Study population

FIND was conducted between September 2012 and October 2018. It was a double-blind, placebo-controlled RCT designed to investigate the effects of 5 year vitamin D3 supplementation on the incidence of major chronic diseases among generally healthy men and women from Finland [11]. The prespecified primary and secondary outcomes were CVD and cancer, and incidence of type 2 diabetes was a prespecified tertiary outcome.

The inclusion criteria were male participants aged ≥60 years and post-menopausal female participants aged ≥65 years without a history of CVD (including myocardial infarction, stroke, transient ischaemic attack, angina pectoris, coronary artery bypass grafting or percutaneous coronary intervention) or cancer (except non-melanoma skin cancer). The exclusion criteria were history of kidney stones, renal failure or dialysis, hypercalcaemia, hypo- or hyperparathyroidism, severe liver disease (cirrhosis) or sarcoidosis or other granulomatous diseases such as active chronic tuberculosis or Wegener’s granulomatosis; and use of vitamin D >800 IU/day or calcium >1200 mg/day from all supplemental sources combined (or, if taking, not willing to decrease or forego such use during the trial).

The original goal was to recruit 18,000 participants from three regions of eastern Finland (North Savo, North Karelia and Kainuu) [11], where, at the time of recruitment in 2012, 13% of the total population of about 500,000 people were men aged ≥60 years and 12% were women aged ≥65 years [12]. Biological sex was determined based on social security number and we aimed to recruit an approximately equal number of men and women. Data on ethnicity were not collected, because in eastern Finland the population is predominantly White in these age groups.

In total, the recruitment process yielded only 2495 participants, who were randomised to receive 1600 IU/day of vitamin D3, 3200 IU/day of vitamin D3 or placebo (Fig. 1) [11]. A random subcohort of 551 participants took part in the more detailed baseline examinations at the University of Eastern Finland, including blood sampling, with follow-up visits after 6, 12 and 24 months, respectively (see electronic supplementary material [ESM] Fig. 1) [11]. A non-study group statistician carried out sex-stratified simple randomisation in a 1:1:1 ratio, based on computerised random number generation. We did not have information on diabetes status at baseline, but we used self-reported use of diabetes medications as an exclusion criterion in the current analyses. Of the 2495 participants, 224 reported using diabetes medications at baseline; therefore, 2271 participants were included in the analyses of incident type 2 diabetes and 505 participants were included in the subcohort analyses (Fig. 1, ESM Fig. 1).

Fig. 1

Pharmaceutical grade supplements (oral pills) were prepared specifically for the FIND trial by Galena Pharma (Kuopio, Finland). The pills contained either 0, 1600 IU or 3200 IU of vitamin D3. The pills were annually either mailed to the participants or given at study visits. Double-blinding was maintained throughout the study.

The participants completed questionnaires at baseline, after 12, 24 and 36 months and at the end of the trial at 60 months (an English translation of the questionnaire is available in the ESM) [11]. At baseline, 36 and 60 months the questionnaire also included a validated 142-item food frequency questionnaire. The final questionnaire included a question for assessing adherence to the study supplementation. Choices ranged from ‘<50%’ to ‘100% or almost 100%’.

BMI was calculated based on the weight and height reported in the baseline questionnaire. BMI based on self-reported data was used in the analyses that evaluated the effects of supplementation based on BMI. In the subcohort, waist circumference and weight and height were also measured by a study nurse during study visits. BMI based on these measurements was used in the analyses that investigated changes in BMI during follow-up in the subcohort.

In the subcohort of 504 participants with available blood samples, serum 25-hydroxyvitamin D3 (25(OH)D3) concentrations were measured using HPLC from samples collected at months 0, 6 and 12 [11, 13]. In a smaller subgroup of 60 participants, a direct competitive chemiluminescence immunoassay (Liaison 25 OH Vitamin D Total Assay, DiaSorin, Stillwater, MN, USA) was also used for serum 25(OH)D3 measurement from samples collected at months 0, 6, 12, 24 and 52 [11]. The serum 25(OH)D3 results were standardised against National Institute of Standards and Technology (NIST) reference values to generate representative values that were comparable to the HPLC results for statistical analyses [11].

All participants were free to withdraw from the study at any time without explanation. All participants provided written informed consent. The appropriate study approvals were obtained from the ethics committee of Kuopio University Hospital (#30/2010) and the Finnish Institute for Health and Welfare.

Measurement of plasma glucose and insulin concentrations

Plasma glucose concentrations were assayed using the photometric hexokinase method (KoneLab 20XT, Thermo Fisher Scientific, Vantaa, Finland) and plasma insulin concentrations were measured using a chemiluminescence assay (DiaSorin Liaison, DiaSorin, Dietzenbach, Germany) from samples collected at baseline and at the 12 and 24 month study visits. HOMA-IR was calculated according to the previously published formula [14].

Assessment of type 2 diabetes

The E11 codes from the ICD-10 (https://icd.who.int/browse10/2019/en) were used for diagnosis of type 2 diabetes mellitus. The codes were retrieved from two care notification registers, Hilmo for specialised healthcare [15] and Avohilmo for primary healthcare [16] (Finnish Institute for Health and Welfare, data agreement THL/523/5.05.00/2020). Causes of death were based on information provided by the Causes of Death Register (Statistics Finland, TK/1007/07.03.00/2022) [17]. The first date when the code E11 appeared in any of the registers was considered the date of diabetes diagnosis.

Statistical analysis

Cox proportional hazards regression models adjusted for age and sex were used to predict the hazard of type 2 diabetes. Participants contributed follow-up time from randomisation until type 2 diabetes diagnosis, death, end of the 5 year follow-up or withdrawal from the study for personal reasons (n=412). The results are presented as HRs with 95% CIs. In the prespecified analyses, the two vitamin D arms were compared separately with the placebo arm. We also performed analyses that were not prespecified and should therefore be considered as exploratory. First, because of the low number of events, the effects of vitamin D supplementation were analysed after combining the two vitamin D arms. Second, potential latent effects of supplementation were investigated by excluding the type 2 diabetes events that occurred during the first 2 years of follow-up and by including the events that occurred during the post-supplementation follow-up period until 31 December 2021. In the subcohort with available data, changes in plasma glucose and insulin concentrations and HOMA-IR and in BMI and waist circumference between baseline and months 12 and 24 were analysed using linear mixed modelling. The logrank (Mantel–Cox) test was used to detect trends. Interactions were considered as multiplicative interactions. IBM SPSS version 29 (IBM, Armonk, NY, USA) was used for analyses. A two-sided p value of <0.05 was used to determine statistical significance. The p values were not adjusted for multiple comparisons.