Covariation of the incidence of type 1 diabetes with country characteristics available in public databases - PubMed (original) (raw)

Covariation of the incidence of type 1 diabetes with country characteristics available in public databases

Paula Andrea Diaz-Valencia et al. PLoS One. 2015.

Abstract

Background: The incidence of Type 1 Diabetes (T1D) in children varies dramatically between countries. Part of the explanation must be sought in environmental factors. Increasingly, public databases provide information on country-to-country environmental differences.

Methods: Information on the incidence of T1D and country characteristics were searched for in the 194 World Health Organization (WHO) member countries. T1D incidence was extracted from a systematic literature review of all papers published between 1975 and 2014, including the 2013 update from the International Diabetes Federation. The information on country characteristics was searched in public databases. We considered all indicators with a plausible relation with T1D and those previously reported as correlated with T1D, and for which there was less than 5% missing values. This yielded 77 indicators. Four domains were explored: Climate and environment, Demography, Economy, and Health Conditions. Bonferroni correction to correct false discovery rate (FDR) was used in bivariate analyses. Stepwise multiple regressions, served to identify independent predictors of the geographical variation of T1D.

Findings: T1D incidence was estimated for 80 WHO countries. Forty-one significant correlations between T1D and the selected indicators were found. Stepwise Multiple Linear Regressions performed in the four explored domains indicated that the percentages of variance explained by the indicators were respectively 35% for Climate and environment, 33% for Demography, 45% for Economy, and 46% for Health conditions, and 51% in the Final model, where all variables selected by domain were considered. Significant environmental predictors of the country-to-country variation of T1D incidence included UV radiation, number of mobile cellular subscriptions in the country, health expenditure per capita, hepatitis B immunization and mean body mass index (BMI).

Conclusions: The increasing availability of public databases providing information in all global environmental domains should allow new analyses to identify further geographical, behavioral, social and economic factors, or indicators that point to latent causal factors of T1D.

PubMed Disclaimer

Conflict of interest statement

Competing Interests: The authors have declared that no competing interests exist.

Figures

Fig 1

Fig 1. Correlations between T1D incidence and 77 country indicators.

The correlations were computed in the 80 WHO countries where T1D incidence could be estimated. Dots: no significant correlations. Squares: significant correlations with _p≤_0.2. Significant correlations after Bonferroni correction (_p_-value ≤ 0.000649) are highlighted. Abbreviations: PM: particular matter, UV: ultraviolet, GDP: Gross Domestic Product, GNI: Gross National Income, BMI: Body Mass Index, CVD: Cardiovascular Diseases, CA: Cancer, DM: Diabetes Mellitus, CHRD: Chronic Respiratory Diseases, Adj. R2: Adjusted-R2. Red: positive correlations, blue: negative correlations. (a) Age-standardized estimate; (b) Per 100,000 individuals; (c) Coverage among 1-year-olds (%). See S1 Database for the entire database.

Fig 2

Fig 2. Stepwise identification of predictors of T1D Incidence (a) by domain and (b) final.

The lines are the 35/77 variables with p<0.2. Green bars indicate variables that were excluded during the stepwise multiple regression (the length of the bar indicates the number of steps at which it was excluded). Blue bars indicate that variables were selected after SMLR was applied. The dark blue bars indicate that the independent predictors of T1D were highly significant; panel (A): in the by-domain analysis; panel (B): in the final model. The final analysis was performed on the variables selected in the by-domain analysis (shown in dark blue in panel (A). Abbreviations: Dom: Domain, PM: particular matter, UV: ultraviolet, GDP: Gross Domestic Product, BMI: Body Mass Index, CVD: Cardiovascular Diseases, CA: Cancer, DM: Diabetes Mellitus, CHRD: Chronic Respiratory Diseases, Adj.R2: Adjusted-R2. (a) Variable dropped after applied Akaike information criterion (AIC).

Fig 3

Fig 3. The predicted incidence of T1D among 80 countries vs. the observed incidence.

Red dot: Finland. See S2 Fig. for predicted indices obtained after 10-fold Cross-validation.

Similar articles

Cited by

References

    1. Harjutsalo V, Sund R, Knip M, Groop PH. Incidence of type 1 diabetes in Finland. JAMA: the journal of the American Medical Association. 2013;310(4):427–8. 10.1001/jama.2013.8399 - DOI - PubMed
    1. Ogle GD, Lesley J, Sine P, McMaster P. Type 1 diabetes mellitus in children in Papua New Guinea. P N G Med J. 2001;44(3–4):96–100. - PubMed
    1. Frongia O, Mastinu F, Sechi GM. Prevalence and 4-year incidence of insulin-dependent diabetes mellitus in the province of Oristano (Sardinia, Italy). Acta Diabetol. 1997;34(3):199–205. - PubMed
    1. Garancini P, Gallus G, Calori G, Formigaro F, Micossi P. Incidence and prevalence rates of diabetes mellitus in Italy from routine data: a methodological assessment. Eur J Epidemiol. 1991;7(1):55–63. - PubMed
    1. Rewers M. Challenges in Diagnosing Type 1 Diabetes in Different Populations. Diabetes Metab J 2012;36:90–7. 10.4093/dmj.2012.36.2.90 - DOI - PMC - PubMed

Publication types

MeSH terms

Grants and funding

This work was supported by grants from the Programme Hospitalier de Recherche Clinique and from Colciencias, the Administrative Department of Science, Technology and Innovation for Colombia. The funders had no role in study design, data collection and analysis, decision to publish, or preparation of the manuscript.

LinkOut - more resources