Study design, setting and population
The Di@bet.es Study is a national, cross-sectional, population-based study which was conducted in 2008-2010 using a random cluster sampling of the Spanish population [16]. The study sample consisted of 5072 not pregnant adults (> 18 years), randomly selected from the National Health System registries distributed into 110 clusters (primary health care centers). Thyroid function studies [thyroid-stimulating hormone [TSH], free thyroxine (FT4), free triiodothyronine (FT3) and thyroid peroxidase antibodies (TPO Abs)], were performed in 90% of this sample [17]. Individuals with missing data did not differ in age, sex or other characteristics of interest from individuals with complete data.
For the present analysis, we excluded all the subjects with a previous thyroid disease diagnosis, and/or taking interfering medications (levothyroxine, thionamides, amiodarone or lithium). We also excluded individuals with a positive TPO Abs test (≥ 50 IU/ml), or with very high (>20 mIU/L) or suppressed (<0.1 mIU/L) TSH levels.
This research was carried out in accordance with the Declaration of Helsinki of the World Medical Association [18]. Written informed consent was obtained from all the participants. The study was approved by the Ethics and Clinical Investigation Committee of the Hospital Regional Universitario de Málaga (Málaga, Spain) in addition to other regional ethics and clinical investigation committees all over Spain.
Variables and procedures
The participants were invited to attend a single examination visit at their health center. Information was collected by means of an interviewer administered structured questionnaire, followed by a physical examination and blood sampling. Information on age, gender, educational level (none/basic/high school/college), and smoking habit (current, former or never been smokers) was obtained by questionnaire. Medical history and medications were also recorded. Menopause was considered in women who reported more than 12 months of amenorrhea without any other obvious pathological or physiological cause. Weight and height were measured and the body mass index (BMI) was calculated by standardized methods.
Blood samples were obtained in fasting conditions, were immediately centrifuged and the serum was frozen until analysis. A casual urinary sample was also collected and samples were frozen until analysis. Samples were managed by the biochemistry laboratory of the Hospital Regional Universitario de Málaga, the IBIMA Biobank and by the CIBERDEM Biorepository (IDIBAPS Biobank). TSH, FT4, FT3 and TPO Abs concentrations were analyzed using an electrochemiluminescence immunoassay (Modular Analytics E170, Roche Diagnostics, Basel, Switzerland). The functional sensitivity of the TSH assay was 0.014 mIU/L. The intra-assay coefficients of variation were: TSH, 1.5–1.2%; FT4 1.8–1.6%; FT3 1.3–2.0% and TPOAb 4.8–2.8%. The inter-assay coefficients of variation for the low and high levels of serum TSH, FT4, FT3 and TPO Abs quality control materials were 3.5 and 2.7%, 4.17 and 2.64%, 3.78 and 2.21%, and 8.5 and 5.2%, respectively. All samples were analyzed at the biochemistry laboratory of the Hospital Regional Universitario de Málaga. Urinary iodine (UI) was analyzed using the modified method of Benotti and Benotti [19]. The intra and inter-assay coefficients of variation of UI assay were 2.01% and 4.53%, respectively. The UI assay was subjected to a program of external quality assessment for the determination of iodine in urine of the Spanish Association of Neonatal Screening (AECNE) and of Ensuring the Quality of Iodine Procedures (EQUIP) Program. All UI samples were analyzed in the Research Laboratory of the Hospital Regional Universitario de Málaga (Spain).
Exposure assessment
Modeled mean annual PM2.5 and NO2 concentrations in Spain for the period 2008 to 2010 were calculated with the CHIMERE chemistry-transport model [20]. This model calculates the concentration of gaseous species and both inorganic and organic aerosols of primary and secondary origin, including primary particulate matter, mineral dust, sulphate, nitrate, ammonium, secondary organic species and water. This model has been broadly evaluated in Spain by comparison with measured air pollutants at a large set of monitoring sites [21, 22]. The model was applied to a domain covering the Iberian Peninsula at a horizontal resolution of 0.1x0.1º. The modelled concentrations were corrected with observed values, by considering a methodology described by Martín et al [23] in which 1) a bias is calculated with regard to the observations in the Spanish air quality network of monitoring sites, 2) these biases are spatially interpolated using a krigging methodology to obtain a gridded bias, and 3) this gridded bias is applied to the modelled concentration grid. This methodology considers a different bias grid for rural and urban sites that are then combined and weighted by population density. We assigned the average annual exposure to air pollutants corresponding to the health examination year of each participant by interpolating the estimated concentrations to the centroid of their residential postal codes.
Data on mean annual temperature (°C) from each municipality of residence were obtained from the Spanish National Meteorological Agency [24]
Statistical analysis
We applied linear regression models to assess associations between air pollutant and thyroid hormone levels (TSH, FT4, FT3), which were log transformed to normalize distributions and also to limit the influence of extreme values. First, we explored any indications of non-linearity in the associations using Curve Estimation procedures. None of the models tested improved the linear model and there was no indication of a threshold effect. Association estimates were presented as percent changes with corresponding 95% confidence intervals (calculated by 100 × [exp(b)-1]), per each interquartile range (IQR) increase in air pollutant concentrations which equated to 4.86 μg/m3 PM2.5, and 12.62 μg/m3 NO2.
We also used logistic regression models to investigate the associations of ambient air pollutants with high TSH levels (defined as a percentile (p) >95), low FT4 (≤p5) and low FT3 (≤p5). These results are presented as odds ratios (ORs) with corresponding 95% CIs again per each interquartile range (IQR) increase in air pollutants.
All these models were controlled for possible confounders such us age, sex, UI, BMI, education level, smoking status and ambient temperature. In addition, we performed subgroup analyses to test potential effect modifications in subgroups according to sex and menopausal status. Homogeneity of the ORs between subgroups was tested with the Breslow-Day test. All the statistical analyses were performed with IBM SPSS statistics 23.0. Reported p values were based on two-sided tests with statistical significance set at 0.05.