The Malmö Diet and Cancer Study (MDCS)
The study population consists of women enrolled in the MDCS, which is a prospective population-based cohort study in Malmö, Sweden. The MDCS and the baseline investigations have been described in detail elsewhere[15,16]. Briefly, the baseline examination took place between 1991-1996, and include a dietary assessment, blood samples and a self-administered questionnaire. The questionnaire included questions regarding socioeconomic status, medical history, lifestyle habits and for women also menopausal status and reproductive history. Height, weight, body composition and blood pressure were assessed by physical examination.
Women born between 1923-1950 were invited to participate. Consequently, this resulted in a total female cohort of 17 035 women, representing a participation rate of 43% for women.
Identification of breast cancer cases and controls
By record linkage with the Swedish Cancer Registry, breast cancer cases until December 31st 2013 were identified. Women diagnosed with breast cancer prior to inclusion in the study were excluded (n=576), resulting in a total of 1186 eligible incident cases. Using two different selection methods, an identical number of controls (n=1186) were included in the study. Approximately half of the controls were chosen based on a previous study by Almquist et al. (2010) where incidence density matching was used in order to match each case to a subject at risk at the time of case occurrence. Using age as a time scale, controls were matched on menopausal status, time of inclusion and age. Among the matched controls, 694 remained breast cancer free until December 31st, 2013. To make it an equal number of cases and controls, the remaining controls (n=492) were selected from a randomized subsample of the MDCS, the cardiovascular cohort. The purpose of choosing the cardiovascular cohort was that genetic material was available as needed in a parallel future study.
As a consequence of missing tumor material, 63 patients were excluded from risk analyses of tumor characteristics. In addition, twenty cases had bilateral breast cancer and were excluded due to problems in deciding which side to be used in analyses of histopathology, receptor status, lymph node status and tumor size. Further, 100 cases were excluded from the risk analyses of tumor characteristics as a result of tumor data showing carcinoma in situ. A flowchart, adapted from Sandsveden et al.[19,20], of patient inclusion and exclusion in the study is illustrated in Figure 1.
Blood sample collection and laboratory analysis
Venipuncture was done on non-fasting participants at baseline. Serum was extracted within one hour of blood sample collection and afterwards stored at −80°C. In October 2015, the saved serum zinc was analyzed by ALS Scandinavia AB, Luleå, Sweden, as previously described by Sandsveden et al. Serum samples were analyzed on ICP-SFMS (Thermo Element 2) using single-element standard solutions, NIST, traceable to the International System of Units. An amount of 0.15 mL was mixed with an alkaline liquid containing 0.1% ammonia and 0.005% EDTA/Triton-X to at quantity of 10 mL. Seronorm, obtained from Sero AS, Norway (Lot 0608414), was analyzed together with the serum samples as a reference material. The detection limit of zinc was 10 ng/mL, and the inter-assay coefficient of variation was 3.3%. Albumin had previosly been analysed as part of antoher study.
Due to insufficient amount of saved serum from 262 women, these women were reported as having missing serum zinc data and were consequently excluded from the analyses using serum zinc as an indicator of zinc status.
Dietary assessment method
The methodology used in The MDCS have good ranking compared to a reference method consisting of 18 days of weighed food records. As previously described more in detail[20,22], it consists of three parts: a) a 168-item semi quantitative diet history questionnaire gathering information about the overall meal pattern e.g. potion-size and frequency of foods consumed regularly b) a 7-day food diary for registration of cooked meals, beverages, nutrient supplements, pharmaceutical drugs and natural remedies c) a 45-60 minutes diet history interview where portion sizes and cooking preparations in the questionnaire and menu book were described more in detail. Based on the estimate of portion sizes and frequencies from the questionnaire and the food diary, a mean daily intake of foods was determined. To translate the food intake to nutrient and energy intake the PCKost2-93 from the National Food Administration in Uppsala, Sweden, was used. PCKost2-93 contains roughly 1600 basic foods with additional food codes and recipes added specially for the MDCS. In the present study, the sum of food intake and supplemental intake of zinc are expressed as dietary intake of zinc.
Data on tumor characteristics in the identified breast tumors was collected in three separate time periods. 1) Tumor material from cases diagnosed until 31st December 2004, as described in two previous studies[23,24], were re-evaluated by a senior breast pathologist concerning the histopathological diagnosis, i.e. histological type in agreement with The World Health Organization classification guidelines and histological grade according to Elston and Ellis. In addition, micro array (TMA) was constructed in order to re-evaluate proliferation (Ki67), human epidermal growth factor 2 (HER2)- and hormone receptor status. 2) Likewise, for cases diagnosed from 2005 to December 31, 2007, TMA was used to re-evaluate Ki67, HER2- and hormone receptor status. However, histological grade and type were gathered from medical records and pathology reports. 3) From December 31, 2007, information about tumor characteristics was exclusively collected from medical records. During all periods, data on tumor size and nodal status were gathered from hospital records.
HER2 status was gathered either from immunohistochemical (IHC) score (1991-2004), or from regional and national cancer registries and hospital records which comprised analysis data from both in situ hybridization (ISH) and IHC (2005-2013). Breast tumors were considered HER2 positive when ISH was amplified or when scored as 3+ on the IHC staining. HER2 was regarded as negative when ISH was not amplified or when scored as 0 or 1+ on the IHC staining. Breast tumors with a score of 2+ was classified as missing if no data from ISH amplification was available. Tumors were divided, based on the expression of Ki67, into tertiles (low, intermediate or high) separately for each time period 1991-2004, 2005-2007 and 2008-2013. ER- and progesterone receptor (PgR) status were evaluated according to the nucleus expression of ER and PgR and tumors were dichotomized into negative (≤10% staining intensity) and positive (≥10% staining intensity).
Breast cancer surrogate intrinsic subtypes classification
Based on ER, PR, and HER2 receptor status along with histologic grade and Ki67 positivity, surrogate intrinsic subtypes were created[19,29]. Tumours were divided into Luminal A-like (ER+ and HER2- with (a) grade 1 or (b) grade 2 and low Ki67 or (c) grade 2, intermediate Ki67 and PgR+), Luminal B-like (ER+ and HER2- with (a) grade 3 or (b) grade 2 and high Ki67 or (c) grade 2 intermediate Ki67 and PgR-), HER2-positive (all tumors classified as HER2 positive), or triple-negative (TNBC) (all tumors that were ER-, PgR- and HER2-).
Dietary intake of zinc was adjusted for total energy intake using the residual model. The residual is the difference between the actual zinc intake and the predicted zinc intake in a regression model with total energy intake as the independent variable and absolute zinc intake as the dependent variable.
Subsequently, the study population was divided into groups according to quartiles (Q) of serum zinc levels and dietary intake of zinc. Quartile cut-offs were based of the distribution of all women in our study, both cases and controls. Quartiles of residuals in our study are presented as the median of total dietary intake of zinc.
To measure the level of agreement between serum zinc and dietary intake of zinc, Cohen’s Kappa coefficient with a p-value was calculated from a cross table of serum zinc and dietary intake of zinc quartiles.
Logistic regression was used to estimate odds ratios (ORs) and 95% confidence intervals (CIs) for breast cancer and having a certain tumor characteristic in different quartiles of serum zinc and zinc intake. Thereafter, the same analyses were done for dichotomized groups of serum zinc and zinc intake, defined as low (a merge of Q1 and Q2) and high (a merge of Q3 and Q4), and for groups combining low and high serum zinc and dietary intake of zinc.
In a second model, all analyses were adjusted for factors with at least five percentage points difference between cases and controls in supplementary table S1: Age, socio-economic index, use of oral contraceptives, menopausal status at baseline, use of hormone replacement therapy (HRT) and year sample was taken.
Several sensitivity analyses were conducted. In the analyses using dietary intake of zinc as an indicator for zinc status, a sensitivity analysis was performed adjusting for the abovementioned factors plus season of collection of dietary data, interviewer who conducted the diet history interview and dietary method before and after September 1st, 1994, but not adjusting for baseline year. In the analyses using serum zinc, a sensitivity analysis was done adjusting for factors with at least five percentage points difference in supplementary table S1 plus time of year sample was taken. An additional sensitivity analysis was made excluding incident cases occurring within the two first years of follow-up. Since phosphorus have been shown to affect the availability of zinc absorption, all analyses were additionally adjusted for intake of phosphorus, both from diet and supplements. Phosphorus was adjusted for total energy intake using the residual method, as described above. Finally, because 70% of all serum zinc in bound to albumin, a subsample, consisting of 694 cases and 788 controls, was used to conduct a sensitivity analysis in the analyses with serum zinc, adjusting for factors with at least five percentage points difference in supplementary table S1 plus albumin levels.
All analyses were performed using SPSS Statistics version 25.