Circulating Levels of Phthalate Metabolites Are Associated With Prevalent Diabetes in the Elderly

Eligible for the PIVUS study were all subjects aged 70 years living in the community of Uppsala, Sweden. The subjects were randomly chosen from the register of community living. A total of 1,016 subjects participated during 2001–2004, giving a participation rate of 50.1%. The study was approved by the ethics committee of Uppsala University.

All subjects were investigated in the morning after an overnight fast. No medication or smoking was allowed after midnight. The subjects were asked to answer a questionnaire about their medical history, educational level, exercise habits, smoking habits, and regular medication. Educational level was divided into three groups (<9, 9–12, and >12 years). Exercise habits were divided into four groups (less than two times per week of light exercise [no sweat], two or more times per week of light exercise, one to two times per week of heavy exercise [sweat], and more than two times per week of heavy exercise). Venous blood samples were collected and stored at −70°C until analysis. Lipid variables and fasting plasma blood glucose were measured using standard laboratory techniques. Basic characteristics of subjects are given in Table 1.

Diabetes was defined as a history of diabetes or a fasting glucose value >7.0 mmol/L. In the total sample, 119 subjects showed diabetes. Of those, 88 had a history of diabetes. In that group, the mean duration of diabetes was 8.9 years (SD 7.7). Only four subjects reported a diabetes duration >20 years.

Human serum (0.5 mL) was analyzed for levels of 10 phthalate metabolites using an isotope liquid chromatograph/tandem mass spectrometer (API 4000LC-MS/MS) at ALS Environmental Canada (http://www.alsglobal.com/environmental.aspx) following the general procedures presented by the Centers for Disease Control and Prevention as previously described in detail (26). In brief, quality control of the analysis was maintained by analyzing a method blank (calf serum) and two spiked calf serum samples (20 ng/mL, all analytes) along with every 17 samples. The limit of detection (LOD) was 0.2 ng/mL. Four of 10 phthalate metabolites, namely mono(2-ethylhexyl) phthalate (MEHP), monoethyl phthalate (MEP), monoisobutyl phthalate (MiBP), and monomethyl phthalate (MMP), were detectable in all but 5–12 subjects (at least 96% of subjects). The fact that some subjects showed undetectable levels rules out a general contamination of these compounds. Only the four metabolites with detectable levels were used in the statistical analysis. For the rest of the metabolites, 31–100% of the observations were below the LOD. The measured serum concentrations of MEHP, MEP, MiBP, and MMP are given in Table 1. The analysis of the phthalates took place 5–8 years following the collection of the samples.

At the laboratory of the Department of Public Health and Caring Sciences/Geriatrics, University Hospital, Uppsala, plasma proinsulin and insulin concentrations were determined using the Proinsulin ELISA and the Insulin ELISA immunoassays (Mercodia, Uppsala, Sweden) on a Bio-Rad Coda automated EIA analyzer (Bio-Rad Laboratories, Hercules, CA).

The ratio of fasting proinsulin to insulin was used as an index of insulin secretion, since this ratio increases with failing β-cell function (27,28). The homeostasis model assessment-insulin resistance (HOMA-IR), as an index of insulin resistance, was calculated as fasting insulin × glucose/22.5 (29).

All four phthalate metabolites, as well as the indices of insulin secretion and resistance, were skewed toward high levels but were normally distributed following ln-transformation.

Relationships between phthalate metabolites and prevalent diabetes were evaluated by logistic regression models, first using the phthalate metabolites as continuous variables and thereafter following division of the phthalate metabolites into quintiles. For the continuous analysis, two steps of adjustments were used: 1) adjustment for sex only and 2) multiple adjustments for sex, serum cholesterol and triglycerides, BMI, smoking and exercise habits, and educational levels. In the quintile analysis, P values are given for linear trend analysis as well as for quadratic trend. In the quintile analysis, only multiple-adjusted P values are given.

A similar approach was used when relating phthalate metabolites to either the proinsulin-to-insulin ratio or to HOMA-IR index, except that linear regression was used. In this analysis, only nondiabetic subjects were included. General additive models were used (30) to visualize the relationship between some of the phthalate metabolites and prevalent diabetes in a continuous fashion in Supplementary Fig. 1. These models were adjusted for the same covariates as the multiple-adjusted models described above.

Mean values or proportions of the established risk factors are shown in Table 1. Relationships between the four evaluated phthalate metabolites are given in Supplementary Table 1. MMP was significantly related to the other phthalate metabolites (although inversely compared with MEHP). MiBP was significantly related to MEHP but not significantly so compared with MEP. MEP and MEHP were not significantly related.

When the four phthalate metabolites were related to prevalent diabetes in logistic regression models using continuous variables, both MEP and MiBP were significantly related to diabetes following adjustment for sex only. Following multiple adjustments, however, only MiBP levels were significantly related to diabetes (odds ratio [OR] 1.30 [95% CI 1.10–1.55], P = 0.0025; for details see Table 2). In Supplementary Table 2, details on different levels of adjustments are presented.

No significant interactions between the four phthalate metabolites and sex, obesity, smoking, and education levels were detected when interaction terms between the phthalate metabolites and these four confounders were introduced in the models. A sensitivity analysis for these four confounders is presented in Supplementary Table 3. Of note is that the OR for MiBP is higher in men than women (1.48 [95% CI 1.19–1.85] in men and 1.08 [0.80–1.46] in women). However, the interaction term between MiBP and sex regarding prevalent diabetes is not significant (P = 0.11).

In quintile analysis, MEP and MiBP, as well as MMP, showed a significantly elevated prevalence of diabetes in the highest quintile compared with the lowest (OR 2.00–2.50), and the linear trend tests were significant. Furthermore, for MEP the quadratic trend test was significant. The highest OR was seen in the third quintile (OR 2.87 [95% CI 1.37–6.03]; for additional details see Table 3).

MEHP and MiBP were significantly related to a high proinsulin-to-insulin ratio, both following sex adjustment only and following multiple adjustments, respectively. MEP was weakly related to the proinsulin-to-insulin ratio in an inverse way in linear regression models using continuous variables (Table 4). In quintile analysis, the quadratic trend test for MIBP was significant, being consistent with the finding that the highest proinsulin-to-insulin ratio was seen in the third quintile (for details see Supplementary Table 4).

MEP and MMP were significantly related to a high HOMA-IR index, both following sex adjustment only and following multiple adjustments, respectively (Table 4). In quintile analysis, the quadratic trend test for MEP was significant, being consistent with the finding that most of the effect on HOMA-IR was seen in the third quintile (for details see Supplementary Table 5).

In the analysis of HOMA-IR and the proinsulin-to-insulin ratio, no diabetic subjects were included because the diabetic state or pharmacological treatment for diabetes may affect the proinsulin and insulin measurements.

The current study showed that some of the measured circulating phthalate metabolites, MEP, MMP, and MiBP, were associated with prevalent diabetes in the present cross-sectional study. Furthermore, MiBP levels were inversely related to a marker of poor insulin secretion, whereas MMP and MEP levels were related to insulin resistance, two major factors involved in regulation of an impaired glucose metabolism and independent risk factors for type 2 diabetes development.

In a recent cross-sectional study of Mexican women, self-reported diagnosis of diabetes was related to several phthalate metabolites (25). However, neither MEP nor MiBP was found to be significantly related to diabetes in the Mexican study, and MMP was not evaluated. The current study has the advantage that three times as many diabetic patients were included. Furthermore, we used the combination of self-reported diabetes and measurements of fasting glucose and/or use of pharmacological treatment of diabetes to define prevalent diabetes. The addition of fasting glucose measurements resulted in the addition of another 31 diabetic case subjects to the 85 subjects that reported a history of diabetes. Thus, if self-reported diagnosis only and no fasting glucose measurements had been used to define cases, 26% of the diabetes cases would have been misclassified and observations made would have been diluted.

In the sensitivity analysis in the current study, a higher OR for MiBP regarding diabetes was found in men compared with women (1.48 vs. 1.08). However, the interaction term between MiBP and sex regarding diabetes was not significant, so future studies have to determine whether this observed relationship between MiBP and diabetes really differs between men and women.

In an evaluation of male participants in the National Health and Nutrition Examination Survey 1999–2002, high levels of MBP, monobenzyl phthalate (MBzP), and MEP were associated with insulin resistance (23). Thus, the finding of a relationship between MEP and a high HOMA-IR index is consistent between the studies. MBP and MBzP were measured in the PIVUS study, but because only a small fraction of the participants showed detectable levels in the circulation, we were not able to perform any meaningful evaluation of these phthalate metabolites. To the best of our knowledge, no other study has investigated whether phthalate metabolites are related to markers of insulin secretion.

Because the parent phthalates are so abundant, it is a hopeless task to measure those compounds without major contamination. Therefore, usually their metabolites, not abundant in the environment, are measured in the circulation or in the urine (31). The majority of previous studies on human phthalate exposure is compromised on urine testing. Measurements in the urine have the advantage that higher levels are found and thereby more metabolites could be properly detected. Thus, a limitation of the current study is that only four phthalate metabolites could be investigated in detail with regard to diabetes.

In the current study, MiBP, MMP, MEP, and MEHP were detectable in almost all analyzed serum samples. The common feature of the metabolites MEP, MiBP, and MMP is that all these three metabolites are derived from degradation of associated phthalate parent compounds (diethyl phthalate, di-isobutyl phthalate, and dimethyl phthalate, respectively) used as solvents/carriers of fragrances used in personal care products. Di-2-ethyl phthalate, with the metabolite MEHP, is on the other hand mainly used as plasticizer to make plastic compounds more flexible.

In a Danish study, the correlations between levels of 13 metabolites in different matrices (urine, semen, and serum) were examined in 60 young men. Both MEP and MiBP levels were correlated in serum and urine, indicating that serum levels could be used as biomarkers of human exposure (1). In the Danish study, MiBP, MEHP, MBzP, mono-(2-ethyl-5- oxohexyl) phthalate, and mono-(2-ethyl-5-hydroxyhexyl) phthalate were found in lower levels and in fewer samples, but the LOD was higher than in the current study. Frederiksen et al. (1) found higher levels and a detection rate of monoisononyl phthalate, whereas mono-n-octyl phthalate was not detectable in any of their samples. For these metabolites, their LOD was comparable with the current study.

Sources of exposure of these compounds and comparisons with the levels in other studies are described in more detail in our previous recent publication by Olsén et al. (26) and at http://www.atsdr.cdc.gov/.

Because phthalate metabolites are known ligands to PPARs (13–19), receptors known to influence glucose homeostasis, impairments in PPAR-signaling pathways are most likely to contribute to the actions of phthalates on glucose metabolism and diabetes development. PPAR activation has been shown to be involved in different steps in glucose homeostasis, such as influence on insulin resistance (22) and on insulin secretion (32), influence on circulating levels of lipids (21), and altering the amount of visceral and subcutaneous fat (21,33). Additional experimental studies on phthalates have to be performed to elucidate the exact actions whereby phthalates could influence these PPAR-mediated actions.

This study was performed in a sample of elderly white subjects. Thus, we cannot extrapolate these findings to other ethnicities and age-groups. The study also was conducted as a cross-sectional study, and, as such, a risk of reverse causality always exists. Thus, the present data have to be confirmed in prospective studies. Some pharmaceuticals contain phthalates. Whether this is an issue for patients with diabetes taking medication is not known, and details on certain brands of antidiabetes drugs are not collected in this study.

In the current study, we used serum measurements of phthalate metabolites. It is more common to use urinary measurements. The advantage of urinary measurements is that usually higher levels are found compared with serum and thereby more phthalate metabolites could be quantified above the lower detection limit. Therefore, we can only report associations regarding four metabolites, although in fact 10 metabolites were evaluated. Serum levels also might change more rapidly than urinary levels and therefore repeated measurements would be desirable for a more precise measure of exposure.

In conclusion, the findings in this cross-sectional study showed that several phthalate metabolites are related to diabetes prevalence, as well as to markers of insulin secretion and resistance, and support the view that these commonly used chemicals might influence glucose metabolism in humans at the level of exposure seen in the general elderly population.

This study was supported by the Swedish Research Council and the Swedish Research Council for Environment, Agricultural Sciences, and Spatial Planning. B.Z. is employed by the Medical Products Agency, Uppsala, Sweden.

No potential conflicts of interest relevant to this article were reported.

P.M.L. conceived of the project and contributed to the critical revision of the manuscript for important intellectual content. B.Z. was responsible for laboratory analyses of insulin and proinsulin measurements and calculations and contributed to the critical revision of the manuscript for important intellectual content. L.L. performed data analysis and contributed to the critical revision of the manuscript for important intellectual content. L.L. is the guarantor of this work and, as such, had full access to all the data in the study and takes responsibility for the integrity of the data and the accuracy of the data analysis.