Women diagnosed with stage I–III breast cancer between 2005 and 2020 were identified from the NBCR. Stage IV (metastatic) breast cancers were not included in this study. Most stage I–III breast cancer patients have surgery as the main component of their treatment [3], while for stage IV breast cancer patients, systemic therapy is the mainstay of treatment, and only some undergo surgery [15]. The NBCR combines the Auckland, Waikato, Wellington and Christchurch Registers, which includes 98% of prevalent breast cancer patients in these regions [16]. Once a patient is diagnosed with breast cancer, their information is included in this confidential register. The Waikato and Auckland Registers were collecting data prospectively before 2005 and the Christchurch and Wellington Registers started in 2009.

The NBCR includes extensive data on the demographics of these women (and men) and their tumour characteristics. This study used data on age at diagnosis, menopausal status, ethnicity, and domicile of residence, diagnosis date, mode of detection (screen-detected or symptomatic), TNM cancer stage (I, II, III and IV), grade (1, 2 and 3), biomarkers (oestrogen receptor (ER), progesterone receptor (PR) and human epidermal growth factor receptor 2 (HER2), cancer treatment including surgery, radiotherapy and systemic treatments, and whether the surgery was received in a public or a private hospital. NBCR data were linked by National Health Index (NHI) number to national-level health data, to determine diabetes status (from the VDR), comorbidities (from hospital admissions recorded in the National Minimum Dataset: NMDS) and radiotherapy (from Radiation Oncology Collection: ROC). The NHI number is a unique identifier for people receiving healthcare services in New Zealand. The VDR records diabetes patients identified using an algorithm which defines whether an individual has diabetes based on national level data on inpatient hospitalisation ICD diagnosis codes (using the NMDS), relevant outpatient events (National Non-Admitted Patient Collection, NNPAC), retinal screening (using regional retinal screening programme datasets), pharmaceutical dispensing (PHARMS) and pathology test claims (Laboratory Claims dataset). The VDR does not differentiate between types of diabetes.

Patients were classified into Māori, Pacific, Asian and European/Other ethnic groups. Ethnicity is self-identified in New Zealand. The categories of ethnicities were based on the 2018 census ethnic groups used by Statistics New Zealand [17]. Socioeconomic deprivation was defined using the New Zealand Index of Deprivation 2018 (NZDep 2018) analysed as quintile, from 1 (least deprived) to 5 (most deprived) [18]. NZ Dep is an area-level measure based on aggregated census data, not on individual data. The study period was separated into groups: 2005–2009, 2010–2014 and 2015–2020. Breast cancer subtypes were categorised into five groups according to biomarker status [19,20,21,22]: (1) luminal A: ER+ , PR+ and HER2−; (2) luminal B HER2-: ER or PR+ (but not both+), HER2−; (3) luminal B HER2+ : ER+ and/or PR+ , HER2+ ; (4) HER2 non-Luminal: ER−, PR−, HER2+ ; and (5) triple negative: ER−, PR−, HER2−. Patients ever recorded as having diabetes before cancer diagnosis in the VDR were considered to have diabetes at the time of cancer diagnosis. All comorbid conditions recorded on the NMDS for hospitalisations in the 5 years up to the index hospitalisation date were identified to calculate a C3 Index score for each patient [23]. The C3 Index is a cancer-specific index of comorbidity, with scores categorised into ‘0’ (≤ 0), ‘1’ (≤ 1.00), ‘2’ (≤ 2.00) and ‘3’ (> 2.00) [23]. Diabetes was excluded as a comorbidity from the C3 Index score calculation. Missing values in these factors were considered as a separate category to maximise the number of patients included in the data analysis.

The cancer treatments examined included BCS, mastectomy, breast reconstruction after mastectomy, and adjuvant radiotherapy after BCS. We compared the proportions of patients receiving treatments between those with diabetes and without diabetes, and examined differences between groups with; Chi-square tests. Logistic regression modelling was used to estimate the odds ratio (OR) and the 95% confidence interval (95% CI) of having no surgery (either mastectomy or BCS) for patients with diabetes compared to patients without diabetes by cancer stage, after adjustment for year of diagnosis, age, menopausal status, ethnicity, deprivation quintile, mode of detection, C3 Index score. Amongst the patients having surgery, the OR of having mastectomy in the diabetes group was estimated after adjustment for year of diagnosis, age, menopausal status, ethnicity, deprivation quintile, mode of detection, C3 score, cancer stage, grade, subtype and public/private hospital treatment. Adjusting for the same factors, we also estimated the adjusted ORs of having reconstruction after mastectomy and having radiotherapy after BCS. The OR of having surgery delay (> 31 days and > 90 days) for women with diabetes compared to women without diabetes, after adjustment for year of diagnosis, age, menopausal status, ethnicity, deprivation quintile, mode of detection, C3 score, cancer stage, grade, subtype, public/private hospital treatment and type of surgery. The ORs were estimated before and after stratification by cancer stage and ethnic group.

All analyses were performed in R Studio (Massachusetts, United States). Ethical approval for the study was granted through the University of Waikato Human Research Ethics Committee (reference: HREC(Health)2021#89).