Iodine deficiency is a global health issue and the main cause of preventable brain damage(1). The WHO recommends iodine fortification to achieve sufficient iodine intake (150 µg/d in non-pregnant adults)(1). In Denmark, iodine fortification was cautiously implemented in two steps(Reference Laurberg, Jørgensen and Perrild2): voluntary iodine fortification (8 parts per million to salt) in June 1998 and mandatory iodine fortification (13 parts per million to salt) from July 2000. Iodine fortification has been accompanied by a monitoring programme of the incidence of overt thyroid dysfunction(Reference Laurberg, Jørgensen and Perrild2). This monitoring programme has shown that the iodine fortification in Denmark was followed by a transient rise in the occurrence of hyperthyroidism, which was more marked in older adults compared with younger adults(Reference Carlé, Pedersen and Knudsen3,Reference Petersen, Knudsen and Carlé4) .
Iodine nutrition has an impact on the occurrence of thyroid dysfunction. Hyperthyroidism is more frequent in mild and moderate iodine-deficient populations, whereas iodine-replete populations may have a higher occurrence of hypothyroidism(Reference Petersen, Knudsen and Carlé4–Reference Petersen, Knudsen and Carlé7). Hyperthyroidism can lead to complications such as atrial fibrillation, heart failure and increased fracture risk. These complications are associated with increased mortality, especially in vulnerable groups such as older adults(Reference Selmer, Olesen and Hansen8–Reference Gencer, Collet and Virgini10). However, the association between long-term iodine intake level and longevity remains to be described.
We performed a follow-up on a previous cross-sectional study of two groups of older adults in Denmark conducted in 1997–1998(Reference Andersen, Iversen and Terpling5,Reference Andersen, Pedersen and Iversen11) . One group lived in the iodine-replete city of Skagen and the other in the moderately iodine-deficient city of Randers. The prevalence of thyroid dysfunction was markedly different with 26 % of participants in the iodine-deficient city having hyperthyroidism compared with 6 % among the iodine replete(Reference Andersen, Iversen and Terpling5). Our follow-up aimed to assess the association between long-term residency in an iodine-replete environment and longevity with a 20 years’ follow-up on older adults aged 75–80 years.
Methods
Setting
Randers and Skagen are separated by only 140 km on the peninsula Jutland in Denmark. Iodine content in ground water differs due to differences in aquifer source rock as Skagen is based on raised sea bottom. Previous investigations found mean drinking water iodine concentrations of 2 µg/l in Randers and 139 µg/l in Skagen without dispersion between samples taken at different time points(Reference Andersen, Pedersen and Iversen11–Reference Andersen, Petersen and Laurberg13). Furthermore, the clinical investigations in 1997–1998 found markedly different iodine intake levels between the two populations with a median urinary iodine concentration of 55 µg/l in Randers, suggesting mild to moderate iodine deficiency, compared with 160 µg/l in Skagen confirming that this population was iodine replete(1,Reference Andersen, Pedersen and Iversen11,Reference Rasmussen, Larsen and Ovesen12) .
Baseline
Invited participants in Randers were all residents born in 1920, and those in Skagen were all residents born in 1918–1923. They were investigated in late 1997 and 1998 as described in detail previously(Reference Andersen, Iversen and Terpling5). The wider age range in Skagen was to ensure equal sample sizes. Participation rate was 47 %(Reference Andersen, Iversen and Terpling5). Ten percent of non-responders were previously selected at random for a telephone interview. There were no differences in sex, smoking, alcohol intake, co-morbidity categorised according to Charlson co-morbidity index(Reference Charlson, Pompei and Ales14), or known thyroid disease between responders and non-responders(Reference Andersen, Iversen and Terpling5).
Participants in both cities completed a questionnaire about previous or current thyroid disease and treatment, smoking, alcohol and co-morbidities. Data on income for the year of 1996 were gathered from the Danish Registry of Income Statistics and indexed to one at the median level of Skagen. Spot urine samples were analysed for creatinine using a kinetic Jaffé method, and iodine concentrations were determined by the Sandell–Kolthoff reaction modified after Wilson and van Zyl as described previously(Reference Andersen, Pedersen and Iversen11,Reference Andersen, Karmisholt and Pedersen15) .
Follow-up
Participants were followed until death, emigration or end of study by 31 December 2017. Data on mortality were collected from the Registry of Causes of Death. Data on emigration were collected from the Central Persons Registry.
Statistical analyses
Categorical variables are presented as numbers and percentages. Continuous variables, age and income, are given with medians and interquartile ranges due to non-normal distributions. To compare variables between groups, χ2 and Mann–Whitney U tests were used. Kaplan–Meier curves of survival probability are presented.
The association between residency and death during follow-up was analysed using multivariate Cox proportional hazards model with calendar year as the time scale. All associations were tested using two models. First, a basic model included age and sex as covariates. Second, an advanced model included age, sex, smoking, alcohol consumption, Charlson co-morbidity index and income.
Proportional hazards assumptions were checked by inspecting cumulative Martingale residual plots, and they were met. Interaction terms between place of residency and all covariates were tested using likelihood ratio tests. Inspection of Martingale residuals revealed that income was non-linear and was modelled accordingly as a restricted cubic spline with five knots at the 5th, 25th, 50th, 75th and 95th percentiles. Age was recorded as a discrete variable and was modelled as such in the analysis.
The variation inflation factor for the residency variable was 3·35 with age in the model and 1·14 without. This suggested multi-collinearity with age likely due to sampling methods. Therefore, sub-group analysis on participants aged 78 years was performed.
The sample size of 428 participants with an event rate of 91 % in two similar sized groups gave us a minimally detectable hazard ratio of 0·76 at 80 % power and 5 % significance level.
All analyses were performed as complete case analyses as there were few missing data. All analyses were performed using R statistical software version 3.5.1 (R Core Team 2018).
Ethics
Ethical approval was granted by the Regional Research Ethics Committees of Northern Jutland and Viborg County(Reference Andersen, Iversen and Terpling5). Approval for registry follow-up was granted by the Danish Data Protection Agency (P-2019-191).
Results
In the original clinical study, there were 430 participants. Of these, two participants were excluded due to insufficient record linkage. One emigrated during follow-up and was censored at the date of emigration.
Baseline characteristics
Participants from Randers and Skagen had similar sex distributions, smoking habits and Charlson co-morbidity index (Table 1), whereas differences in alcohol consumption (primarily occasional use) and income were seen. As expected according to the inclusion, there were also differences in urinary iodine concentration, age and length of residency.
Baseline characteristics by residency
(Median values and interquartile ranges (IQR); numbers and percentages)
UIE, urinary iodine excretion.
* UIE corrected for age- and sex-specific creatinine excretions (men 0·95 g/l; women 0·7 g/l).
† Income is indexed to median level in Skagen.
Survival analyses
Overall, survival was higher in Skagen than in Randers for 20 years’ follow-up (Fig. 1), and 88 % of participants in Skagen died during follow-up compared with 95 % of participants in Randers. Thus, Skagen dwellers were followed for 11·5 (interquartile range 6·8, 17·0) years, while Randers dwellers were followed for a median of 10·1 (interquartile range 4·8, 14·1) years.
Kaplan–Meier survival curve by residency in iodine-replete Skagen or iodine-deficient Randers. 
, Skagen; 
, Randers.
Cox regression conducted both adjusted for sex and age (basic model) and with additional adjustment for smoking, alcohol consumption, Charlson co-morbidity index and income (advanced model) provide strong statistical support for an association between residency and longevity (Table 2). The hazard ratio for death was 0·60 (95 % CI 0·41, 0·87) in favour of Skagen residency. The sub-group analysis of 78-year-olds showed similar trends as the primary analysis.
Twenty years’ follow-up of older Randers and Skagen residents in 1998
(Median values and interquartile ranges; hazard ratios (HR) and 95 % confidence intervals)
* Adjusted for age and sex.
† Adjusted for age, sex, smoking, alcohol consumption, Charlson co-morbidity index and income.
Discussion
We found that long-term residency in an iodine-replete environment is associated with lower mortality during 20 years’ follow-up in older adults. This is consistent with findings that longevity in older adults is associated with higher thyroid-stimulating hormone (TSH)(Reference Jansen, Roelfsema and van der Spoel16,Reference Gussekloo, van Exel and de Craen17) , as higher iodine intake levels in a population raise the average TSH with age(Reference Hollowell, Staehling and Flanders18,Reference Bjergved, Jørgensen and Perrild19) .
The influence of iodine nutrition on morbidity has been studied extensively(Reference Laurberg, Jørgensen and Perrild2,Reference Petersen, Knudsen and Carlé4–Reference Petersen, Knudsen and Carlé7,Reference Laurberg, Cerqueira and Ovesen20) . Specifically, iodine nutrition is a pivotal determinant of thyroid dysfunction, and this influence is strengthened with advancing age(Reference Laurberg, Cerqueira and Ovesen20). Iodine deficiency is known to be associated with an increased occurrence of hyperthyroidism. In our baseline study, 26 % of participants from the moderately iodine-deficient Randers residents had hyperthyroidism compared with 6 % among the iodine-replete Skagen residents as published previously(Reference Andersen, Iversen and Terpling5). Even slight thyroid dysfunction may influence mortality, and hyperthyroidism increases the risk of complications such as atrial fibrillation, heart failure and fractures(Reference Selmer, Olesen and Hansen8–Reference Gencer, Collet and Virgini10). Thus, the difference in the prevalence of hyperthyroidism could explain some of the differences in longevity between the two cities.
Long-term iodine intake influences the occurrence of thyroid disorders as discussed above. The difference in iodine intake level in these populations was determined by the iodine content of tap water(Reference Andersen, Pedersen and Iversen11,Reference Pedersen, Laurberg and Nøhr21) . Tap water iodine was documented to be stable for decades prior the baseline examination(Reference Andersen and Laurberg22). An iodine fortification programme was implemented during follow-up, and the monitoring programme (DanThyr) showed an approximately 50 µg/24 h rise in urinary iodine excretion(Reference Rasmussen, Carlé and Jørgensen23). A raised iodine intake was followed by a transient increase in hyperthyroidism in a formerly iodine-deficient population parallel to the population in Randers(Reference Petersen, Knudsen and Carlé4), and hence the differences in the present report could be strengthened. Still, for participants not taking iodine supplements, the anticipated population iodine intake level would remain within the recommended level in Skagen and mild iodine deficiency to recommended intake would be anticipated in Randers(Reference Andersen, Pedersen and Iversen11).
Data on the association between iodine nutrition and mortality are scarce. Only one study aimed to assess the relationship between a single-spot urine iodine concentration measurement and mortality based on a sample of the National Health and Nutritional Examination Survey III (NHANES)(Reference Inoue, Leung and Sugiyama24). They did not find that iodine deficiency was related to increased mortality in conflict with our results. However, the population included was iodine replete(Reference Lee, Cho and Shin25) without the same large span of iodine intake levels as found in the present study. Furthermore, the study excluded participants with known thyroid dysfunction and thus prevented for any effect of iodine deficiency mediated through thyroid disease. Additionally, they based their analysis on a single urine sample which potentially could lead to misclassification of iodine nutritional status on the individual level and provides very limited information about long-term iodine intake(Reference Knudsen, Christiansen and Brandt-Christensen26).
An additional report based on the National Health and Nutritional Examination Survey population showed no association between iodine intake levels and CVD after the exclusion of individuals with abnormal thyroid function(Reference Tran, Erskine and Kiefe27). Their study used the age- and sex-adjusted iodine:creatinine ratio, which has been shown to give a more accurate assessment of iodine nutrition status than urinary iodine concentration(Reference Knudsen, Christiansen and Brandt-Christensen26,Reference Andersen, Pedersen and Pedersen28) . Their cut-off score for low iodine levels was based on quartiles, which complicates the determination of iodine nutritional status of subjects in the lowest quartile of iodine intake. These two studies, based on the National Health and Nutritional Examination Survey, in conjunction with our study, suggests that any association between iodine nutrition and mortality may be mediated by thyroid function. However, this needs confirmation in studies using appropriate statistical methods and includes follow-up on thyroid function. Furthermore, the National Health and Nutritional Examination Survey stems from an iodine-replete population. Thus, further studies are needed to rule out non-thyroidal effects in iodine deficiency.
Our study also had some limitations. Importantly, there may be unknown and residual confounders including socio-economic status and other unmeasured differences between the two cities, which could influence our results and thus needs further exploration in prospective cohorts. Our study would have been strengthened by the addition of a city with mild iodine deficiency to explore a potential dose–response relationship. Furthermore, we did not perform follow-up on thyroid function or iodine intake in our participants, and any potential thyroidal or non-thyroidal effects cannot be evaluated. Finally, we chose not to analyse causes of death, as systematic validation of the Causes of Death Registry is lacking, the autopsy rate is low and there are indications of misclassification in the registry(Reference Helweg-Larsen29). Additionally, we were concerned that geographical variations in coding practices could severely confound any findings from analysis of causes of death.
Conclusion
Long-term residency in an iodine-replete environment is associated with increased longevity as evaluated from 20 years’ follow-up in older adults.
Acknowledgements
The authors would like to thank the late Professor Peter Laurberg for his invaluable contribution to design, planning, and Steen Terpling and Finn Iversen for the execution of the Randers–Skagen study.
The work was sponsored by the Department of Geriatric Medicine, Aalborg University Hospital, Aalborg, Denmark. C. T.-P. has received a research grant from BIOTRONIK and a research grant as wells as advisory board honoraria from Bayer.
J. R. and S. A. designed the study. K. M. P. and S. A. conducted the initial clinical study and C. T.-P acquired follow-up data. J. R., C. T.-P and S. A. designed and performed the statistical analysis. J. R., M. B. D., G. V. B. S., M. G. J., S. L. A., A. C., I. B. P., C. T.-P. and S. A. interpreted the results. J. R. wrote the manuscript and all other authors critically revised the manuscript. All authors read and approved the final version of the manuscript.
The authors declare that there are no conflicts of interest.
