differential, and hence it would most likely result in a decrease in true association. Major strengths of this study are the prospective design and its large sample size, allowing sufficient statistical power for subgroup analyses. It is now widely accepted that the factors that modify the risk of BC are not the same when diagnosed before or after the menopause. On the other hand, the importance of distinguishing tumour subtypes according to hormone receptors when evaluation aetiology is now well established [29]. Therefore, the ability to assess within a common framework the associations between ISD and BC risk overall, as well as by menopausal status and tumour receptor status is an advantage.
In conclusion, our findings suggest that a more pro-inflammatory diet is associated with an increased risk of breast cancer, especially among premenopausal women. These results could help provide dietary recommendations, although they require further confirmation, for the prevention of breast cancer. In this line, it may be of interest to study new hypotheses regarding the possible effect of the inflammatory potential of the diet and the progression and prognosis of breast cancer.