Ambient Particulate Air Pollution is Associated with Higher Risk of Microvascular Complications and Diabetic Ketoacidosis Among Persons with Type 1 Diabetes Mellitus in an Observational Survival Study

Abstract

Aims:

The burden of Type 1 Diabetes Mellitus (DM, T1D) is growing and represents a major health care cost. This study investigated the relationship between ambient particulate matter (PM_2.5) exposure and T1D complications. The research was motivated by the potential for air pollution’s known inflammatory effects to exacerbate T1D’s microvascular harms (i.e., damaged peripheral tissues from poor glucose control).

Methods:

In a cohort of 12, 925 Utah participants, competing-risk Cox proportional hazards regression analyzed the increased risk of PM_2.5 exposure and diabetic ketoacidosis (DKA), along with kidney, ophthalmic, and neurological complications.

Results:

An interquartile range increase in one-year PM_2.5 exposure was associated with increased risk of DKA by 28.8% (95% CI: 14.3, 45.2), ophthalmic complications by 33.5% (95% CI: 10.0, 62.0), and neurological complications by 35.1% (95% CI: 14.3, 59.6). No statistically signficant effect was found for diabetic kidney complications. Acute exposures of 30, 60, and 90 days was also associated with increased risk of diabetic ketoacidosis, although less than one-year exposure.

Conclusions:

We hypothesize these effects stem from PM_2.5-induced oxidative stress and systemic inflammation, which may exacerbate metabolic disruptions already present in hyperglycemic individuals. Clients and providers may want to consider environmental factors, like air pollution, in T1D management.

Introduction

The global burden of Type 1 diabetes mellitus (T1D) has grown sharply over the past 50 years. T1D was a rare disease until the mid-20^th century, after which sharp increases in incidence were documented around the globe [1]. Worldwide, the incidence of T1D is currently increasing by 0.34% per year [2]. This global trend is echoed in the United States (US), where an estimated 1.7 million adults and 304,000 children and adolescents younger than age 20 had T1D in 2021 [3], or approximately 5/1000 adults and 4/1000 children [4].

An increasing prevalence of T1D has important public health consequences due to the macrovascular, microvascular, and glycemic event complications that arise from uncontrolled blood glucose levels. These are generally divided into macrovascular complications, microvascular complications, and glycemic events arising from uncontrolled blood sugar levels. Macrovascular complications include cerebrovascular and cardiovascular diseases, both of which confer significant mortality risks [5] even with tight glycemic control [6]. Unlike macrovascular complications, microvascular complications – such as diabetic nephropathy, neuropathy, and ophthalmopathy – are quite responsive to glycemic control [7–9]. Despite this, end-stage renal disease remains the leading cause of death for individuals with T1D of <35 years duration, retinopathy affects 27.4% of persons with T1D—climbing to over 80% after 40 years of disease [10]—and both peripheral and sensorimotor neuropathies are exceedingly prevalent, with studies suggesting cardiovascular autonomic neuropathy afflicting over 40% of individuals with T1D [11]. In fact, a 2020 simulation study estimated the lifetime economic burden of T1D in the United States at an astounding $813 billion [12], highlighting the importance of identifying and mitigating causes of complications and comorbidities for those with Type 1 diabetes.

While environmental factors, such as nutrition, viruses, and intrauterine events have been linked to development of T1D [13, 14], less is known about the impact of environmental toxicants—such as air pollution—on the development of complications among those with T1D. Several studies, including ours, have identified that higher levels of air pollution are associated with risk of developing diabetes [15] or perturbations in glucose metabolism in those with diabetes mellitus [16–18], but these studies did not robustly differentiate T1D and Type 2 DM (T2D). Separately distinguishing T1D is important because T1D-associated microvascular complications are hypothesized to concentrate in glomerular, retinal, and peripheral nerve tissues, which are unable to down-regulate glucose uptake during periods of hyperglycemia, resulting in increased intracellular oxidative stress, and ultimately cellular death [13, 19]. In one of the only prior studies examining the impacts of air pollution on T1D complications, we identified significant associations between air pollution and T1D-related mortality [20]. Similarly, air-pollutant-related perturbations in glucose metabolism [21] or systemic inflammation [22] may be important triggers for acute hyperglycemic events in T1D, such as diabetic ketoacidosis (DKA), but the evidence is limited.

Our goal, therefore, was to explore associations between ambient PM_2.5 exposure and risk of microvascular and glycemic diabetic complications among persons with T1D, including diabetic nephropathy, retinopathy, neuropathy, and DKA.

Methods

Study design and population

This analysis draws from 12,925 people with T1D in the US state of Utah from 1999 to 2021. T1D is defined as a subtype of diabetes mellitus resulting in insulin deficiency and sudden onset of severe hyperglycemia. Cases of Utah T1D diagnoses were obtained from electronic medical records of Intermountain Health Care and the University of Utah Health System [23], which tracked individual biometrics and diagnostic outcomes over time. Data were provided by the Utah Population Database (UPDB) located at the University of Utah [24]. When patient database records included diagnosis codes for both T1D and T2D, we included only participants with at least a 70% proportion of T1D diagnoses, similar to previous thresholds used for this cohort [25]. In addition to being a threshold used by prior studies, the 70% T1DM diagnosis proportion was selected as our primary outcome classification approach to ensure that the vast majority of diabetes ICD codes for an individual were for T1D, decreasing the likelihood of outcome misclassification due to intermittent coding errors in the medical records.

The study was approved by the Institutional Review Boards of Quinnipiac University (IRB 11025), the University of Utah (IRB_00096551), Intermountain Health Care (IRB number 1050657), and the Utah Resource for Genetic and Epidemiological Research (RGE_00002451), which is responsible for oversight of the UPDB.

Exposure

Ambient PM_2.5 measures were acquired from Washington University in St. Louis’s Atmospheric Composition Analysis Group. They created 0.1° × 0.1° grids of monthly average pollution estimates across North America from 1998 to 2022 using a residual Convolutional Neural Network (CNN) with inputs including Aerosol Optical Depth (AOD), GEOS-Chem chemical transport model, and calibrated via ground-based observations [26]. These grids were converted to geographic hexagons per the H3 spatial index system at resolution 7 (5.15 km² average area) [27] and linked to participant residential history (also in H3 format) received from the UPDB via driver’s license records. Each participant’s annual moving-average PM_2.5 exposure was assigned based on monthly averages for each hex they lived in during the year, weighted by how many months they lived there. The same procedure was used for exposure assignment of 30, 60, and 90 days.

Outcomes

Outcomes of interest were based on individual diagnoses from the International Classification of Diseases, Tenth Revision, Clinical Modification (ICD-10) [28]. ICD-10 codes included E10.1xx (Type 1 diabetes mellitus with ketoacidosis), E10.2xx (Type 1 diabetes mellitus with kidney complications), E10.3xx (Type 1 diabetes mellitus with ophthalmic complications), and E10.4xx (Type 1 diabetes mellitus with neurological complications). Ketoacidosis is defined as a complication of DM resulting in dangerously high levels of blood ketones (potentially resulting in coma or death). Kidney, ophthalmic, and neurological complications are defined, respectively, as DM-aggravated damage to renal function, eye tissues (including diabetic retinopathy), and hyperglycemic harm to the peripheral nervous system [28]. The sub-diagnoses of DKA include both presence and absence of coma. Sub-diagnoses of kidney complications include diabetic nephropathy, inter- and intracapillary glomerulosclerosis, and Kimmelstiel-Wilson disease. Sub-diagnoses of ophthalmic complications include retinopathy with and without macular edema plus nonproliferative diabetic retinopathy. Sub-diagnoses of diabetic neuropathy include mononeuropathy, polyneuropathy, autonomic (poly)neuropathy, and amyotrophy.

All enrolled participants were diagnosed with T1D sometime during the study period. For each diabetes complication, the first date of a complication diagnosis served as the survival outcome event, and the earliest date of contact with the participant within the exposure period (1999–2021) served as study entry date (indicating the time they contracted T1D). If an individual died before an outcome diagnosis, then they were censored under a competing-risk framework. In order to avoid possible bias from exposure window misclassification [29], we assessed whether any participants had been diagnosed with T1D prior to 1999. One such person was found, but sensitivity analysis excluding them showed no changes to parameter estimates.

Covariates

Covariates to adjust for potential confounding included sex (binary), race (categorical with levels Asian, Black, Native American, Pacific Islander, White, and Other or Unknown), age (time-varying continuous for each year of follow-up), BMI (time-varying continuous for each year of follow-up), and socioeconomic status (SES, continuous for each year of follow-up). Because BMI measurement did not necessarily directly correspond with the date of annual exposure measurement, the nearest patient contact date with a recorded BMI was used. SES comprised census-tract-level social vulnerability index (SVI) converted to H3 hexagons and linked with participant residential history. We used the most recent SVI recorded in 2020 and included it in models as a non-time-varying covariate. SVI is a composite number combining poverty level, employment status, housing cost, high school graduation, and lack of health insurance. Higher values indicate higher vulnerability and hence lower SES [30]. For census tracts that did not have a recorded SVI, we used a participant’s mean SVI level calculated from other lived locations.

Statistical Analysis

Models

Parameters were estimated using competing-risk Cox proportional hazards regression with death as the competing risk. We constructed four primary models for the four diagnosis outcomes, each with three sub-models: base (partial covariate adjustment including sex, age, SES, and year), fully adjusted, and base + BMI (adjusted for non-BMI covariates). BMI was modeled separately, given past evidence of associations between air pollution and both obesity and T1D complications that suggests obesity may be a mediator rather than a confounder [31]. The independent variable of interest was an interquartile range (IQR) change in ambient annual average PM_2.5 exposure. As mentioned earlier, sex and race remained constant through all years of follow-up, but age, BMI, and SES were time-varying. Given that DKA may develop from more acute exposures, we further explored 30, 60, and 90 day moving average exposure windows for this outcome. Model diagnostics were assessed with Schoenfeld residuals [32].

In all four fully-adjusted models, we explored effect modification by age, sex, and race. Each variable was cut into binary predictors: for age, above and below the median; for sex, male versus female; and for race, white versus nonwhite. These were interacted with PM exposure and assessed for multiple-comparison inflation using false discovery rate adjustment with Benjamini and Yekutieli’s method [33].

Sensitivity Analysis: To ensure our results were not being significantly impacted by the >70% definition of T1D, we reran our main models using various cutoffs for participant exclusion based on the ratio of T1D to T2D diagnoses as shown in Supplemental Figure 1.

Software

All statistical analysis was performed with R 4.2.2 and SAS 9.4 [34, 35].

Results

This survival analysis included 12,925 T1D participants in Utah whose characteristics are shown in Table 1.

Table 1:

Participant Characteristics

Total N = 12,925
		PM2.5 exposure in μg/m3 (mean ± SD) or years (mean ± SD, range) by outcome*
Characteristic	n(%), mean ± SD	DKA	Kidney	Ophthalmic	Neurological
Race
Asian	84 (0.7%)	8.57 ± 2.27	8.08 ± 1.25	8.42 ± 1.40	7.72 ± 0.99
Black	139 (1.1%)	8.77 ± 2.60	8.94 ± 1.44	9.19 ± 1.32	8.94 ± 1.44
Native American	45 (0.3%)	8.22 ± 1.75	8.06 ± 1.27	8.59 ± 1.17	8.15 ± 1.16
Pacific Islander	77 (0.6%)	8.35 ± 1.73	8.00 ± 2.04	6.84 ± 1.75	7.87 ± 1.93
White	11,800 (91.3%)	8.12 ± 2.58	8.25 ± 1.75	8.36 ± 1.78	8.30 ± 1.73
Other or Unknown	780 (6.0%)	8.55 ± 2.84	8.26 ± 1.43	7.99 ± 1.70	8.16 ± 1.56
Sex
Female	6,256 (48.4%)	8.17 ± 2.62	8.34 ± 1.70	8.37 ± 1.78	8.34 ± 1.70
Male	6,669 (51.6%)	8.11 ± 2.54	8.18 ± 1.77	8.35 ± 1.77	8.27 ± 1.76
Age at outcome diagnosis
< 18	–	8.48 ± 3.17	8.48 ± 1.64	8.60 ± 1.94	8.61 ± 1.66
18 – 29	–	8.01 ± 2.09	8.51 ± 1.72	8.63 ± 1.84	8.52 ± 1.69
30 – 39	–	7.93 ± 2.06	8.33 ± 1.70	8.35 ± 1.67	8.40 ± 1.65
40 – 49	–	7.88 ± 2.02	8.20 ± 1.79	8.37 ± 1.77	8.33 ± 1.75
50 +	–	7.43 ± 1.93	7.90 ± 1.74	8.06 ± 1.73	7.88 ± 1.77
Years elapsed until complication	–	6.93 ± 5.63	10.71 ± 5.91	11.51 ± 5.66	10.73 ± 5.74
Age at study entry	24.3 ± 18.80
BMI at study entry	24.2 ± 6.69
SES (increasing social vulnerability)	2.40 ± 1.06

^*includes only participants who received the outcome diagnosis

The vast majority of participants were White (91.3%). Male partipants exceeded female participants by a statistically signficant margin of 3.2% (p < 0.001). At study entry, mean age was 24.3 years (median 19.0) and mean BMI was 24.2 (median 24.0), which is considered within the normal weight range for adults [36], but values ranged from 10 to 60. Mean SVI in this cohort is slightly lower than the 2020 national population-weighted average SVI of 2.45 (on a scale of 1 = low socioeconomic vulnerability to 5 = high vulnerability).

Figure 1 shows the annual distributions of participants’ one-year moving-average exposures of PM_2.5 during the study period. The overall mean PM_2.5 level for all participants across all years was 7.94 μg/m³, median 7.80, IQR 2.48. The trend of decreasing PM_2.5 concentration over this time period reflects an overall national trend of declining ambient particulate pollution due to federal regulatory controls. For example, from 2001–2017, PM_2.5 levels in the contiguous United States dropped 38.5% [37]. Although the overall mean lies below the US Environmental Protection Agency’s 2024 National Air Quality Health Standard for annual average particulate matter (set at 9.0 μg/m³ and requiring power, transportation, and industrial sectors to implement emissions reductions) [38], individual earlier years exceed that level and the prior standard of 12.0 μg/m³.

Figure 1:

Distributions of annual PM_2.5 exposure levels

Figure 2 shows the associations between annual average PM_2.5 exposure and risk of all complications of interest among people with T1D.

Figure 2:

Results of competing-risk Cox Proportional Hazards regression for all complications (ICD10: E10.1, E10.2, E10.3, E10.4) with 1-year exposures of PM_2.5 (per IQR)

* Base model includes covariates of sex, age, SES, and year. Fully adjusted model adds BMI and race.

Regression of 1-year exposure with DKA outcomes resulted in 1,275 diagnoses (9.9%), 855 deaths (6.6%), and 10,748 (83.5%) censored final events. For the base model, an IQR increase in PM_2.5 exposure was associated with 30.2% (95% CI: 16.0, 46.2) increase in risk. The fully adjusted model shows an IQR increase in PM_2.5 exposure was associated with 28.8% (95% CI: 14.3, 45.2) increase in risk.

Regression of kidney complications resulted in 736 diagnoses (5.7%), 833 deaths (6.4%), and 11,356 (87.9%) censored final events. In the base model, an IQR increase in PM_2.5 exposure was associated with 14.5% (95% CI: −7.4, 41.5) increase in risk (non-significant). The fully adjusted model shows an IQR increase in PM_2.5 exposure was associated with 11.9% (95% CI: −9.6, 38.5) increase in risk (non-significant).

Regression of ophthalmic complications resulted in 900 diagnoses (7.0%), 841 deaths (6.5%), and 11,184 (86.5%) censored final events. For the base model, an IQR increase in PM_2.5 exposure was associated with 36.6% (95% CI: 12.6, 65.6) increase in risk. The fully adjusted model shows a 1-IQR increase in PM_2.5 exposure was associated with 33.5% (95% CI: 10.0, 62.0) increase in risk.

Regression of neurological complications resulted in 1,190 diagnoses (9.2%), 829 deaths (6.4%), and 10,906 (84.4%) censored final events. In the base model, an IQR increase in PM_2.5 exposure was associated with 37.0% (95% CI: 16.1, 61.5) increase in risk. The fully adjusted model shows a 1-IQR increase in PM_2.5 exposure was associated with 35.1% (95% CI: 14.3, 59.6) increase in risk.

In all of the above results, excluding BMI from the models did not materially alter parameter estimates (ranging from relative changes of 0.6% to 1.9% in fully adjusted HR estimates).

Given the potentially rapid onset and mortality implications of ketoacidosis, Figure 3 shows further investigation of DKA outcomes with acute exposures of 30, 60, and 90 days.

Figure 3:

Results of competing-risk Cox Proportional Hazards regression for Type 1 diabetes mellitus with ketoacidosis (ICD10: E10.1) with acute exposures of PM_2.5 (per IQR)

* Base model includes covariates of sex, age, SES, and year. Fully adjusted model adds BMI and race.

In fully adjusted models, 30-day exposure was associated with 18.9% (95% CI: 10.1, 28.3) risk increase; 60-day exposure was associated with 18.8% (95% CI: 10.0, 28.2) risk increase; and 90-day exposure was associated with 18.6% (95% CI: 9.8, 28.1) risk increase.

Figure 4 investigates binary effect measure modification for age, sex, and race across each of the outcomes.

Figure 4:

Binary effect modification results for 1-year and acute exposures to PM_2.5 (per IQR)

*No statistically significant (p < 0.05) interactions after false discovery rate adjustment.

All models fully adjusted. Median age was 26.7 years across all follow-up.

False discovery rate analysis for p-inflation due to multiple comparisons yielded null results for all effect modification results. Linear fits of Schoenfeld residuals suggested no violations of proportional hazards assumptions in main or effect modification models.

Sensitivity analyses: In models using other outcome definitions (i.e., >50%, >60%, >80% ratio of T1D diagnoses) our results did not importantly differ (Supplemental Figure 1).

Discussion

We found clinically relevant and statistically significant air pollution exposure-related increases in risk for several key T1D complications. The largest magnitude association was for 1-year PM_2.5 exposure and neurological complications followed by ophthalmic complications. For ketoacidosis, 30-, 60-, and 90-day exposures were associated with similar HRs in each model. We hypothesized that the acute nature of DKA would result in greater sensitivity to shorter pollution exposure timeframes; our findings did not support that. One-year exposure was associated with the highest observed risk. In effect modification models, we found potential evidence that air pollution exposures among non-white and younger participants were associated with higher risk of DKA, while air pollution exposures among those below the median age were associated with higher risk of neurological complications. However, after controlling for multiple comparisons, no statistically signfiicant effect modificiation was observed.

While our findings are generally consistent with the extant prior literature on air pollution and diabetes mellitus, most prior literature focuses on T2D or undifferentiated DM, which complicates this comparison [39]. For example, in China a research team found each 10 μg/m³ increase in PM_2.5 was associated with an odds ratio (OR) of 1.41 (95% CI: 1.27, 1.57) for developing diabetic retinopathy [40]. Similarly, a study in Taiwan used a health insurance database to examine associations between air pollution and diabetic retinopathy in undifferentiated diabetes mellitus patients, finding associations for 10 μg/m³ increases in PM_2.5 and PM₁₀ (HRs of 1.29 [95% CI: 1.11, 1.50] and 1.18 [95% CI: 1.08, 1.28], respectively) [41]. Further, in a longitudinal study of 25 people with T1D and T2D, acute exposure to PM₁₀ was associated with biomarkers of systemic oxidative stress, which may help explain the biological pathways through which air pollutants impact the risk of microvascular and glycemic complications in T1D [42].

However, because these studies did not differentiate between T1D and T2D, it limits their utility in understanding the impacts of air pollutants on the development of complications among individuals with T1D. T1D and T2D have important etiologic and pathophysiologic differences as disorders of glucose metabolism, which may impact the way in which air pollutants affect disease progression in the respective conditions. For example, in a recent study, exposure to acute PM_2.5 was associated with reduced vascular reactivity in brachial arteries for both T1D and T2D participants. Flow-mediated dilation showed −8.8% (95% CI: −17.0, −0.1; n = 183) reduction for T2D and −4.8% (95% CI: −19.9, −13.1; n = 45) reduction for T1D, while nitroglycerin-mediated dilation showed a similar −8.5% (95% CI: −14.1, −2.5; n = 169) reduction for T2D but a null finding for T1D. The authors hypothesized that people with T2D may be more vulnerabile than people with T1D to particulate mechanisms, such as chronic inflammation and oxidative stress, imbalanced vasoreactors, or elevated superoxide flux on the endothelium further exacerbated by PM, highlighting the need to study the environmental determinants of T1D and T2D complications separately [43].

Prior literature on the impacts of air pollutants on microvascular and glycemic complications in T1D populations is sparse. In a 2021 review, Shubham et al. hypothesized several mechanisms that might explain air pollution contribution to chronic kidney disease (which might be further exacerbated by T1D diagnosis). These include direct neurotoxic effects of particulate components, inflammation and oxidative stress that are known outcomes of particulate inhalation, hypercoagulability leading to epithelial injury and hence kidney dysfunction, vascular injury, hypoxia, and autoimmune response [44].

However, in an analysis of a Taiwanese population, no association between long-term PM_2.5 and renal dysfunction was found. For PM₁₀ and course PM, participants without diabetes showed significantly increased odds of chronic kidney disease, but persons with diabetes did not [45]. Our findings of positive, consistent, and statistically significant associations may indicate that glomerular cells are particularly vulnerable to PM-induced oxidative stress and inflammation in persons with T1D.

To our knowledge, no prior literature explores the impacts of air pollution on retinopathy in T1D, specifically. For example, a recent study in a Japanese population showed increased years of exposure to PM_2.5 were associated with increased odds of diabetic retinopathy in an undiffereniated diabetic population, with a 5-year exposure associated with an OR of 1.93 (95% CI: 1.21, 3.07) per IQR increase [46]. The link between air pollution and diabetic retinopathy is not fully understood, but particulate pollutants may contribute by triggering oxidative stress, systemic inflammation, and increased cytokine levels (e.g., TNF-α, VEGF, IL-6). Metals like nickel, copper, and arsenic in fine particles have been associated with elevated inflammatory markers, potentially explaining part of PM’s role in diabetic retinopathy. These effects may compound the inflammation and oxidative stress already caused by diabetes [46]. Additionally, air pollution has been linked to endothelial dysfunction and atherosclerosis, which are key features of diabetic retinopathy, and known complications of T1D, which may explain our findings of consistent and significant effects of PM on retinopathy in this T1D cohort in Utah.

The strongest prior evidence supports our findings of acute glycemic events due to air pollution. In a nationwide Chinese case-crossover study, short-term air pollution was associated with increased rates of diabetic mortality and DKA [47]. In Chile, pooled geographic-sector estimates of relative risk of hospitalization for diabetic ketoacidosis and coma were associated with changes in single-pollutant concentrations with 6-day lag: RR of 1.108 (95% CI: 1.063, 1.155) for IQR increase of PM_2.5 [48]. In a German study of under-21-year-olds diagnosed with T1D (n = 44,383), long-term exposure to both PM₁₀ and PM_2.5 was associated with increased event rates of severe hypoglycaemia and hypoglycaemic coma [49]. In contrast, Kim et al. found associations between NO₂ and emergency department visits for diabetic coma (RR of 1.092 [95% CI: 1.005, 1.186]) at lag 0–3, but not for PM₁₀ and other pollutants [50].

The biological mechanisms underpinning how PM may impact glycemic events such as DKA are beginning to be understood, with the impacts of oxidative stress and inflammation implicated in exacerbating perturbations in glucose metabolism. For example, Puett et al. described relationships between PM_2.5 exposure and inflammatory markers in youth with T1D. Specifically, a 3-day lagged exposure to PM_2.5 was associated with 5.0% (95% CI: −0.01%,10.3%) higher IL-6 level when comparing lowest and highest quartiles of exposure [51]. Further, in a longitudinal study of 25 people with T1D and T2D, acute exposure to PM₁₀ was associated with biomarkers of systemic oxidative stress [42]. In a T1D rat model, compared to filtered air, exposure to PM_2.5 was associated with significant increases in mean levels of Glycated hemoglobin A1c, interleukin 6, and fibrinogen. Authors maintain that PM exposure caused “focal myocarditis, aortic medial thickness, advanced glomerulosclerosis, and accentuation of tubular damage of the kidney” [52]. In the context of this prior literature, we hypothesize that the observed increased risk of DKA in our population may result from the known impacts of PM on systemic inflammation, which, in the presence of T1D, helps contribute to the cascade of hyperglycemia, ketonemia, and acidosis which characterizes this life-threatening complication.

Limitations and Strengths

Our findings should be interpreted in the light of several limitations. First, the demographic cohort in Utah is quite homogenous—predominantly white race (91.3%) and nonsmoking—making generalization to other demographic groups challenging. Second, we were not able to consider drug prescriptions or patient adherence to disease-management medicines, which may be potential confounders. Third, we used only census tract-level measures to account for SES. Fourth, it is possible that there was some residual outcome misclassification due to coding errors in the medical record which might occur between T1D and T2D. To minimize the likelihood of this we used a high (70%) threshold in our primary analyses. Sensitivity analyses (Supplemental Figure 1) showed only minimal changes in effect sizes when using thresholds between 50% and 80%, which is reassuring that our results are not unduly influenced by the choice of threshold. These limitations are, however, counterbalanced by several strengths. This is a large cohort of existing T1D diagnoses tracked over a considerable timeframe. We employed a high-quality, deep-learning-derived spatiotemporal model for PM_2.5 exposure assessment. And individual exposure assignment takes into consideration granular residential history for each participant.

Conclusions

We investigated for the first time the relationship between ambient particulate matter (PM_2.5) exposure and T1D complications in a large US cohort of 12,925 individuals from 1999–2021. We found that PM exposure is associated with consistent and significant increases in the risk of ketoacidosis, neurological complications, and ophthalmic complications, but not kidney complications. The study supports the limited previous research suggesting air pollution can contribute to diabetic complications through mechanisms involving vascular dysfunction, inflammation, and oxidative stress. Thus, it is important for clients and providers to consider environmental factors in T1D management and recognize that ambient air pollution may be an important yet underappreciated contributor to T1D complications.

Data availability

Due to participant confidentiality issues, data is not generally available.