Diethylene glycol (DEG) is a clear, colorless liquid used in the production of a wide variety of commercial and industrial products. ¹ It is found in trace, non-harmful amounts in other products such as some dietary supplements and cosmetics (probably as a manufacturing contaminant). ^2,3 When ingested in large amounts (1–1.5 g/kg), DEG can be a potent nephrotoxic and neurological poison. ^1,4 Unfortunately, DEG’s physical and chemical properties give it similar properties to solvents safely used in drug delivery such as propylene glycol and glycerin. This fact, and its lower cost compared with solvents such as pharmaceutical grade glycerin have resulted in more than 13 medication-associated, DEG mass poisonings. These incidents were associated with more than 500 deaths and thousands of sub-lethal exposures around the world since 1937, many of which occurred in children. ^1,4–7 Many if not all appear to be the result of substitution for diluents such as propylene glycol and glycerin. ^4,6–7 Despite the long history of medication-associated DEG poisoning, surprisingly little is known about the pathophysiology of human disease. As a result, the role of traditional antidotal therapy used in other, similar toxic alcohol poisonings (e.g., methanol and ethylene glycol) is unclear in DEG poisoning.

DEG is not hydrolyzed to two ethylene glycol molecules (animal data only) and further metabolized to known ethylene glycol metabolites, as once thought. ⁸ Limited animal and in-vitro evidence supports that DEG is metabolized to 2-hydroxyethoxy acetaldehyde by alcohol dehydrogenase (ADH) and then to 2-hydroxyethoxyacetic acid (HEAA) by aldehyde dehydrogenase. ^8,9 (Figure 1) This work has also identified diglycolic acid as a DEG metabolite in animal poisoning experiments, ^9,10 identified diglycolic acid as a substantial contributor to DEG induced acute kidney injury (AKI), ^11,12 and shown that inhibition of DEG metabolism by an ADH inhibitor such as fomepizole (4-methylpyrazole) decreases kidney injury and lethality. ^8,9 Unfortunately, there is no published information identifying the presence and/or pattern of HEAA and diglycolic acid concentrations in human DEG poisoning. Such information, if available, could be used to inform decision-making regarding whether or not to administer an agent such as fomepizole and/or perform hemodialysis. A single case report exists of a person who ingested a large amount of a pure form of one of these substances (diglycolic acid). He developed acute renal failure, peripheral neuropathy, coma, and ultimately died; unfortunately, no serum or urine toxicology testing was done. ¹³ The primary objective of this study was to identify and characterize DEG metabolite patterns in human serum, urine, and cerebrospinal fluid (CSF) specimens obtained from DEG poisoning victims. Our secondary objective was to compare these results to biologic specimens obtained from control-patients and describe the implications of these findings. This information may help clinicians characterize and quantify the severity and extent of human DEG poisoning by 1) identifying potential biomarkers for further validation; 2) informing laboratory assay development; 3) assisting clinicians in interpreting future biological testing results, and 4) provide data which can be used to inform future research and clinical decision-making.

Twenty serum specimens from both cases and controls were analyzed. Eleven and 22 urine specimens were analyzed from cases and controls, respectively. Urinary oxalic acid concentration could not be determined because of strong matrix effects encountered during laboratory analysis. One case had insufficient quantities of urine to perform testing for glycolic acid, HEAA, and diglycolic acid and was excluded from the analysis.

Diglycolic acid was detected in all case serum and urine samples. Diglycolic acid was not detected in any control serum samples but was detected in five (22.7%) control urine samples (Table 1). The median case serum diglycolic acid concentration was 40.7 mcg/mL and the median case urine diglycolic acid concentration was 28.7 mcg/mL. In controls, the median value for each was less than the lower limit of quantitation. The median and ranges for all analytes in serum and urine in cases and controls are presented in Table 2.

When considered dichotomously (detected or non-detected), the presence of diglycolic acid was the only analyte significantly associated with case status (serum and urine, OR=>999; exact p < 0.0001) (Table 1). Some analytes (serum oxalic acid, glycolic acid, and HEAA along with urinary glycolic acid) had sufficient numbers (>90%) of samples with quantifiable concentrations in both cases and controls to analyze by exact logistic regression when dichotomized at the median into a “high” and “low” group. This demonstrated that elevated serum oxalic acid and serum HEAA concentrations were also significantly associated with case status (OR = 14.6; 95% CI= [2.8,101], for both analytes). Urinary glycolic acid concentrations were significantly lower among cases when compared with controls (OR=0.057; 95% CI=[0.001,0.546]). The Wilcoxon Rank Sum test gave similar results to the above-mentioned logistic regression analyses for serum (oxalic acid, HEAA, and diglycolic acid being higher in cases vs. controls) and urine (glycolic acid and diglycolic acid being lower and higher in cases vs. controls, respectively) (exact p<0.0001).

For the CSF samples (n=8), detection frequencies were as follows: DEG (n= 3; 38%), ethylene glycol (n=0), glycolic acid (n=8; 100%), HEAA (n=5; 63%), and diglycolic acid (n= 7; 88%). Median concentrations and ranges for all analytes in mcg/mL were as follows: DEG (<LLQ; range, <LLQ–4.82), glycolic acid (2.84; range, 1.77–3.77), HEAA (1.01; range, <LLQ–121), and diglycolic acid (2.03; range, < LLQ-7.47). The sole CSF specimen used to develop the matrix blank had non-detectable concentrations for each analyte except for glycolic acid (2.91 μg/mL).

Diglycolic acid and HEAA were identified in human DEG poisoning among not just one biological matrix, but three (serum, urine, and CSF). The strong association of diglycolic acid (serum and urine) and HEAA (serum only) with case status strongly suggests DEG as the etiology. The most striking results are the marked differences in serum and urine diglycolic acid concentration between cases and controls. It is interesting to briefly compare these results to those reported in DEG poisoned rats by oral gavage. The median diglycolic acid serum concentration of 41 μg/mL is approximately 0.3 mmol/L, which is much higher than concentrations reported in high-dose (10 g/kg) exposed rats (mean, 0.04 mmol/L; range, <LLQ, 0.2 mmol/L).¹⁰ This might be due to a different exposure scenario from the single, acute ingestion model used in these rat studies. In the original Panama mass poisoning, patients had been instructed to consume a dose of the implicated cough syrup containing DEG as much as three times a day for as long as needed. The kinetic data in rats also showed that diglycolic acid concentrations peaked at a later time than HEAA in serum. This might explain the high serum and urine diglycolic acid concentrations seen in our human cases relative to other analytes, since biological samples from cases were collected well after onset of illness. ⁷ Other animal and in-vitro work in cultured human kidney cells implicate diglycolic acid as a nephrotoxic agent in DEG poisoning. ^11,12 The in-vitro work suggests that HEAA is not nephrotoxic. ¹¹ Diglycolic acid appears to be the most likely etiology of DEG’s nephrotoxicity. ^11,12

A previous study did not detect urinary diglycolic acid in a rat model of DEG poisoning. ⁸ The authors hypothesized that HEAA might have formed a cyclic compound, 1,4-dioxanone, not subject to further degradation. However, they used DEG doses of 1.1 g/kg which may have been too small to produce substantial diglycolic acid concentrations. ⁸ Similar rat models have found detectable urinary diglycolic acid concentrations following higher DEG doses, from 2 to 10 g/kg. ¹⁰ Furthermore, the analytical methods used by Weiner et al. (high-performance liquid chromatography) ⁸ may have been unable to detect very small amounts of diglycolic acid, even if present, as compared with the GC/MS method used in later work. ¹⁰ Rat studies with dioxane suggest that HEAA and 1,4-dioxanone may co-exist in a pH-dependent equilibrium that is heavily favored toward HEAA, ¹⁶ although 1,4-dioxanone may be favored in very acidic environments (<pH 5). ^16–18 However, such a pH is not encountered even in the urine of rats with severe acidosis from high doses of DEG ⁹ and this compound is not observed under normal aqueous conditions in vivo. ¹⁶

Case serum and urine DEG concentrations were not significantly higher than controls. This is not surprising given that biological samples were typically collected days after last exposure to the cough syrup, allowing sufficient time for DEG metabolism (half-life of 5–13 h). ^5,7,10 The few control samples that did have detectable serum DEG concentrations may have resulted from background environmental exposures as DEG can be found in packaging materials, cosmetics, and certain foodstuffs such as dietary supplements. ^2,3 During the initial outbreak investigation in 2006, urinary DEG concentrations were determined by CDC on a small sample of case and control-patient specimens: a significant difference was found. ⁷ The reason for the difference in findings between the 2006 study and this one may be due to different samples being tested although the possibility that DEG underwent some degradation while in storage is possible.

Case serum samples had significantly higher HEAA concentrations compared with controls as expected given that HEAA is a known metabolite of DEG. A possible explanation for the fact that many control samples had low but detectable serum HEAA concentrations is background environmental exposure. This may have occurred from personal care products containing either DEG as noted above or 1,4-dioxane which is metabolized to HEAA in vivo. ^19,20 Three samples (one case serum, one case urine, and one control urine sample) contained detectable ethylene glycol concentrations. This may be due to background environmental exposure among controls or possibly from further reaction of one of the DEG acid or aldehyde intermediates ¹⁰ (not by DEG metabolism to two ethylene glycol molecules). Finally, although control patients in the study had to have been hospitalized for any condition besides AKI, ⁷ they could have received the implicated DEG-containing cough syrup while in the hospital. The implicated medication was formulated in a pharmaceutical manufacturing plant operated by the hospital system and distributed to patients of this system. This might also explain why a few control-patients had detectable concentrations of DEG and HEAA.

Urinary glycolic acid concentrations were significantly lower among case samples when compared with controls. Limited human data suggest that normal urinary glycolic acid concentrations are approximately 38.8 mg/day (SD: 13.8 mg) and depend largely on diet. ²¹ If one assumes an average adult urine volume of 1.5 L, control patients were probably excreting similar amounts, approximately 22.5 mg/day (15 mcg/mL× 1500 mL/day) of glycolate. These values are proportionally much greater than the median urinary glycolic acid concentrations seen among cases (3 mcg/mL × 1500 mL/day or 4.5 mg/day). The lower values among cases may be due to an impaired ability to excrete glycolic acid due to AKI from DEG’s nephrotoxic effects. Impaired kidney function in cases may have reduced elimination of glycolic acid in urine, resulting in the availability of serum glycolic acid for further metabolism. As glycolic acid was present in the serum due to AKI, it may have been metabolized to oxalic acid (as well as other unmeasured organic acids) via normal metabolic pathways ²² and oxalate accumulated in the serum of the cases due to impaired kidney function. Additionally, late sample collection relative to exposure may have resulted in simply missing the peak serum glycolic acid concentration. Unfortunately, our samples did not capture serial measurements over time and alternative explanations for these findings may exist. As expected, almost all cases and controls did not have detectable serum and urine ethylene glycol concentrations.

The pathophysiology of DEG-induced neurotoxicity is much less clear and unstudied. The detection of HEAA and diglycolic acid in nearly all case CSF samples suggests that these metabolites may be associated with DEG-associated neurotoxicity. Signs and symptoms of non-inebriation related neurotoxicity tend to only appear following nephrotoxicity, ¹ hence these findings make intuitive sense (the nephrotoxic agents seem likely to also be the neurotoxic agents). Unfortunately, the data and the study design were insufficient to adequately characterize this issue and make a determination on causation. True control CSF samples with undetectable concentrations of DEG metabolites would have enabled us to say more; alas none were available for testing. Nevertheless, the single bio-bank supplied CSF sample had no detectable HEAA or diglycolic acid concentrations providing at least one comparison value. The finding of glycolic acid in the bio-bank CSF sample is not unexpected since it is an endogenous metabolic intermediate that is likely present at some small level in all body fluids.

Taken together, these findings suggest that the DEG metabolites diglycolic acid and HEAA are associated with human DEG poisoning: these results appear similar to those found in animal models of DEG poisoning. Although this study did not assess causation, these data provide further insight into the metabolism of DEG in humans after poisoning and more data to inform future work in studying potential DEG poisoning therapies. Although the role of HEAA and diglycolic acid in DEG-associated neurotoxicity remains unclear, the limited CSF findings from this study suggest a possible association. The possibility that other unmeasured analytes may occur following DEG poisoning exists as well and these may also contribute to poisoning.

The initial outbreak study was a case-control study that employed a matched study design. Due to the need for subject de-identification, we could not perform a matched, multivariable analysis. The true association between diglycolic acid and case status may be stronger than what we found, since ignoring the matching in the analysis of matched data typically biases results toward the null. ²³ Another limitation is that the time interval between last dose and biological sample collection is unknown for each case. This probably resulted in substantial variation between analyte concentrations among cases, which limits the utility in using the quantitative data to examine dose-response relationships. We also did not measure 1,4-dioxanone concentrations due to the unlikeliness of its presence at the urine and blood pH that are encountered in human DEG poisoning; such measurements would have further supported that this compound is not formed in appreciable amounts in vivo (consistent with rat studies). ¹⁶ Nor did we measure dioxane which can be metabolized to HEAA and might have affected HEAA measurements and their interpretation. It is likely that many of the cases were receiving hemodialysis during their clinical course. This may have affected analyte concentrations and quantitative measurements should be interpreted with this possibility in mind. Finally, the quantitative methods used for the laboratory analysis lacked the necessary specificity to exclude with complete certainty all other substances that have the same retention time and mass spectra on the low resolution GC/MS methods used. Further refinement of these methods to include detection of two confirmatory ions, in addition to the primary ion of interest, is needed to completely eliminate any other possibility. Nevertheless, we believe the likelihood of an alternative substance mimicking any of these analytes on GC/MS is extremely low based on several factors. The use of stable-isotope labeled internal standards confirmed analytes’ retention times for each sample analyzed. Limited analyses of the CSF samples by a complementary ion chromatography–mass spectrometry method, used on previous animal studies with DEG, afforded comparable results to the GC/MS data presented here (data not shown). Additionally, most of the measured metabolite concentrations are consistent with previously published animal DEG pharmacokinetic data.

Diglycolic acid (serum and urine) and HEAA (serum) concentrations are identifiable in, and associated with, human DEG poisoning. Diglycolic acid and perhaps HEAA appear to be useful biomarkers for human DEG poisoning but require further validation. These results may also be useful in interpreting biological testing results in future instances of DEG-associated illness. Further work characterizing the role of ADH inhibiting therapies such as fomepizole and ethanol, as well as hemodialysis, in treating DEG poisoning is needed. Finally, more work in human specimens is needed to validate these findings and to elucidate the role of HEAA and diglycolic acid in DEG-associated neurotoxicity.