Between December 1993 and May 1997, the Danish Diet, Cancer, and Health Study recruited 57,053 men and women who were between the ages of 50 and 65 years, who had no history of cancer (excluding non-melanoma skin cancer), and who were living within the greater areas of Copenhagen and Aarhus. Using unique personal identification numbers assigned to all inhabitants of Denmark, the following databases were cross-linked to the cohort on an individual level: The Civil Registration System [21], The National Death Registry [22], The Danish National Patient Register (DNPR) [23], The Danish National Prescription Registry [24], the Register for Selected Chronic Diseases (RUKS), and the Education Registry [25]. We excluded participants with prevalent CVD (n = 2705), defined as record of ischemic heart disease, ischemic stroke, peripheral artery disease, atrial fibrillation or heart failure prior to baseline (more information in Supplementary Table 1). Additionally, participants with missing exposure or covariate data (n = 1511) were excluded, leaving 52,247 participants for analysis in the present prospective cohort study (Supplementary Fig. 1).

Establishment of the cohort was approved by relevant scientific ethics committees and the Danish Data Protection Agency, and all participants gave written informed consent.

Participants, prior to their first study visit, completed a validated semi-quantitative food frequency questionnaire (FFQ) where they reported their usual intake of 192 food and beverages over the preceding 12 months [26,27,28]. A detailed methodology for the calculation of nitrate and nitrite intakes has been documented earlier [29]. In brief, nitrate and nitrite intakes, except for tap water, were derived using two comprehensive databases [9, 17] and from government analyses as part of national monitoring programmes. Preference was given to values from Danish sources. To compute nitrate/nitrite intakes, we multiplied the reported consumption quantity of each food item (g/day) by its designated median nitrate/nitrite value (mg/g), adjusting for a 50% decrease in nitrate content for boiled vegetables.

For tap water nitrate assessment, we utilized the public national geodatabase Jupiter [30]. By spatially linking this data with the addresses of the cohort participants, individual-level nitrate consumption from tap water for each participant between 1978 and 2016 was estimated. Comprehensive details have been outlined in a previous publication [29]. To estimate baseline intakes of tap water-sourced nitrate, intakes of tap water were summed from the FFQ (considering the consumption of tap water, tea, coffee, and water added to fruit syrup) and multiplied by the time-weighted average of the nitrate concentration at every address each cohort participant lived at in the 12 months prior to their enrolment into the study. As constituents in tea, coffee, and fruit syrup might hinder the formation of NOCs [14], we additionally examine associations for intakes of tap water itself.

Food and beverages were sorted into four primary categories: tap water, foods from plant sources (fruits, vegetables, legumes, and wholegrains), foods from animal sources (red meat, poultry, processed meat, offal, dairy, eggs, fish, and other seafood), and other sources (alcoholic beverages and discretionary foods) [29]. Given that vegetables are the predominant dietary nitrate source [9], for relevant public health recommendations, we also looked specifically at vegetable nitrate/nitrite. Furthermore, we distinguished between inherent versus added nitrate/nitrite in animal-sourced foods, examining ‘naturally occurring animal-sourced nitrate/nitrite’ and ‘additive permitted meat-sourced nitrate/nitrite’ (i.e., bacon, ham, salami, sausage, liver paste, and other processed meats), separately.

Information regarding the vital status and date of death of each participant was acquired from the Civil Registration System while data concerning the cause of death was obtained from the National Death Register. Cancer-related mortality was defined as a death with cancer (ICD-10: C00-C97) listed as the primary cause while CVD-related mortality was defined as a death with CVD (ICD-10: I00-I99) listed as the primary cause.

Upon enrollment into the Danish Diet, Cancer, and Health Study, participants provided information regarding various demographic and lifestyle factors. Specifically, participants reported their age, sex, smoking status (i.e., current, former, or never), and smoking history (i.e., packyears). Additionally, participants reported their leisure-time physical activity in winter and summer (number of hours per week which was converted to total daily metabolic equivalent score; MET) in a self-administered questionnaire. Height and weight were objectively measured at baseline and body mass index (BMI) was calculated in kg/m². Data pertaining to the participants’ education level (≤ 7 years, 8–12 years, or ≥ 12 years) and living situation (living with a partner/single) were obtained from the Education Registry and the Civil Registration System, respectively. Intakes of alcohol (g/d), total polyphenols (mg/d; calculated using the Phenol-Explorer database [31]), folate (µg/d), vitamin C (µg/d), and vitamin E (mg α tocopherol equivalents/d) were estimated from the FFQ. Prevalent chronic kidney disease (CKD) and chronic obstructive pulmonary disease (COPD) were defined as a record of the respective disease in either the DNPR or the RUKS registry, while prevalent diabetes was defined as a record of either type 1 or type 2 diabetes in the RUKS registry (see Supplementary Table 1 for more information).

Baseline characteristics of the 52,247 cohort participants are presented for the whole cohort as well as for the lowest and highest quintiles of plant-sourced nitrate, animal-sourced nitrate, and water-sourced nitrate intakes. Participants’ time-to-event was calculated from the date of enrolment until the date of death, emigration, or end of follow-up (31 December 2020), whichever came first. To allow for non-linear relationships between exposures and outcomes, continuous exposure variables were fitted as restricted cubic splines (with 4 knots placed at the 5th, 35th, 65th, and 95th percentiles and the median intake in the lowest quintile taken to be the reference) within separate Cox proportional hazards models for each exposure/outcome combination. The resulting HRs (95% CIs) were graphed with x-axes truncated at 3 standard deviations above the mean for visual simplicity. In tables, HRs and 95% CIs are presented for the median intake in each quintile. No violations in proportional hazards assumptions, assessed via visual inspection of the parallelism of log-log plots of the survival function versus time, were observed. To examine consistency of associations, analyses were stratified by sex (male vs. female), smoking status (ever vs. never), and intakes of vitamin C, vitamin E, folate, and polyphenols (tertile 3 vs. tertile 1 for all). Covariates were chosen a priori using prior knowledge of potential confounders of nitrate intake and premature death. Three models of adjustment were used: Model 1 included age and sex; Model 2 included age, sex, BMI, smoking status, smoking packyears, alcohol consumption, education level, physical activity level and living situation; Model 3 adjusted for the covariates in Model 2 plus intakes of (a) red meat, processed meat, poultry, dairy, fish, sugar and confectionary, soft drinks, refined grains, coffee, and tea when plant-sourced nitrate or nitrite were the exposures of interest, (b) wholegrains, refined grains, vegetables, fruits, vegetable oils, sugar and confectionary, soft drinks, coffee, and tea when animal-sourced nitrate or nitrite were the exposures of interest and (c) wholegrains, refined grains, red meat, processed meat, poultry, dairy, fish, vegetables, fruits, vegetable oils, sugar and confectionary, and soft drinks when water-sourced nitrate was the exposure of interest. We used the ‘all-components model’ approach for Model 3, i.e. adjustment for food groups excluding the exposure food group, as it provides unbiased estimates compared to other methods of energy adjustment [32] and accounts for underlying dietary patterns. In a sensitivity analysis, we additionally adjusted for the exposure food group. All continuous covariates were modelled with restricted cubic splines. As nitrate samples in private wells are sparse, in a sensitivity analysis we excluded participants who were suppled by a private well in the 12 months prior to baseline. To quantify effect estimates on the absolute scale, we furthermore analysed associations between source-dependent nitrate intakes and all-cause mortality using regression analyses of restricted mean survival time based on pseudo-observations [33]. Differences in restricted mean survival time (RMST) at 20 years, according to the different exposure variables, were estimated (Model 3 adjustment). Similar to the Cox regression models, exposure variables were specified as restricted cubic splines and contrast estimates with 95% confidence intervals for the median intake in each quintile compared to the median intake in the first quintile are presented. In a sensitivity analysis we analysed nitrate from drinking water as a time-varying exposure variable, expressing average 15-year exposure. At each point in time, the exposure was assessed by taking the cumulative average concentration of nitrate in drinking water supplied to the addresses where each participant had lived over the previous fifteen years and modelled with a restricted cubic spline. For this analysis, a complete case approach was applied and, as exposure information was only available until 2016, end of follow-up was 31 December 2016. Analyses were undertaken using R statistics (R Core Team, 2022) and SAS 9.4 (SAS Institute, Cary NC). Statistical significance was set at p ≤ 0.05 (two-tailed) for all tests.

The funding source had no role in study design, preparation of this manuscript, or decision to submit the paper for publication.