Eating habit of adding salt to foods and incident sleep apnea: a prospective cohort study

UK Biobank recruited over 500,000 participants aged 40–69 between 2006 and 2010 [19]. Participants attended 1 of the 22 assessment centers across England, Wales, and Scotland, where they completed a touch-screen questionnaire, had physical measurements taken, and provided biological samples, as described in detail elsewhere [20]. Health-related outcomes of the participants were collected through linkages to electronic health records, including inpatient hospital records, primary care records, and death registrations. UK Biobank has ethics approval from the North Multi-centre Research Ethics Committee (11/NW/0382). All participants provided informed written consent. This research has been conducted using the UK Biobank Resource under project numbers 63454 and 69,424.

We excluded participants who withdrew (n = 94), those with incomplete data on the frequency of adding salt to foods (n = 1128), participants with sleep apnea or other diagnosed sleep disturbances at baseline (n = 5284), and participants with sleep apnea or died within the initial 2 years of follow up (n = 3444). Our final analysis included 488,196 participants (Fig. 1).

Participants attended 1 of 22 assessment centers across the UK, where they completed a touch screen questionnaire. One of the questions asked, “Do you add salt to your food? (Do not include salt used in cooking)”, and participants could select one answer from the given options: “Never/rarely”, “Sometimes”, “Usually”, “Always”, and “Prefer not to answer”. In the primary analysis, salt use was coded as an ordered categorical variable with four levels of discretionary salt use.

Prevalent and incident sleep apnea cases within the UK Biobank were ascertained through data linkage to hospital inpatient records, death registries, primary care, and self-reported fields. The date of diagnosis was set as the earliest date of sleep apnea codes recorded regardless of the source used. According to the International Classification of Disease (ICD) edition 10, sleep apnea was defined as code G47.3. Participants’ self-reported diagnosis was identified from self-reported non-cancer illness code (1123) in the data field 20,002 (Additional file 1: Table S1).

Potential confounders were acquired using a baseline touch-screen questionnaire, including age, sex, ethnicity, household income, highest qualification, employment status, smoking status, and alcohol intake. The Townsend deprivation index was applied to indicate socioeconomic status, and a high score was equated with a high socioeconomic deprivation [21]. In accordance with the healthy physical activity recommendations [22], we defined physical activity as inactive (no documented moderate or vigorous physical activity), insufficient (moderate activity of < 150 min/week and vigorous activity of < 75 min/week), and active (moderate activity of ≥ 150 min/week or vigorous activity of ≥ 75 min/week or equivalent). The mean values of vegetable intake, fruit intake, processed meat intake, and unprocessed red meat intake were used in this study. BMI was calculated by dividing a participant’s weight, measured to the nearest 0.1 kg using the Tanita BC-418 MA body composition analyzer (Tanita Corporation of America, IL), by the square of their standing height in meters, measured with a Seca 202 device (SECA, Hamburg, Germany). Hypertension, diabetes, atrial fibrillation, congestive heart failure, stroke, and asthma at baseline were ascertained using ICD diagnosis codes, record-linkage data from local general practitioners, or self-reported previous diagnoses. Menopause and history of hormone replacement therapy were self-reported. A casual (spot) urine was obtained at the end of about a 2-h visit and stored at -80℃. Sodium and potassium concentrations were measured at the UK Biobank laboratory from stored urine aliquots by the Ion Selective Electrode (potentiometric) method using Beckman Coulter AU5400. We estimated 24-h sodium excretion and 24-h potassium excretion (mg/day) from the spot urinary concentration values based on the Kawasaki formula [23]. Information on sleep duration, chronotype, and major dietary changes in the last 5 years were acquired using baseline touch-screen questionnaire. Drug use, including depressants and drugs for blood pressure and blood sugar, was ascertained by a touch-screen questionnaire and verbal interview. More details are shown in Additional file 1: Table S2.

Descriptive characteristics of the sample are presented as either mean (SD) or number (percentage). Follow-up time was calculated from the baseline date to diagnosis of the outcome, death, or the censoring date (30 September 2021 in England, 31 July 2021 in Scotland, and 28 February 2018 in Wales), whichever occurred first. The sleep apnea incidence rate (per 1000 person-years) was calculated, and the Poisson regression model was assumed to calculate the confidence intervals (CIs). Cox proportional hazards regression models were used to estimate the hazard ratio (HR) and 95% CI. The proportional hazard assumption was evaluated by tests based on Schoenfeld residuals [24], and no violation of this assumption was observed. We ran four incremental models for outcome: ‘model 1’ included age, sex, race (White and non-White); ‘model 2’ additionally included sociodemographic covariates: educational attainment (college or university, vocational, upper secondary, lower secondary, or others), socioeconomic status (quintiles of Townsend deprivation index), household income (< £18 000, £18 000-£30 999, £31 000-£51 999, £52 000-£100 000, > £100 000, or unknown), employment status (employed, unemployed, or retired), and 22 assessment centers; ‘model 3’ additionally included lifestyle factors: smoking status (never, previous, current, or unknown), alcohol intake (moderate alcohol intake [yes/no]), physical activity (inactive, insufficient, or recommended level), vegetable intake (quartiles), fruit intake (quartiles), processed meat intake (quartiles), unprocessed red meat intake (quartiles), BMI (< 18.5 kg/m², 18.5 ~ 25 kg/m², 25 ~ 30 kg/m², ≥ 30 kg/m²); ‘model 4’ additionally included multimorbidity: diabetes (yes/no), hypertension (yes/no), atrial fibrillation (yes/no), congestive heart failure (yes/no), stroke (yes/no), and asthma (yes/no). If covariate information was missing, we used a missing indicator approach for categorical variables and with mean values for continuous variables [25].

To assess the potential effect modification, stratified analyses, and interaction analyses were performed by the following factors: sex (women or men), age (< 60 or ≥ 60), race (White, non-White), socioeconomic status (dimidiate Townsend deprivation index), physical activity (active, insufficient, and inactive), smoking status (never, ever, and current), moderate drinking (yes, no), BMI (< 25, 25.0–29.9, or ≥ 30), hypertension (yes, no), diabetes (yes, no), atrial fibrillation (yes, no), congestive heart failure (yes, no), stroke (yes, no), and asthma (yes, no).

We conducted several sensitivity analyses. First, to examine robustness against missing covariates, we repeated the main analysis using a complete analysis approach. Second, to examine robustness against severe medical conditions, we repeated the main analysis by excluding those with CVD or cancer at baseline. Third, to assess the risk of the habit of adding salt to foods with the outcome, salt use was coded into a binary variable (“generally add salt to foods” and “do not generally add salt to foods”) as an independent variable in Cox models. Fourth, to examine robustness against great changes in eating habits, we repeated the main analysis by excluding those who reported great changes in eating habits. Fifth, we repeated the main analysis in participants with low OSA risk to minimize the undiagnosed cases of sleep apnea. Based on Berlin Questionnaire, the risk of OSA was determined by reports of snoring, daytime sleepiness, body mass index (BMI), and blood pressure [26]. Sixth, the hormonal status might be related to the prevalence of sleep-disordered breathing in females [27, 28]. We further adjusted for menopause and history of hormone replacement therapy among females. Seventh, we adjusted for other potential covariates, including sodium and potassium in the urine, sleep duration, chronotype, weight change compared with 1 year ago, waist circumference, depressant use, drugs for blood pressure and blood sugar, and diuretics. Eighth, we evaluated the potential mediation effect by modeling the cross-product term of the blood pressure, BMI, and C-reactive protein level with the frequency of salt added to foods [29].

This study has several major strengths, including the large sample size and the wealth of information, enabling us to adjust possible covariates and conduct comprehensive sensitivity and subgroup analyses. The following inevitable limitations deserve mentioning. First, it is challenging to distinguish subtypes of sleep apnea, including CSA and OSA, in this study. However, in reality, given considerable overlap between these two conditions, the more general term sleep apnea to encompass both subtypes’ manifestations of disordered ventilatory control might be more appropriate [34]. Second, the frequency of adding salt to foods is partly related to the salt intake of individuals. We conducted further adjustments for the estimated 24-h excretion of sodium and potassium in sensitivity analyses, and the results were not substantially changed. Third, this study was also limited by a single assessment of adding salt to foods at baseline. Major dietary changes may accompany by the change of habits of adding salt to foods. Therefore, we conducted a sensitivity analysis to exclude participants with major dietary changes in the last 5 years. Fourth, sleep apnea is a highly underestimated disease due to the low rate of diagnosis. To minimize the influence of undiagnosed cases of sleep apnea, we restricted our analyses to those participants with low OSA risk in the sensitivity analysis. Fifth, although we conducted the sensitivity analyses, the reverse causality between the habit of adding salt to foods and incident sleep apnea cannot be eliminated due to the nature of the observational study.