ABSTRACT
Breast cancer (BC) is the second most common cancer worldwide, with over 2,300,000 new cases estimated per year. Diet has been identified as a modifiable risk factor for BC development and prognosis. The Mediterranean diet (MD) has shown to be inversely associated with chronic diseases including BC. The aim of the present study was to assess dietary patterns according to BC molecular subtypes in a subgroup of patients at their baseline visit of a lifestyle trial conducted by our institute. A principal component analysis (PCA) was conducted to assess the best dimensional space where to summarize dietary information. An explorative unsupervised automatic clustering technique was performed to identify diet-risks groups. Final groups were analyzed as dietary patterns and comparisons made by synthetic statistics with univariable analysis. The first PCA factor was characterized mainly by vegetables (27.6%), nuts & extra-virgin olive oil (EVOO) (16.7%), and sweet & sugars (11.8%). Legumes and fats separately represented just over 10% of the first PCA factor. The second factorial axis was represented mainly by cereals (40.9%), sweet & sugars (20.2%) and nuts & EVOO (15.4%). PCA showed different behaviors between dietary variables in each molecular subtype, especially among patients with triple negative TNBC (n = 37) the strongest contribution to the first PCA factor was given by sweet & sugars (20.7%), then vegetables (17.1%), fruits (11.9%) and legumes (11.0%) while animal proteins (24.3%), nuts & EVOO (16.0%), fruits (14.1%) and fish (12.1%) determined the second factor. From k- means, three clusters of patients were found. Cluster 1 (Healthy pattern) was associated with healthier dietary habits compared with the other two groups, with approximately twice the vegetables (204 grams vs. 119 grams for cluster 2 and 95 grams for cluster 3, p < 0.01). Cluster 2 (Western pattern) was characterized by greater refined cereals and animal protein, sedentary behavior and higher body mass index (BMI) and central obesity (35% with ≥30 kg/m2 compared with 13% in cluster 1 and 25% in cluster 3. Cluster 3 (Ultraprocessed pattern) was characterized by greater intakes of sweet & sugars and non-EVOO fat, cluster 2 was composed mainly of Luminal BC subtype, while TNBC were found mostly in cluster 1 and cluster 3. Our findings revealed three main dietary risk group by BC patients at the baseline visit of a lifestyle Trial: a healthy dietary group (cluster 1), a western diet group (cluster 2) and an ultra-processed food diet group (cluster 3). The former is considered part of a healthy Mediterranean diet which is known to improve the metabolic and hormonal risk factors for BC and reduce total mortality. A concern emerged for the high-risk group of TNBC patients who tend to be younger and appeared to consume more sweets and fats which are known risk factors for chronic diseases and poor cancer prognosis.

IMPACT STATEMENT 
Different dietary patterns emerged according to Breast Cancer molecular subtypes through a principal component analysis (PCA) that analyzed dietary information in a lifestyle trial conducted by our institute; moreover, an explorative unsupervised automatic clustering technique was performed to identify diet-risks groups.