Abstract

Materials and Methods

Results

The 25 759 participants ranged in age from 1 to 93 years. Whereas the mean age was 23.0 ± 16.9 years (Table 1), there were 7324 participants (28%) who were 26 years of age or older. Subjects were 50% female, 82% non-Hispanic white, and had a mean duration of diabetes of 10.6 years at the time of data collection. Median HbA1c ± SD was 8.4% ± 1. 6%. Additional cohort characteristics are shown in Table 1. Among the participants, 27% of study participants (n = 6870) had a diagnosis of at least one AID. Of these, 20% (n = 5262) had one; 5% (n = 1374) had two; less than 1% (n = 214) had three; less than 1% (n = 19) had four, and less than 1% (n = 1) had five AIDs.

Table 1

Study Participant Characteristics

Characteristic	n, % (n = 25 759)
Gender, female	12 895 (50%)
Age, y, mean ± SD	23.0 ± 16.9
<6	1275 (5%)
6 to <13	6962 (27%)
13 to <18	6321 (25%)
18 to <26	3877 (15%)
26 to <50	4479 (17%)
50 to <65	2145 (8%)
≥65	700 (3%)
Diabetes duration, y, mean ± SD	10.6 ± 11.8
Age of diagnosis of T1D, y, mean ± SD	11.6 ± 10.6
<18	21 528 (84%)
≥18	4231 (16%)
Race/ethnicity
White non-Hispanic	21 141 (82%)
Black non-Hispanic	1314 (5%)
Hispanic or Latino	2049 (8%)
Asian/Pacific Islander	344 (1%)
Native American	91 (<1%)
More than one race	721 (3%)
Participants with coexisting AIDs	6819 (26%)

Open in a separate window

Factors associated with AIDs

Participants with one or more AIDs were more likely to be older (P < .001), female gender (P < .001 after adjusting for age and duration), and of white non-Hispanic race/ethnicity (P < .001 after adjusting for age and duration, Table 2). Although the frequency of one or more AIDs was greatest in non-Hispanic whites (29% of this racial group affected), the likelihood of one or more AIDs was also high in the other ethnic groups: 21% of Hispanic/Latino (n = 420), 12% of black non-Hispanics (n = 156), and 21% of other races (n = 243), illustrating increased autoimmune propensity in T1D-affected individuals of any race, compared with what would be expected in the general population (Table 2).

Table 2

Factors Associated With Additional AID Diagnosis

	Additional AID (Total = 6962), n (row %)	Odds Ratio (95% Confidence Interval)a	P Valuea
Age, y			<.001
<6	243 (19%)	Referent
6–<13	1540 (22%)	1.20 (1.04, 1.40)
13–<18	1504 (24%)	1.32 (1.14, 1.54)
18–<26	917 (24%)	1.31 (1.12, 1.54)
26–<50	1544 (34%)	2.23 (1.91, 2.60)
50–<65	884 (41%)	2.97 (2.52, 3.50)
≥65	330 (47%)	3.80 (3.10, 4.64)
Diabetes duration, y			<.001
<1	588 (18%)	Referent
1 to <10	3009 (23%)	1.34 (1.22, 1.48)
10 to <20	1502 (29%)	1.84 (1.65, 2.06)
20 to <50	1737 (40%)	3.01 (2.70, 3.35)
≥50	126 (48%)	4.17 (3.22, 5.41)
Genderb			<.001
Male	2496 (19%)	Referent
Female	4462 (35%)	2.18 (2.06, 2.31)
Race/ethnicityc			<.001
Black non-Hispanic	158 (12%)	Referent
White non-Hispanic	6120 (29%)	2.71 (2.30, 3.21)
Hispanic or Latino	420 (21%)	2.04 (1.70, 2.57)
Other	244 (21%)	2.06 (1.65, 2.57)
Age of Diagnosis of T1D, y			<.001
<18	5451 (25%)	Referent
≥18	1511 (36%)	1.34 (1.24, 1.44)
HbA1c, %d			.02
<7.0	1097 (29%)	Referent
≥7.0	5819 (27%)	1.09 (1.00, 1.18)

Open in a separate window

^aOdds ratios, confidence intervals, and P values were based on univariate logistic regression models. P values for age, duration, T1D diagnosis age, and HbA1c used a continuous variable. P values for gender, race, and HbA1c were adjusted for age and T1D duration. P value for T1D diagnosis was age adjusted for T1D duration.

^bExcludes nine transgender individuals (four with at least one additional AID and five without).

^cRace/ethnicity missing for 99 (20 with at least one additional AID and 79 without).

^dHbA1c taken as the average over the previous year. Missing for 189 (46 with at least one additional AID and 143 without).

Participants with one or more AIDs were also more likely to have a longer T1D duration (P < .001) and be diagnosed with T1D later in life (P < .001 after adjusting for duration). However, the burden of additional AIDs was not associated with metabolic control after adjusting for confounding from age and T1D duration: frequency of one or more AIDs was 29% among participants meeting American Diabetes Association glycemic target of HbA1c less than 7.0% and 27% among those not meeting the target (P = .02; Table 2).

The distribution of AIDs by age is shown in Figure 1. Most affected participants across all age groups have a single AID (black bars), as opposed to more than one (gray bars). Notably, there is an age-dependent increase in the percentage affected with one AID starting in participants 18 years old and older with no plateau effect. An increase in the participants with two or more AIDs was also observed with older age.

Open in a separate window

Figure 1

Age-dependent increase of additional AIDs. More study participants are affected by a single AID (black bars) than multiple AIDs (gray bars). Notably, there is an age-dependent increase in individuals affected with AIDs with no plateau effect, with 47% of those aged 65 years or older having one or more AIDs (n = 330 of 700).

Prevalence of AIDs by disease type

The types of AIDs and their frequencies are presented in Table 3. Thyroid diseases were the most common reported AID, affecting 20% (n = 5121) of all study participants. Among these, most had hypothyroidism or Hashimoto’s disease (n = 4895, 96% of all thyroid diseases), whereas fewer had hyperthyroidism or Graves’ disease (n = 400, 8% of all thyroid diseases). Less common were rheumatoid arthritis (RA; n = 244, <1%), vitiligo (n = 169, <1%), psoriasis (n = 170, <1%), and Addison’s disease (n = 76, <1%).

Table 3

AID Diagnosis According to Gender and Race/Ethnicity, n, %

	Overall (n = 25 759)	Gendera		Race/Ethnicityb
		Male (n = 12 855)	Female (n = 12 895)	Black Non-Hispanic (n = 1315)	White Non-Hispanic (n = 21 141)	Hispanic or Latino (n = 2048)	Other (n = 1156)
Additional AIDs, n
0	18 940 (74%)	10 361 (81%)	8574 (66%)	1165 (89%)	15 134 (72%)	1643 (80%)	919 (79%)
1	5225 (20%)	2017 (16%)	3204 (25%)	123 (9%)	4585 (22%)	321 (16%)	180 (16%)
2	1358 (5%)	413 (3%)	945 (7%)	25 (2%)	1205 (6%)	74 (4%)	52 (4%)
≥3	236 (<1%)	64 (<1%)	172 (1%)	2 (<1%)	217 (1%)	10 (<1%)	5 (<1%)
Thyroid disease	5121 (20%)	1709 (13%)	3408 (26%)	115 (9%)	4526 (21%)	288 (14%)	178 (15%)
Hashimoto’s and/or hypothyroidism	4895 (19%)	1649 (13%)	3242 (25%)	104 (8%)	4334 (21%)	272 (13%)	171 (15%)
Graves’ and/or hyperthyroidism	400 (2%)	103 (<1%)	297 (2%)	19 (1%)	343 (2%)	24 (1%)	13 (1%)
Addison’s disease	76 (<1%)	36 (<1%)	40 (<1%)	0	72 (<1%)	3 (<1%)	1 (<1%)
GI disease	1626 (6%)	701 (5%)	925 (7%)	11 (<1%)	1467 (7%)	92 (4%)	48 (4%)
Celiac	1517 (6%)	650 (5%)	867 (7%)	10 (<1%)	1370 (6%)	87 (4%)	45 (4%)
Colitis	79 (<1%)	32 (<1%)	47 (<1%)	0	72 (<1%)	4 (<1%)	2 (<1%)
Crohn’s	39 (<1%)	21 (<1%)	18 (<1%)	2 (<1%)	33 (<1%)	1 (<1%)	1 (<1%)
Collagen vascular disease	456 (2%)	149 (1%)	307 (2%)	10 (1%)	396 (2%)	27 (1%)	14 (1%)
RA	244 (<1%)	62 (<1%)	182 (1%)	14 (1%)	208 (<1%)	16 (<1%)	6 (<1%)
Psoriasis	170 (<1%)	87 (<1%)	83 (<1%)	1 (<1%)	152 (<1%)	8 (<1%)	8 (<1%)
Lupus	43 (<1%)	3 (<1%)	40 (<1%)	2 (<1%)	37 (<1%)	3 (<1%)	1 (<1%)
Scleroderma	11 (<1%)	1 (<1%)	10 (<1%)	1 (<1%)	10 (<1%)	0	0
Skin disease	272 (1%)	117 (<1%)	155 (1%)	11 (<1%)	221 (1%)	26 (1%)	13 (1%)
Vitiligo	169 (<1%)	81 (<1%)	88 (<1%)	6 (<1%)	139 (<1%)	13 (<1%)	10 (<1%)
Alopecia	93 (<1%)	34 (<1%)	59 (<1%)	4 (<1%)	74 (<1%)	11 (<1%)	4 (<1%)
Dermatomyositis	15 (<1%)	2 (<1%)	13 (<1%)	1 (<1%)	12 (<1%)	2 (<1%)	0

Open in a separate window

Abbreviation: GI, gastrointestinal. Table excludes diseases that were rare in our sample: multiple sclerosis (one participant) and Sjogren’s syndrome (one participant).

^aExcludes nine transgender participants.

^bRace/ethnicity missing for 99.

We observed a strong female predominance across the spectrum of AIDs. Of those with thyroid diseases, 67% (n = 3408) were female, with a similar distribution among those with Hashimoto’s thyroiditis (66% female) and Graves’ disease (74% female). Among collagen vascular diseases, systemic lupus erythematosus has by far the highest female predominance at 93% (n = 40), followed by scleroderma at 91% (n = 10) and RA at 75% (n = 182). Other disease categories showed minor female preponderance, ranging from 57% (n = 867) in celiac disease to 53% (n = 40) in Addison’s disease. As shown in Table 3, females also were more likely to have two or more AIDs compared with males.

Discussion

This is the largest US-based study assessing the frequency of AIDs in individuals with T1D. Although the T1D Exchange is not a population-based registry, our cohort represents the diverse sociodemographic and clinical features in children and adults with T1D throughout the United States. The overall prevalence of additional AIDs in our participants with T1D was 27%, which significantly exceeds the estimated 3%–8% prevalence of AIDs in the general population (17, 18) as well as some of the previously reported AID frequency in T1D populations (6, 19–22). Whereas our registry participants are divided evenly among female and male gender, most of those affected by additional AIDs are female (63%, n = 4321) compared with male (37%, n = 2494), which agrees with the established female predominance in the autoimmune literature.

The prevalence of specific AIDs in T1D differs based on disease type. Autoimmune thyroid disease is well recognized as the most common AID in T1D. We found autoimmune thyroid disease in 24% of our registry participants, a figure that is consistent with prior reports (23). Of note, some studies that relied on autoantibody testing yielded higher prevalence data, including Hanukoglu et al (24), who had shown 27% thyroidperoxidase and/or thyroglobulin antibody positivity in teens with T1D (n = 29 of 109), Barker et al (25), who reported 29% thyroid antibody positivity (n = 236 of 814), and Umpierrez et al (21), who reported 33% as having either autoantibodies or thyroid dysfunction (n = 19 of 58). Because ours was a retrospective observational study, autoantibody data were not available for all patients, which likely led to an underestimation of the true prevalence of thyroid autoimmunity in this population.

In other disease categories, our findings contrasted variably from existing data. Addison’s disease was seen in 0.3% (n = 75 of 25 759) of our T1D participants, lower than previously published 1.7% (n = 11 of 629) (14). RA was seen in less than 1% (n = 228) in our study, whereas its incidence has been reported as higher in other series, particularly in older patients (7, 27). Both these diseases tend to affect individuals after the age of 35–40 years; therefore, it is no surprise that they had lower rates in our cohort (mean age 23.0 y, median age 16.5 y). In contrast, celiac disease, which is often seen in the young, had a prevalence of 6% in our series, which fits within the 4%–11% range that has been previously reported (24–26, 28). The young age of our cohort also made it possible to miss a portion of adult-onset autoimmune diabetes patients, who tend to have a high frequency of associated autoimmunity, particularly of the thyroid (29, 30).

Our cohort demonstrated an age-related increase in AID frequency, with no threshold effect in the very young or a plateau effect in older age groups. Beginning with younger than 6 years of age, 18% of these children already exhibited one or more AIDs in addition to diabetes. The AID frequency remains largely stable through adolescence and early adulthood but increases after the age of 26 years. The older patients in our registry had the highest AID frequency, up to 42.2% of those older than 50 years (n = 1202) and 46.9% of those older than 65 years (n = 328). Without age-of-onset information for each AID diagnosis, we cannot distinguish new-onset AID in each age group from mere accumulation over time. However, there is strong evidence from the literature that certain AIDs cluster in older populations, including Sjogren’s, multiple sclerosis, and RA (27, 31–33), which may be the case in our series. Interestingly, despite what may be higher autoimmunity in the elderly, the clinical manifestations of late-onset disease are often milder, reflecting altered immune regulation and possibly expansion of protective regulatory mechanisms with advancing age (34–36). It is also noteworthy that the increased burden on AIDs did not appear to adversely affect overall metabolic control in our participants.

In addition to cohort size, another advantage of this study is the inclusion of a large number of black and Hispanic participants. For a disease such as T1D that most commonly affects non-Hispanic whites, the large number of black and Hispanic participants in the registry provided a unique opportunity to study the prevalence of AIDs in these minority populations. Whereas non-Hispanic white participants had the highest prevalence of AIDs, our data demonstrate that the prevalence of associated autoimmunity is also substantial in racial and ethnic minorities, including greater than 20% of Hispanic/Latino and 12% of black non-Hispanic T1D individuals. A significant difference likely exists in the prevalence of specific AIDs in each population. In support of this, studies have found an increased risk of systemic lupus erythematosus in Hispanic and black women as well as more prevalent Graves’ disease in non-Hispanic blacks and Asians (37, 38). In both lupus and Graves’ disease, early diagnosis and intervention can reduce end-organ damage and improve overall clinical outcomes; therefore, it is important that T1D providers be cognizant of the possibility of these AIDs in at-risk groups. With additional analyses of our data set in the future, we may be able to identify specific AID prevalence by race and to draw more conclusions about how genetic factors contribute to the development of each AIDs in the context of preexisting T1D.

Our data have several implications for AID screening in individuals with T1D. First, because there is a large increase in AID frequency after adolescence, screening efforts should be reinforced in the early-adult population (age 18–26 y) to identify clinically silent but developing tissue-specific autoimmunity. Screening may need to continue in patients with advanced age because our cohort showed no maximal age cutoff for identifying additional AIDs after an individual has developed T1D. In all age groups, thyroid disease and celiac disease should be a priority for screening, given their high prevalence and potential long-term complications of untreated disease (12, 39, 40). For other AIDs with lower prevalence, our evidence does not support routine screening, unless there are signs and symptoms suggestive of a new autoimmune diagnosis. Screening for Addison’s disease should be done for patients with a positive family history.

Limitations of this study include its retrospective nature and the relative young age of the cohort. Also, because we relied on clinical history rather than autoantibody positivity, AID prevalence was likely underestimated. The fact that the cohort is analyzed at a single time point also limits our ability to extrapolate true AID frequency in older age groups. Future studies should endeavor to identify AID risk at the earliest time point, whether it is autoantibody positivity or DNA analysis for human leukocyte antigen (HLA) genotyping. For example, children carrying the HLA-DRB1 and DQB1 alleles may be at increased susceptibility to celiac disease and RA in addition to T1D (41–43). Non-HLA loci including CTLA4, PD1, and pathways regulating inflammation have now been identified as potential modulators in T1D and may also predict susceptibility to other AIDs (44–48), including CTLA4 as a shared susceptibility gene for T1D and autoimmune thyroid disease (49, 50). These markers may reveal why it is that AIDs tent to cosegregate in certain demographic groups and what are the contributions from genetic vs environmental factors for the development of AIDs. Future studies should also include date of onset for AIDs to show whether the presence of T1D affects the development of other autoimmune diseases, or vice versa, and to better understand disease associations.

Acknowledgments

Abbreviations

Reference