The aim of this observational study is to investigate the association between mid-life changes in cardiorespiratory fitness and the risk of breast cancer incidence and mortality among Swedish women.

The main questions to answer are:

- Are changes in cardiorespiratory fitness associated with the risk of being diagnosed or dying from/with breast cancer later in life?

Participants performed at least two occupational health assessment tests, which consisted of a submaximal ergometer cycle test, measurement of body mass and height to calculate BMI, and a questionnaires on physical and life style habits.

PLAN FOR THE STATISTICAL ANALYSES

This plan has been made before any data analysis has been conducted.

The primary statistical analysis aims to determine if there is any association between changes in estimated cardiorespiratory fitness (both absolute L/min and relative ml/min/kg) and breast cancer incidence and mortality.

Cox proportional hazards models will be analysed with inverse probability treatment weights of time-varying confounders such as BMI, physical activity smoking and alcohol. Time stable confounders will be baseline cardiorespiratory fitness, education, co-morbidities. Risk time will be calculated as the time from last cardiorespiratory fitness test to the diagnosis of breast cancer or the end of study. In addition, restricted cubic splines will be performed.

Secondary analysis will be performed to determine if the association remains after dividing into prespecified groups. The first analysis will be conducted by grouping individuals who decrease, stay stable or increase in cardiorespiratory fitness. The second analysis, we will conduct is an analysis to investigate if age (proxy for menopausal status) modifies the relationships between change in cardiorespiratory fitness and breast cancer incidence or mortality. This will be performed in two analyses by splitting on age 50 and 60, respectively. In the third analysis we will investigate if baseline fitness modifies the association. This will be done with fitness both as a continuous and stratified into tertiles.

Proportional hazard assumption was checked using scaled Schönfelts residuals.

Analysis will include four models and include confounders relevant to the outcomes and research questions. These will be as follows:

Model 1: Baseline cardiorespiratory fitness Model 2: Baseline cardiorespiratory fitness, age, education and date of last health assessment test Model 3: Baseline cardiorespiratory fitness, age, education and date of last health assessment test, BMI Model 4: Baseline cardiorespiratory fitness, age, education and date of last health assessment test, BMI, physical activity, smoking and alcohol intake.

Sensitivity analysis: To address potential reverse causality we will exclude individuals who are diagnosed with breast cancer within two years after the last health assessment.

Inclusion Criteria:

• Older than 18 years
• Performed at least two health assessment tests

Exclusion Criteria:

• Cancer diagnoses prior to the second health assessment test
• Missing data on any exposure or confounder

Only women will be included in the study

This study used data from the Health Profile Assessment (HPA) database managed by the HPI Health Profile Institute AB in Stockholm, Sweden. The HPAs, which include a questionnaire on physical activity, lifestyle, and perceived health, as well as body mass and height measurements and a submaximal cardiorespiratory fitness test, have been conducted since the 1970s and stored in a central database since 1990. Participation was optional and offered to employees at no cost through the occupational and health services contracted by HPI and HPI was responsible for the standardization and administration of the database.

The HPI-database was coupled to the national death registry to obtain data on cancer incidence and mortality of the involved participants.

The data underlying the findings in this study are currently not publicly available as the original ethical approval application and the informed consent from did not include such direct free access to the data. Data are stored by and can by requested from the HPI Health Profile Institute support@hpi.se.