Association of endemic goitre and exceptional longevity in Sardinia: evidence from an ecological study

Abstract

This study aims to test the hypothesis that a high prevalence of endemic goitre, considered as a proxy measure for subclinically reduced thyroid function in the population, is geographically associated with exceptional longevity. Using historical data available for 377 Sardinian municipalities in the first half of the twentieth century, we performed an ecological study to investigate the geographic distribution of goitre and its spatial association with demographic indicators of population longevity. This analysis was conducted by using both conventional ordinary least square and geographically weighted regression models to take into account spatial autocorrelation and included other longevity-associated factors previously identified in Sardinia. The spatial analysis revealed that the goitre rate (p < 0.0001), the proportion of inhabitants involved in pastoralism (p = 0.016), the terrain inclination (p = 0.008), and the distance from the workplace as a proxy for physical activity (p = 0.023) were consistently associated with population longevity at an aggregated level in the 377 municipalities. Within the limits of an ecological study design, our findings support the existence of a significant association between high goitre prevalence and increased probability to survive into old age. The present study confirms previous results and is consistent with animal studies and epidemiological surveys in other long-lived areas known as Blue Zones. Potential mechanisms underlying this association need to be further investigated.

Electronic supplementary material

The online version of this article (10.1007/s10433-019-00510-4) contains supplementary material, which is available to authorized users.

Introduction

One of the many factors that have been shown to be associated with increased life expectancy in humans, a moderate reduction in thyroid function, has attracted interest over the past decades. In animal models, thyroid gland ablation (Ooka and Shinkai 1986) or a genetic predisposition to produce lower thyroid hormone levels (Buffenstein 2005) generally result in extended lifespan compared with control animals with normal thyroid activity. Long-living humans among the Ashkenazi Jews have been described as having a mildly reduced thyroid function, revealed by elevated circulating levels of thyroid-stimulating hormone (TSH) (Atzmon et al. 2009a, b). Similar results were obtained in the Leiden Longevity Study, where abnormally elevated TSH concentrations and lower free thyroxine levels were found to be associated with increased survival (Rozing et al. 2010a, b). An increase in TSH levels is a marker of mechanisms compensating thyroid function. These findings have been interpreted as indicating that thyroid metabolism plays an important role in the duration of life and, specifically, that a mild hypothyroid state could make individuals capable of surviving into very old age. However, results obtained in animal studies cannot always be applied to humans and, moreover, human reports have been limited so far to a few communities encompassing a small number of long-lived subjects. Thus, the hypothesis of a positive influence of low thyroid function on human survival needs further support and in particular, should be tested in populations where the proportion of long livers is considerably higher. One of these populations lives on the Mediterranean island of Sardinia (Poulain et al. 2004, 2011; Pes and Poulain 2016), which was recognised as one of the world’s longevity hot spots (Poulain et al. 2013).

The research carried out in the past two decades in Sardinia has pointed to important determinants of population longevity, such as physical exercise (Pes et al. 2013, 2018), leisure activities (Fastame et al. 2018), environmental factor (Pes et al. 2013), nutritional factor (Pes et al. 2013, 2015), anthropometric factor (Salaris et al. 2012), genetic factor (Pes et al. 2004), and social context (Poulain et al. 2011). Epidemiological and/or ecological studies investigating the relationship between exceptional longevity and decreased thyroid function markers in Sardinia are missing. A possible approach to investigate this relationship might be based on historical rather than current data, taking into account that factors promoting longevity had probably already come into play in the early stages of the individual's life. Fortunately, data on the distribution of endemic goitre in all the Sardinian municipalities had already been published in the first quarter of the twentieth century, enabling areas with higher and lower goitre prevalence to be outlined (Ottonello 1927; Putzu 1928; Desogus 1938).

In general, endemic goitre has been a problem for centuries on the island, like in other mountainous areas of the European continent including the rest of Italy. In most cases, it has been related to iodine deficiency and, to a lesser extent, to genetically determined autoimmune processes leading to hypothyroidism (Martino et al. 1994). Goitre prevalence in the general Sardinian population partially decreased in the 1990s thanks to preventive health programs (i.e. dietary supplementation with iodised salt) or replacement therapy with L-thyroxine for hypothyroidism (Fiore et al. 2014). Yet goitre did not disappear totally. In actual fact, goitre prevalence was estimated in Sardinia at between 16 and 61% of the general population in the 1980s, with a wide range of variation depending on geographic sites (Aghini-Lombardi et al. 1998). In a previous study of a large cohort of schoolchildren aged 5‒15, a prevalence of 21% was detected (Martino et al. 1994). In most cases, untreated goitre is asymptomatic or associated with a slightly reduced thyroid function (Pinchera et al. 1998), with the exception of thyroid enlargement that produces mechanical engorgement of the neck requiring prompt surgery (Sørensen et al. 2014). As a result, high goitre prevalence in a limited geographic area can serve as a proxy measure for reduced thyroid function in a sizeable percentage of the population. This assumption enables a preliminary assessment of the strength of association between reduced thyroid activity and longevity.

The observation that longevity in Sardinia shows a spatially clustered pattern, centred on the inner mountain area, has prompted the search for potential causal factors mirroring this unique spatial distribution. For instance, a study by Tentoni et al. reported some overlap between longevous areas and propensity for late maternity in Sardinia (Tentoni et al. 2012). The implicit assumption of this approach is that since both phenomena show similar spatial distribution—no matter how different they might be in appearance—they have a priori a non-negligible probability of being causally related.

The objectives of this ecological study were (1) to explore the spatial correlation between the prevalence of endemic goitre in Sardinia in the first half of twentieth century and the emergence of population longevity in the same area nearly half a century later and (2) to compare the strength of the association between goitre and longevity with that of other longevity determinants previously identified in Sardinia.

Methods

Data source

The historical data on endemic goitre in Sardinia analysed in this study were those published by Ottonello in (1927) encompassing the provinces of Sassari, Oristano and Nuoro, and by Putzu who in (1928) mapped the major goitrogenic spots in 56 municipalities out of a total of 256 in the same provinces, as well as in the province of Cagliari (Ottonello 1927; Putzu 1928). These data were further integrated with a list of villages where sporadic goitre was recorded, published by Desogus in (1938). As suggested by Corda (1935), the municipalities not included in this list were considered as having a near-zero goitre rate; eventually, the statistical units used in the analysis comprised all Sardinia’s 377 municipalities.

Measures

The demographic indicator adopted for quantifying longevity in this study was the Extreme Longevity Index (ELI) previously used to measure the longevity level in Sardinian municipalities (Poulain et al. 2004). ELI_x is defined as the probability of a newborn child in a given municipality reaching the threshold age x. The x value may vary depending on the definition of which level of longevity is considered “extreme”, although ELI₁₀₀ is the usual value (Poulain et al. 2011) defined as the ratio between the number of centenarians born in a given municipality between 1880 and 1900 and the number of people born in the same place and the same year of centenarians, regardless of their final residence. The main advantage of ELI₁₀₀, compared with centenarian prevalence, i.e. the number of centenarians relative to the total population, is that (1) it considers a larger number of surviving or already dead centenarians covering 21 birth cohorts from 1880 to 1900 inclusive; (2) it limits potential bias due to migration, as centenarians are counted according to their place of birth and compared with the total number of newborn individuals in the same place; and (3) it is not affected by changes in population fertility during the past century (Poulain et al. 2004). The ELI₁₀₀ value computed for each municipality was also used to define three spatially non-overlapping municipality groups with progressively decreasing ELI₁₀₀ values. More specifically, a restricted Longevity Blue Zone (RLBZ) grouped all municipalities with ELI₁₀₀ above 500 (95th percentile), extended Longevity Blue Zone (ELBZ) with ELI₁₀₀ grouped those ranging from 200 to 500 (75th and 95th percentiles), and a non-Longevity Blue Zone (NLBZ) encompassed the remaining municipalities. In addition to goitre prevalence, other variables collected from historical documents at a municipal level and already analysed as longevity predictors in a previous study (Pes et al. 2013) were included in the statistical models: the proportion of inhabitants involved in animal husbandry (pastoralism) (Fermi 1934), the average daily distance from the workplace (Fermi 1934), the average terrain inclination (Pes et al. 2013), the diet quality (Fermi 1934), the average stature (Fermi 1934), and robustness (Poulain et al. 2017). More specifically, the daily distance for the active population to reach the workplace was considered a proxy of the physical activity during most of working life. Similarly, the terrain inclination, calculated from available Geographical Information System, was considered proportional to lifelong active energy expenditure. The quality of the diet of the population in each village, originally expressed through adjectives (from “very poor” to “very good”) in the Fermi’s database, was re-coded into an ordinal variable ranging from 1 to 5. The average population stature and the average population body robustness reported in the Fermi’s database were derived from military lists and reflected the nutritional status of the population. Finally, the possible correlation between endemic goitre and iodine levels was evaluated by using data published by Costa in samples of drinking waters from Sardinia (Costa 1973).

Statistical analyses

As a first step, nonparametric correlation between the ELI₁₀₀ index, goitre prevalence and the other selected variables was calculated using the Spearman correlation coefficient ρ, taking each Sardinian municipality as the smallest statistical unit. Next, the degree of global spatial autocorrelation was evaluated for each selected variable according to Moran’s I index with inverse distance weighting, using Matlab 7.0 software. Spatial autocorrelation is characterised by a correlation in a signal among adjacent locations in space. Moran’s I is a measure of spatial autocorrelation developed by Patrick Alfred Pierce Moran (Moran 1950) and defined as:

$$I=\frac{N}{{\sum }_i{\sum }_j{w}_{\mathit{ij}}}\frac{{\sum }_i{\sum }_j{w}_{\mathit{ij}}\left(X_i-\overline{X}\right)\left(X_j-\overline{X}\right)}{{\sum }_i{\left(X_i-\overline{X}\right)}^2}$$

where N is the number of spatial units indexed by i and j; X is the variable to be analysed; X_i and X_j are the values of the variable X at sites i and j; $\overline{X}$ is the mean of X; and w_ij represents the spatial weights matrix (N × N) which in the present study was calculated with the procedure described by Chen (Chen 2013). Briefly, the spatial weights matrix describes the relationship between a spatial element and its surrounding elements, and the weights Wij were calculated as the inverse distance between pairs of locations. Moran’s I index assumes positive values when the corresponding variable is spatially clustered and negative values when it is dispersed. Its values were transformed into Z-scores to test the null hypothesis that the spatial autocorrelation of any variable included in the study was zero; when values are greater than 1.96 or smaller than − 1.96, the null hypothesis is rejected, and the variable is said to be spatially autocorrelated at the 0.05 threshold. The next step was to apply a weighted ordinary least squares (WOLS) regression model using the Statistical Package for Social Sciences (SPSS) version 16.0 software (Chicago, US) to assess the association of longevity with goitre prevalence, as reported in historical records, and the other study variables. Regression coefficients were taken as indicators of the magnitude of linear association between the dependent and the independent variables. As weight vector, the cumulated number of newborns in each municipality between 1880 and 1900 was used. Residuals of the WOLS analysis were mapped to check for being autocorrelated or spatially independent.

However, the conventional WOLS analysis considers each municipality as an isolated unit, thereby losing some information due to the spatial relationships existing between adjacent space units—which could lead to violation of the assumption of observations independence and, in turn, to unreliable estimates of model parameters (Haining and Law 2011). For this reason, the association of the longevity index with the potential explanatory variables was subsequently tested by geographically weighted regression (GWR) modelling using GWR 4.0 software, which generates parameter estimates for each unit by borrowing information from the surrounding units (Fotheringham et al. 2002). A key characteristic of GWR is that it takes into account the random spatial effects in the resulting modelling and covariate effects at the ecological level. More specifically, this model may include globally fixed and geographically varying coefficients:

$${\text{ELI}}_i=\sum _k{\beta }_k\left(u_i,v_i\right)x_{k,i}+\sum _l{\gamma }_l{z}_{l,i}+{\epsilon }_i$$

where ELI_i refers to the Extreme Longevity Index at the location i, coefficients β_k (u_i, v_i) are local parameters of the k-th independent variable and vary according to the location of each spatial unit, z_l,i is the l-th independent variable with a fixed coefficient ${\gamma }_l$ and ε_i is the Gaussian error at location i.

Results

Table 1 shows the descriptive statistics of the variables analysed in this study. Variables were stratified according to three spatially non-overlapping areas characterised by low, medium and high longevity according to ELI₁₀₀ calculated for the birth cohorts 1880–1900 (Pes et al. 2013). Overall, goitre prevalence as reported in historical documents was three times higher in the “restricted” Blue Zone where the proportion of centenarians reached a maximum (above 95th percentile of ELI₁₀₀), compared to areas with medium and low longevity. Interestingly, the significantly greater prevalence of goitre in the high-longevity area is evident also in more recent studies conducted in schoolchildren (62% in the restricted Blue Zone compared with 12% in Sardinia as a whole) (Table 1). The iodine content in samples of drinking water is significantly lower in areas of high and medium longevity (2.2 and 2.6 μg/L, respectively) compared to the regional average (4.4 μg/L). As for the other variables, the practice of pastoralism, the terrain inclination, the average distance from the workplace and the body robustness showed significantly higher values in the restricted Blue Zone, confirming the results of our previous study (Pes et al. 2013), whereas the diet score and the body height did not show significant differences across the various areas.

Table 1.

Longevity index, prevalence of endemic goitre according to historical data (see references) and schoolchildren survey, and some environmental and population variables in the 377 villages of Sardinia, stratified by longevity areas

Study variables	RLBZ1	ELBZ2	NLBZ3	Total Sardinia	References
No. of municipalities	16	176	185	377
No. centenarians	98	604	306	1008	Poulain et al. (2004)
ELI100 (persons/105)	553*	263	72	183	Pes et al. (2013)
Prevalence of visible endemic goitre estimated from historical documents of 20th century, (%)	0.69%**	0.25%	0.10%	0.20%	Ottonello (1927) Putzu (1928) Desogus (1938)
Prevalence of endemic goitre in schoolchildren 1986–1994, (%)	62%	32%	8%	12%	Martino et al. (1994) Loviselli et al. (2001)
Iodine in drinking water (µg/L)	2.2 ± 1.0***	2.6 ± 1.8	7.1 ± 5.7	4.4 ± 4.4	Costa (1973)
Pastoralisma
No shepherds	25.0%	31.3%	61.1%	45.6%	Fermi (1934)
Farmers > shepherds	0.0%	1.7%	15.1%	8.3%
Shepherds > farmers	12.5%	36.9%	14.6%	24.9%
Only shepherds	62.5%	30.1%	9.2%	21.2%
Distance to workplace, kma	10.8	10.6	6.8	8.7	Fermi (1934)
Terrain inclination, (%)	20.9	14.7	11.1	13.2	Pes et al. (2013)
Diet scorea
Very poor	18.8%	13.1%	7.6%	10.6%	Pes et al. (2013)
Poor	50.0%	52.3%	74.6%	63.1%
Average	31.2%	15.9%	14.0%	15.7%
Good	0.0%	11.3%	3.8%	7.2%
Very good	0.0%	7.4%	0.0%	3.4%
Average population stature, cma	162.1	163.0	162.9	162.9	Fermi (1934)
Average population body robustnessa	2.38 ± 0.50***	1.90 ± 0.83	1.40 ± 0.97	1.68 ± 0.93	Fermi (1934)

¹RLBZ, restricted LBZ; ²ELBZ, extended LBZ; ³NLBZ, non-LBZ as according to reference Pes et al. 2013

*ANOVA, p < 0.0001; ** ANOVA, p < 0.0001; ***Student’s t test, p = 0.007 (restricted + extended LBZ vs non-LBZ)

^aThe quality of the diet of the population in each village, originally expressed through adjectives (from “very poor” to “very good”), was re-coded into an ordinal variable ranging from 1 to 5

Table 2 shows the spatial autocorrelation of the variables analysed in this study and their correlation with ELI₁₀₀. Moran’s I values indicated that the distribution of all variables, including endemic goitre, in all Sardinian municipalities displayed significant positive spatial autocorrelation with the exception of distance to workplace, diet score and average population stature. Spearman’s correlation coefficients between ELI₁₀₀ (per 100,000 in the period 1980‒2000) and the other potential explanatory variables selected in the analysis are also reported in Table 2. Goitre rate showed a significant correlation with ELI₁₀₀ (ρ = 0.210; p < 0.0001) preceded in size only by pastoralism (0.272; p < 0.0001) and average terrain inclination (0.263; p < 0.0001). A correlation matrix for all the selected variables is reported in the supplementary material.

Table 2.

Spatial autocorrelation of study variables and correlation with extreme longevity Index₁₀₀ (person/10⁵) in the 377 villages of Sardinia and some environmental and population variables

Variables	Spatial autocorrelation		Correlation with ELIa100
	Moran’s I	p value	Spearman ρ	p value
ELIa100	0.0753	0.025	1.000	‒
Prevalence of endemic goitre in historical documents 20th centuryb	0.0860	0.013	0.210	< 0.0001
Pastoralismc	0.1113	0.002	0.272	< 0.0001
Distance to workplacec	0.0513	0.088	0.182	< 0.0001
Terrain inclination	0.0809	0.018	0.263	< 0.0001
Diet scorec	0.0386	0.151	− 0.044	0.401
Average population staturec	0.0614	0.054	0.116	0.026
Average population body robustnessc	0.0910	0.009	0.109	0.036

^aThe Extreme Longevity Index in each municipality was computed as: ELI = 10,000*C/B, where C is the number of people born in that municipality between 1880 and 1900 who reached age 100 and B is the number of births in the same cohorts of centenarians, in the same municipality and in the same time interval

^bData as defined in references: Ottonello (1927), Putzu (1928) and Desogus (1938)

^cData as defined in reference: Fermi (1934)

Figure 1 shows the geographic distribution of crude prevalence of goitre in Sardinia according to historical documents. Two major clusters of significantly increased prevalence are clearly evident (red shaded), the largest of which matches the high-longevity area (Blue Zone) as shown in Fig. 1b and very similar to the distribution of goitre rate reported in Sardinian schoolchildren in 1986 (Loviselli et al. 2001). As illustrated in Fig. 1a, goitre prevalence in Sardinia in 1938 was highest in some of the areas where the population, half a century later, displayed the Longevity Blue Zone phenomenon. The epicentre of goitre as well as longevity can be located in the central‒eastern area of the island, named Ogliastra.

Fig. 1.

a Geographic distribution of endemic goitre in Sardinia in the first half of twentieth century (data not smoothed); b Geographic distribution of ELI₁₀₀ in Sardinia between 1880 and 1900 (data not smoothed)

The model summary for WOLS and GWR analysis is reported in Table 3. Goitre is a significant predictor of longevity (p < 0.0001): a β coefficient of 74.3 represents the expected increase of the conditional mean of ELI₁₀₀ for every one-unit increase of goitre prevalence. Also, pastoralism (p = 0.016), terrain inclination (p = 0.008) and physical activity (p = 0.023) were significant predictors, as already reported in our previous ecological study (Pes et al. 2013). The independent variables included in this study explained about 16% of the variance (12% excluding goitre) in ELI-estimated spatial longevity.

Table 3.

Parameters of weighted ordinary least squares and geographically weighted regression models

	Weighted ordinary least squares			Geographically weighted regressiona
	ß	St. ß	p value	Mean	SD
Constant	− 47.546	‒	‒	181.52	23.43
Prevalence of endemic goitre	74.359	0.191	< 0.0001	35.36	11.02
Pastoralism	19.626	0.128	0.016	20.12	6.81
Distance to workplaceb	3.712	0.117	0.023	19.19	9.02
Terrain inclination	3.907	0.138	0.008	27.64	4.67
Diet Scorec	1.250	0.006	0.911	− 2.53	6.82
Average population stature	22.375	0.071	0.314	12.48	8.95
Average population body robustness	− 1.702	− 0.008	0.906	3.47	6.06
Adjusted R2	0.16			0.25
AICd	4979			4964

^aEstimated coefficients of geographically varying (local) terms

^bConsidered as a proxy for physical activity

^cTreated as a continuous variable

^dAkaike information criterion

The GWR analysis further confirmed these results. The summary of the GWR model results is reported in Table 3. When the model was fitted with variables as local terms, the coefficient of the goitre was the highest (35.3 ± 11.0), followed by the terrain inclination (27.6 ± 4.6) and pastoralism (20.1 ± 6.8) indicating that these variables had a better explanatory capacity after adjustment for autocorrelation. The spatial distribution of the local coefficients in the GWR model is illustrated in Fig. 2. Their highest significant values were found for goitre, pastoralism, terrain inclination and body stature, indicating that these variables were better predictors in the central‒eastern areas of the island where longevity is prevalent. In contrast, the coefficients for diet and body robustness did not show significant high values in correspondence of the high longevity area. Although the GWR model showed better explanatory capacity than WOLS model (the Akaike information criterion dropped from 4979 to 4964), the residual scatterplot clearly evidenced clustering, suggesting that still unmeasured factors may have contributed to the geographic pattern of longevity.

Fig. 2.

Spatial distribution of the local coefficients in the geographically weighted Poisson regression model (smoothed data)

Discussion

The population living in Sardinia, especially in the mountain area, has been recognised as one with the highest centenarians’ rate in the world (Poulain et al. 2004; Pes et al. 2013; Pes and Poulain 2016). Therefore, there is growing interest in investigating the causes of its exceptional longevity. In order to understand further the possible determinants of the phenomenon, endemic goitre was included as an additional variable in a preexisting ecological dataset for testing its association with longevity. A visual inspection of the spatial distribution maps of both goitre and longevity shows a clear overlap between the two phenomena, further confirmed by correlation and regression analyses. Previous reports correlated pastoralism, terrain inclination and average daily distance to workplace to longevity. In the present study, these variables were confirmed to be associated with survival into old age. However, the true novelty of our findings is that goitre was an independent predictor of longevity after adjusting for potential confounders. In particular, the largest goitrogenic area matches municipalities where the longevity index between 1980 and 2001 reached its highest magnitude (Longevity Blue Zone). Goitre prevalence proved to be an ecological variable, the association of which with the exceptional longevity index in Sardinia is comparable with social variables such as the proportion of inhabitants involved in pastoralism. This finding raises the question of a possible causal relationship between the two phenomena.

Sardinia was renowned for centuries for its high incidence of endemic goitre. Already in the first half of the last century, a number of pioneering studies were published addressing the spatial distribution of the condition at a micro-geographic level (Ottonello 1927; Putzu 1928; Desogus 1938). These early studies pointed out that goitre was not uniformly spread throughout the island but was rather localised in limited areas, especially amongst the poorest social groups (Ottonello 1927), and where naturally occurring iodine was relatively low.

From a speculative point of view, one might ask what are the causes responsible for the geographic correlation between goitre and longevity.

One hypothesis might imply changes in thyroid activity through inadequate iodine intake: (1) iodine content in soil, drinking water, and foods was remarkably low in the mountain areas of Sardinia. A study of several insular localities found that iodine levels in water were as low as 4 µg/L (Atzeni et al. 1989), whereas in the World Health Organization reports values were higher than 8 µg/L. This iodine shortage in the water supplies was the cause of widespread thyroid deficiency in the population, especially in times when a public water supply was not available; (2) although this population was close to the sea, before nutrition transition began in the 1960s the average intake of seafood was virtually negligible (Pes et al. 2015) and the main sources of iodine were milk and dairy products. However, dairy food, which made up a large part of the diet in the high longevity area, was very low in iodine (average values 23.02 μg/100 g in sheep’s cheese and 21.71 μg/100 g in goat’s cheese; normal values above 50 μg/100 g) (Pulina et al. 2008). For this reason, iodine supplementation trials with sheep to improve milk quality have been undertaken recently (Nudda et al. 2009).

In pre-transitional Sardinia, iodine deficiency could have also been exacerbated by the high dietary content of calcium and vitamin D, which enhance iodine requirements by reducing the action of thyroid hormones on target tissues, as suggested by experiments in both rodents (Etling et al. 1986) and humans (Boyle et al. 1966); (3) although iodised salt supplementation has been a trend in Italy since 1938 (Ambrosi 1941), the Sardinian population has long resisted the adoption of such preventive measures (Peretti 1943). Even when iodisation became mandatory in 2005 by national law (Pastorelli et al. 2015), Sardinia remained the region with the lowest iodine consumption in Italy (Olivieri and Vitti 2014). Overall, the daily intake of iodine among Sardinians was far lower than the average daily requirement of 150 µg for adults as recommended by the World Health Organisation/International Council for the Control of Iodine Deficiency Disorders. In addition, widespread L-thyroxine administration by general practitioners to treat subclinical hypothyroidism around 1990 resulted in a partial reduction in goitre rate. In spite of these measures, the phenomenon did not fully disappear, as highlighted by a 21% goitre prevalence in a study carried out between 1986 and 1994 on some 8000 school children aged 6‒15 years from all the island provinces. It should be noted, however, that before the introduction of ultrasonography goitre could be detected only by visual inspection and neck palpation; this explains the much lower estimated prevalence in historical documents (Nordmeyer et al. 1997; Anwar et al. 2015). A further study in Sardinia showed a low urinary iodine excretion (0 ‒ 108 mg/L, N.V. > 100) (Loviselli et al. 2001) which suggested the persistence of local causes of the hypothyroidism up until the present. The practice of prolonged breastfeeding, too, being frequent in the Sardinian Longevity Blue Zone, may have contributed to the worsening of a condition of maternal hypothyroidism (Speller and Brodribb 2012).

The evidence of a spatial overlap between areas of goitre and population longevity—observed in our study—might indicate that the association may be more than a simple coincidence. As previously suggested, the most reasonable hypothesis is that iodine deficiency modifies thyroid physiology, inducing a general reshaping of energy metabolism with the further result of extended lifespan. Although overt hypothyroidism is usually associated with adverse health outcomes (Ladenson 2008) such as increased cardiovascular risk in adults, this was rarely observed in subclinical hypothyroidism (Laulund et al. 2014; Van den Beld et al. 2005). Moreover, there is growing evidence that a mild degree of down-regulation of thyroid function exerts a beneficial effect during ageing, representing a protective factor (Aggarwal and Razvi 2013). The lower thyroid function may lead to a lower resting metabolic rate which is associated with improved survival in animal models (Buffenstein 2005). As suggested by Van den Beld et al. (2005), a lower metabolism could trigger a cascade of events ending in reduced production of reactive oxygen species, thereby limiting the damage to intracellular macromolecules such as DNA and proteins in the ageing organism. The existence of a spatial association between goitre and increased longevity, however, does not imply that the latter is a healthy longevity: future research will help clarifying this important issue.

From a worldwide perspective, the finding of an epidemiological association between goitre prevalence and longevity does not seem to be limited to Sardinia, but partially shared also by other populations where long-lived subjects are numerous. Most of the Longevity Blue Zones in the past were niches of endemic goitre as well. In Costa Rica, where the Nicoya LBZ was identified in 2007 (Rosero-Bixby 2008), high prevalence of endemic goitre has been reported since the 1950s (Perez et al. 1956) possibly aggravated by a gross excess of calcium ingested with drinking water (Boyle et al. 1966). In another LBZ, Ikaria island, the iodine level in spring water is remarkably low (Pes and Poulain 2014).

This study has some limitations. Although the extreme longevity index (ELI), calculated as the probability of a newborn (zero age) to become a centenarian, has been used in the previous study by the same research group (Pes et al. 2013), it is influenced by infant mortality, juvenile mortality and adult mortality, thus reflecting only in part the chance to reach exceptional old age. In addition, the present study included only ecological variables in the regression models, and therefore, its results cannot be extended to individual subjects. As a matter of fact, our ecological approach did not allow to test a direct cause‒effect relationship and should be regarded merely as exploratory or hypothesis generation (Greenland and Morgenstern 1989; Greenland 2001). The geographically weighted regression analysis was performed because most of the variables selected showed some degree of autocorrelation and we wanted to avoid a potential bias by the ordinary least square regression. The analysis we carried out solved partly this problem, but also revealed that a sizable proportion of residual variance remains unexplained.

Conclusions

Taking into consideration the limits of ecological studies and in the absence of detailed clinical data on the population under evaluation, our analysis nonetheless provides a further support to potential goitre/longevity association entailing possible clinical consequences.

In the past, older people with goitre were frequently overtreated with L-thyroxine with the goal of reducing long-term morbidity and mortality. Goitre may be euthyroid, but it may also co-occur with subclinical hypothyroidism. A recent meta-analysis did not show significant benefits in the hormone therapy of this condition regarding thyroid-related symptoms and general quality of life (Feller et al. 2018). Therefore, if the conclusions of our study will be supported by clinical evidence, it would be better to reconsider the approach to patients with subclinical hypothyroidism, which at the moment is not worthy of treatment by international guidelines.

Electronic supplementary material

Below is the link to the electronic supplementary material.

Compliance with ethical standards

Conflict of interest

The authors declare that they have no conflict of interest.