Recent Insights in the Epidemiology of Autoimmune Diseases: Improved Prevalence Estimates and Understanding of Clustering of Diseases

Abstract

Previous studies have estimated a prevalence of a broad grouping of autoimmune diseases of 3.2%, based on literature review of studies published between 1965 and 1995, and 5.3%, based on national hospitalization registry data in Denmark. We examine more recent studies pertaining to the prevalence of 29 autoimmune diseases, and use these data to correct for the underascertainment of some diseases in the hospitalization registry data. This analysis results in an estimated prevalence of 7.6–9.4%, depending on the size of the correction factor used. The rates for most diseases for which data are available from many geographic regions span overlapping ranges. We also review studies of the co-occurrence of diseases within individuals and within families, focusing on specific pairs of diseases to better distinguish patterns that may result in insights pertaining to shared etiological pathways. Overall, data support a tendency for autoimmune diseases to co-occur at greater than expected rates within proband patients and their families, but this does not appear to be a uniform phenomenon across all diseases. Multiple sclerosis and rheumatoid arthritis is one disease pair that appears to have a decreased chance of coexistence.

1. Introduction

A seminal paper in the epidemiology of autoimmune diseases, published in 1997, gathered and synthesized 30 years of studies on 24 autoimmune diseases [1]. This compilation was a comprehensive analysis of the incidence, prevalence, and temporal changes in disease patterns. It was also the first comprehensive analysis of the totality of the burden represented by this collection of diseases. Applying the individual disease rates to the United States population, Jacobson et al. estimated that the prevalence of these 24 autoimmune diseases was 3.2%. A different approach was taken by Eaton et al. [2], who used national hospitalization registry data in Denmark from 1977 to 2001 to estimate the prevalence of 31 autoimmune diseases, and the co-occurrence of diseases within individuals and within families. The estimated prevalence of these autoimmune diseases was 5.3%. Walsh and Rau extended the analysis of the burden of autoimmune diseases by analyzing its relative ranking in terms of mortality risk among women under the ages of 65 [3]. Within each age group, the collection of 24 autoimmune diseases specified by Jacobsen et al. [1] ranked within the top 10 causes of death. These studies have been instrumental in promoting funding for autoimmune disease research, and in promoting connections between researchers and advocacy groups focusing on specific diseases.

A limitation of the Jacobsen et al. study [1] is that for some diseases, the data used were from studies conducted more than 30 years ago. In addition, data were aggregated across all geographic areas, although incidence and prevalence rates may vary by more than an order of magnitude. In this paper, we examine recent studies pertaining to the prevalence of a broad array of autoimmune diseases, separating out data from different areas. We also review studies of the co-occurrence of diseases within individuals and within families, focusing on specific pairs of diseases to better distinguish patterns that may result in insights pertaining to shared etiological pathways. Specific recommendations regarding the design of future epidemiologic studies that could address limitations identified in the current literature are also discussed.

2. Autoimmune Disease Prevalence Data

The prevalence studies included in this analysis were limited to studies in which the date used to define prevalence was within the last 20 years (1989–2008). For diseases and areas in which numerous studies were available, however, we used a more restrictive time period (1995 – 2008 for myasthenia gravis and rheumatoid arthritis; 2000 – 2008 for celiac disease, Crohn disease, ulcerative colitis, type 1 diabetes, multiple sclerosis, and systemic lupus erythematosus). We did not find any prevalence studies meeting this time restriction for Guillain Barre syndrome, pemphigoid, phemphigus, dermatomyositis, polymyositis, or hematologic conditions (autoimmune hemolytic anemia, idiopathic thrombocytopenic purpura, or percicious anemia), so these diseases are not included in the summary table (Table 1). In addition, we excluded two of the diseases included in Eaton et al.'s study [2], interstitial cystitis and endometriosis, because there is less certainty regarding their classification as autoimmune conditions. Thus, this analysis of prevalence data is based on 29 diseases [4–145].

Table 1.

Recent Prevalence Data for Autoimmune Diseases, by Geographic Area

	Hospital-based Data, Denmarkb	Hospital and non-Hospital- based Data
		Studies from Europe, North America, Australia, New Zealand			Studies from Asia, Middle East, Caribbean, South America
Diseasea	Rate per 100,000	Rate per 100,000	Study Area	Reference	Rate per 100,000	Study Area	Reference
Addison	18	11–14	UK, Italy, Norway	[4–6]
Alopecia	21	1700	US	[7]
Celiac disease	50	180–350	Greece, Netherlands	[8,9]	140–280	Iran, Tunisia	[13–15]
		740–1000	Iceland, Italy	[10,11]	470–600	Brazil, Argentina	[16, 17]
		1900	Finland	[12]	900	Turkey	[18]
Crohn disease	225	28–53	Bosnia-Herzegovina, Hungary	[19,20]	6–53	Puerto Rico, Malaysia, Lebanon	[27–29]
		96–201	US, Spain, Denmark, New Zealand	[21–26]	113	Israel	[30]
Ulcerative colitis	378	143–294	US, Hungary, Denmark, New Zealand	[19,21–23,25, 26]	6	Lebanon	[29]
					102	Puerto Rico	[27]
Diabetes
All ages	946	118	Lithuania	[31]
All ages		340–570	UK, Sweden, Australia	[32–34]
Ages < 20		87–120	Spain, Germany	[35, 36]	31	Bahamas	[40]
Ages < 20		227–355	US, New Zealand	[37, 38]	110–270	Kuwait, Saudi Arabia	[41,42]
Ages < 20		70	US- American Indian	[39]
Liver - Chronic active hepatitis	45	11–17	Spain, Sweden, Norway	[43–45]	3–8	Singapore	[47]
		36	US-Alaska Natives	[46]
Liver - Primary biliary cirrhosis	12	15–40	Norway, Finland, Spain, UK,	[45,48–51]	4–18	Israel	[53]
		4–20	US Australia	[52]
Thyroid - Hyper	629	500	US	[54]	20	Iran	[56]
		626	UK	[55]
Thyroid - Hypo	62	300	US	[54]	350	Iran	[56]
		2980	UK	[55]
Multiple sclerosis	182	177–358	US, Canada	[57–60]	4–20	Colombia, Brazil, Argentina	[72–74]
		100	Canada-First Nations	[60]	13	Japan	[75]
		121–200	Italy, Greece, France, Ireland,	[61–69]	11–62	Israel, Kuwait, Jordan, Iran	[76–79]
		46	Norway	[70]	101	Turkey	[80]
		50	Portugal New Zealand	[71]
Myasthenia gravis	18	8–15	Greece, Estonia, Croatia	[81–83]	3	Colombia	[87, 88]
			Netherlands, Sweden, UK	[84–86]	7	Curacuo and Aruba
Polymyalgia rheumatica	112	739b	US	[89, 90]
		150–370b	Greece	[91]
Psoriasis	197	696–1527	US, UK	[92, 93]
Psoriatic arthritis		57	Greece	[94]
		140–190	Iceland, Norway, Denmark, US	[95–98]
		500	Australia-Aboriginie	[99]
Rheumatoid arthritis	381	310–810	France, Hungary, Spain, Turkey, Greece, UK	[91,100–105]	120–280	Thailand, Phillipines, Vietnam, China	[106–109]
					510–550	India, Pakistan, Argentina	[110]
					197		[111]
Sjögren disease	48	3500	UK	[112]	330–770	China	[118]
		110	Denmark (ages 30–60)	[113]	720–1560	Turkey (women)	[119]
		93–150	Greece	[91,114,115]
		600	Greece - (women)	[116]
		600	Slovenia	[117]
Systemic sclerosis	23	5–34	France, Greece, Spain, Italy, US	[120–125]	20–24	Australia	[127]
		10–66	US-Native Americans	[126]	30	Phillipines	[107]
Systemic lupus erythematosus	32	34–150	US, Spain, Greece	[128–131]	19	Saudi Arabia	[133]
		42	Canada-1st Nations	[132]	45	Australia	[134]
					93	Australia-aboriginal	[134]
Systemic vasculitis		9–14	France, UK	[135, 136]	10	Australia	[137]
Wegener granulo-matosis	10	2–10	France, Norway, Australia, New Zealand	[135, 137–140]
Uveitis (iridocycltis)	149	69–115	US, Finland	[141–143]	730	India	[144]
Vitiligo	29				93	China	[145]

^aAddison = primary adrenal insufficiency; diabetes = type 1 diabetes, systemic vasculitis based on the total of polyarteritis nodosa, microscopic polyangiitis, Wegener's granulomatosis, and Churg-Strauss syndrome.

^bbased on data from Eaton et al., 2007; population-based study in Denmark using hospitalization records from 2001

^crates per population ages >= 50 years

As noted by Eaton et al. [2], the prevalence estimates from the Danish hospitalization registry data are significantly underestimated for diseases with low hospitalization rates. This underestimation is most evident for alopecia, celiac disease, hyperthyroidism, hypothyroidism, psoriasis and vitiligo, for which the rates are 5 to 10 times higher in studies from Europe or the United States using a broader ascertainment method. Although polymyalgia rheumatica and some of the Sjögren disease rates are also lower in the Eaton et al. data, the Eaton et al. estimates are not directly comparable to the estimates from the studies limited to older populations (e.g., ages ≥ 50 years). The celiac disease studies are primarily screening studies which include asymptomatic disease detected through antibody testing in conjunction with follow-up biopsies. In a study of 50,700 adults in the Netherlands, the prevalence of clinically diagnosed celiac disease (defined on the basis of adherence to a gluten free diet in conjunction with diagnosis confirmation) was 16 per 100,000, and the prevalence of undiagnosed disease was 350 per 100,000 [9].

How much would the total prevalence of autoimmune diseases estimated from Eaton et al. increase if corrected for this underascertainment? Multiplying Eaton et al.'s number of cases of alopecia, hyperthyroidism, hypothyroidism, psoriasis and vitiligo by 5, as a conservative adjustment factor, and excluding interstitial cystitis and endometriosis from the calculations, results in a total of 459,422 cases of autoimmune disease. This total needs to be adjusted for duplicate counting of comorbidities among individuals. In Eaton et al.'s analysis, there were 289,228 people with one or more diseases compared with a total number of diseases of 320,358. Applying the ratio of the number of affected people to the number of diseases (289,228 ÷ 320,358, or 0.90) to the new total of 459,422 diseases results in a figure of 414,719 affected individuals, which would be 7.6% of the total population of 5,472,032. Using a higher adjustment rate for four of these diseases (alopecia, hypothyroidism, psoriasis and vitiligo, each multiplied by 10), results in an estimated total prevalence of 9.0% Including a 10-fold correction for the underrepresentation of undiagnosed celiac disease increases the estimated total prevalence to 9.4%

There is a lack of current prevalence data from areas other than Europe and North America for many of the autoimmune diseases (Table 1). The rates for most diseases are similar (or span overlapping ranges) across geographic areas. For uveitis, however, the prevalence reported from one study in India [144] is much higher than the data from studies in the United States and Finland [142, 143]. In contrast, multiple sclerosis rates are, in general, higher in the United States and Europe compared with estimates from other countries (Table 1).

3. Co-Occurrence of Autoimmune Diseases

Autoimmune diseases have conventionally been considered as distinct disorders, likely as a consequence of their being treated by separate medical specialties based on organ system of involvement. Characterization of the extent to which particular combinations of autoimmune diseases occur in excess to that expected by chance may offer insight into shared pathophysiologic mechanisms and aid in the targeting of therapeutic strategies.

Clinical data on the associations among autoimmune diseases have predominantly been derived from anecdotal evidence or small studies from tertiary care settings. Despite the need for more large-scale, population-based epidemiologic research in this area, common clinical perception is that autoimmune diseases tend to co-exist both within individuals and families, and the concept of an autoimmune diathesis is widely accepted. Delineation of clinical patterns of aggregation, in concert with recognition of shared genetic features, will provide etiologic clues to this set of diseases. Since most autoimmune diseases are individually uncommon, often convenience samples from tertiary care settings are used, and achieving adequate statistical power to detect rare disease combinations is difficult. Thus, rather than providing an exhaustive account of all studies of autoimmune comorbidities, this overview is intended to highlight key studies with large or population-based samples, with adequate control or reference population data, if available for given disease combinations. Findings from major intra-individual and intra-family studies examining autoimmune disease associations within individuals are summarized in Tables 2 and 3, and studies within families are summarized in Table 4.

Table 2.

Intra-person coexistence of autoimmune diseases

Reference Location Data source; Measure of Associationa	Index disease cases; Control or reference population	Comorbid autoimmune disease(s)	Probands/Controls n (%); n (%) or n observed / n expected	Measure of Association (95% Confidence Interval)
Somers [146] United Kingdom, General Practice Research Database (pop-based); SIR	RA (n=22,888)	AIT	337 / 208.6	161.5 (145, 180)
	RA	MS	13 / 17.8	73.2 (39, 125)
	AIT (n=26,198)	RA	296 / 224.7	131.8 (117, 148)
	AIT	MS	23 / 20.7	111.0 (70, 167)
	MS (n=4,332)	RA	30 / 37.6	79.8 (54, 114)
	MS	AIT	61 / 42.2	144.4 (110, 185)
	T1DM (n=6,170)	RA	72 / 44.9	160.3 (125, 202)
	T1DM	AIT	175 / 39.0	449.0 (385, 521)
	T1DM; UK general population	MS	15 / 12.5	120.2 (67, 198)
Nielsen [148] Denmark Danish MS & Hospital Discharge Registers; SIR	MS (n=10,596); Danish general population	42 diseasesb	133 / 153.1	0.9 (0.7, 1.0)
		Ulcerative colitis	29 / 14.9	2.0 (1.4, 2.8)
		Pemphigoid	12 / 0.8	15.4 (8.7, 27.1)
		Pemphigus	2 / 0.03	53.6 (13.4, 214.3)
		RA	28 / 53.0	0.5 (0.4, 0.8)
		Temporal arteritis	11 / 20.6	0.5 (0.3, 0.97)
Nielsen [147] Denmark Danish MS & Hospital Discharge Registers; SIR	MS (n=6078); Danish general population	T1DM	11 / 3.38	3.3 (1.8, 5.9)
Ramagopalan [149] Canada Longitudinal, pop-based MS study (19 centers); OR	MS (n=5031); Spousal controls (n=2707)	T1DM	19 (0.4); 24 (0.5)	0.7 (0.3, 1.6)c
		RA	153 (3.0); 66 (2.4)	1.3 (0.9, 1.7)c
		Ulcerative colitis	9 (0.2); 4 (0.2)	1.2 (0.3, 5.4)c
		Crohn disease	11 (0.2); 4 (0.2)	1.5 (0.4, 6.4)c
		Psoriasis	293 (5.8); 146 (5.4)	1.1 (0.9. 1.3)c
		Pernicious anemia	123 (2.4); 25 (0.9)	2.7 (1.7, 4.3)c
		SLE	28 (0.6); 7 (0.3)	2.2 (0.9, 5.9)c
		Vitiligo	35 (0.7); 12 (0.4)	1.6 (0.8, 3.3)c
		AITD	395 (7.9); 116 (4.3)	1.9 (1.5, 2.4)c
		MG	7 (0.1); 3 (0.1)	1.3 (0.3, 7.5)c
		≥1 of above	Not reported	1.1 (95% CI 0.86–1.2)c
Weng[151] California, US Kaiser Permanente Medical Care Plan; OR	IBD (n=12,601); Matched controls from database (n=50,404)	Psoriasis	242 (1.9); 495 (1.0)	1.7 (1.5, 2.0)
		T1DM	122 (1.0); 346 (0.7)	1.2 (0.9, 1.4)
		RA	165 (1.3); 307 (0.6)	1.9 (1.5, 2.3)
		MS	49 (0.4); 74 (0.2)	2.3 (1.6, 3.3)
		SLE	27 (0.2); 72 (0.1)	1.3 (0.8, 2.1)
		Vitiligo	23 (0.2); 53 (0.1)	1.4 (0.8, 2.4)
		AITD	32 (0.2); 104 (0.1)	1.1 (0.7, 1.6)
		Chron glomer	14 (0.1); 29 (0.1)	1.8 (0.9, 3.5)
		6 other diseases	46 (0.4); 77 (0.2)	2.1 (1.4, 3.0)
		≥1 of above	2139 (17.0); 5131 (10.2)	1.5 (1.4, 1.6)
Cohen [152] United States IMS Health & MarketScan (MKT) Claims databases; OR	IBD cases; matched controls from database	AS (MKT)	not reported	5.83 (3.93, 8.64)
		AS (IMS)	not reported	7.79 (5.81, 10.83)
	MKT: IBD cases (n=6,139); controls (n=64,556)	MS (MKT)	not reported	1.56 (1.18, 2.05)
		MS (IMS)	not reported	1.53 (1.23, 1.91)
		Psoriasis (MKT)	not reported	1.4 (1.27, 1.68)
	IMS: IBD cases (n=18,603); controls (n=66,969)	Psoriasis (IMS)	not reported	1.51 (1.32, 1.73)
		RA(MKT)	not reported	2.05 (1.84, 2.29)
		RA (IMS)	not reported	2.72 (2.43, 3.04)
Alkhateeb [169] United States/Canada & United Kingdom Vitiligo Foundation/Society members; SPR	Vitiligo n=2078; Published population rates	Addison	8 (0.38) / 0.1c	†8000 (3454, 15763)c
		IBD	14 (0.67) / 7.69c	†184.2 (101, 309)c
		Pernicious anemia	37(1.78) / 3.12c	†1185.9 (835, 1635)c
		SLE	4 (0.19) / 0.5c	†800.0 (218, 2048)c
		AITD	354 (17.0) / 39.5c	†896.2 (805, 995)c
		7 other diseasesb	see foonoteb	no association
Koulentaki [150] Crete Clinical series; SPR	Ulcerative colitis (n=412); Crete general population	Primary biliary cirrhosis	2 / 0.06	‡3258.0 (395, 11769)
Eaton [2] Denmark Danish Hospital Discharge Register	31 autoimmune diseases	465 pairwise comorbidities	see text for summary	see text for summary

^aOR, odds ratio; SIR, standardized incidence ratio (100 represents unity) and SPR, standardized prevalence ratio (100 represents unity)

^bData not shown for other pairs with no association detected

^ccalculation not published by study authors, therefore calculated by authors of review

Disease abbreviations used in table: AIT, autoimmune thyroiditis (hypo-); AITD, autoimmune thyroid disease (includes hypo- and hyper-); AS, ankylosing spondylitis; Chron glomer, chronic glomerulonephritis; IBD, inflammatory bowel disease; MG, myasthenia gravis; MS, multiple sclerosis; RA, rheumatoid arthritis; SLE, systemic lupus erythematosus; T1DM, type 1 (insulin-dependent) diabetes mellitus.

Table 3.

Potential autoimmune disease associations within individuals described in the literature

Autoimmune hemolytic anemia

Hem

Autoimmune thrombocytopenic purpura

↑

AITP

Primary adrenal insufficiency (Addison's)

Add

Alopecia

Alop

Celiac disease

Cel

Inflammatory bowel disease

↑

IBD

Diabetes (type 1)

T1DM

Liver - Chronic active hepatitis

Hep

Liver - Primary biliary cirrhosis

↑

PBC

Thyroid - Hyperthyroidism (Grave's)

Grav

Thyroid - Hypothyroidism (Hashimoto's)

↑

Hash

Multiple sclerosis

↑

↑

↑

MS

Myasthenia gravis

MG

Polymyalgia rheumatica

PmR

Psoriasis

↑

Psor

Psoriatic arthritis

PsA

Rheumatoid arthritis

↑

↑

↑

↓

RA

Ankylosing spondylitis

↑

AS

Idiopathic inflammatory myopathy

↑

↑

↑

IIM

Sjögren's syndrome

Sjog

Systemic sclerosis (scleroderma)

↑

SSc

Systemic lupus erythematosus

SLE

Systemic vasculitis

↑

↓

Vasc

Uveitis (iridocycltis)

Uvei

Pemphigoid

↑

↑

↑

PemD

Pemphigus

↑

↑

PemS

Vitiligo

↑

↑

↑

Vit

^↑positive association

^↓negative association

Table 4.

Autoimmune co-occurrence in first degree relatives of index disease probands and controls (with disease verification by physician or chart review)

Reference Location, Data source, Measure of Association	Index disease; Control probands or reference population	Comorbid autoimmune disease(s)	First degree relative of probands / controls n (%); n (%) or n observed / n expected	Measure of Association (95% Confidence Interval)
Ginn [158] Maryland, USA Consecutive IIM patients referred to National Institutes of Health, OR	IIM probands (n=21), proband relatives (n=151);	≥1 of 25 diseasesb	33 (21.9) / 7 (4.9)	5.5 (2.3, 12.9)
		AITD	13 (8.6) / 1 (0.7)	13.4 (1.9, 572.3)d
	Control probands (n=21), Control relatives (n=143)
Nielsen [147] Denmark Danish MS & Hospital Discharge Registers, SIR	MS probands (n=11,862), proband relatives (n= 14,771); Danish general population	T1DM	56 / 34.3	1.6 (1.3, 2.1)
Nielsen [148] Denmark Danish MS & Hospital Discharge Registers, SIR	MS probands (n=10,596), proband relatives (n=20,800); Danish general population	≥1 of 42 diseasesb polyarteritis nodosa	288; 235.9	1.2 (1.1, 1.4)
			4; 1.1	3.7 (1.4, 10.0)
		Addison disease	4; 1.2	3.4 (1.3, 9.0)
		Crohn disease	44; 31.3	1.4 (1.04, 1.9)
		UC	51; 39.1	1.3 (0.99, 1.7)
Broadley[155] United Kingdom MS referrals from throughout United Kingdom; spousal controls, OR	MS proband families (n=571); control families (n=375)c	≥1 of 11 diseases	96 (16.8) / 44 (11.7)	1.7 (1.1, 2.6)
Cooper [154] Canada SLE probands from 11 rheumatology clinics; population matched controls, OR	SLE proband relatives (n=626); Control relatives (n=267)	AIT	37 (5.9) / 6 (2.2)	2.7 (1.1, 8.0)d
		Grave	14 (2.2) / 0 (0)	e
		RA	15 (2.4) / 0 (0)	e
		SLE	15 (2.4) / 0 (0)	e
		SSc	3 (0.5) / 0 (0)	e
		Sjögren's	6 (1.0) / 0 (0)	e
		PM/DM	0 (0.0) / 0 (0)	e
		APS	1 (0.2) / 0 (0)	e
		Hemoltic anemia	1 (0.2) / 0 (0)	e
		MS	2 (0.3) / 0 (0)	e
		Vitiligo	0 (0) / 0 (0)	e
		≥1 of above	73 (11.7) / 6 (2.2)	5.7 (2.5, 16.3)d
Anaya [156] Colombia, Mulit-center cohort of consecutive T!DM patients; matched healthy controls	T1DM probands (n=98), proband relatives (n=312); Control probands (n=113), Control relatives (n=362)	≥1 of 18 diseases	26 (8.3) / 9 (2.4)	3.56 (1.64, 7.73)
		SLE	1 (0.3) / 0 (0)	e
		Vitiligo	1 (0.3) / 3 (0.8)	0.4 (0.0, 4.8)d
		AITD	15 (4.8) / 6 (1.7)	3 (1.15, 7.82)
Anaya [157] Colombia Sjogren's clinical cohort (females); matched cases without autoimmune disease from same clinic, OR	Sjögren's probands (n=101), proband relatives (n=876) Control (n=124), control relatives (n=857)	SLE	8 (7.9) / 1 (0.8)	7.9 (1.1, 350.4)d
		RA	15 (14.9) / 10 (8.1)	1.5 (0.6, 3.7)d
		SSc	2 (2.0) / 0 (0)	e
		PBC	1 (1.0) / 0 (0)	e
		Vitiligo	4 (4.0) / 3 (2.4)	1.3 (0.2, 8.9) d
		MS	1 (1.0) / 0 (0)	e
		T1DM	3 (3.0) / 1 (0.8)	2.9 (0.2, 154.6)d
		Grave	1 (1.0) / 1 (0.8)	1.0 (0.0, 76.9)d
		AIT	25 (2.9) / 17 (2.0)	1.5 (0.7, 2.9)d
		≥1 of above	56 (6.4) / 33 (3.9)	1.7 (1.1, 2.7)

^aOR, odds ratio; SIR, standardized incidence ratio (100 represents unity)

^bData not shown for other pairs with no association detected

^canalysis based on number of families, not number of relatives

^dcalculation not published by study authors, therefore calculated by authors of review

^eOR and 95% CIs not calculated due to zero cell count

Disease abbreviations used in table : AIT, autoimmune thyroiditis (hypo-); AITD, autoimmune thyroid disease (includes hypo- and hyper-); APS, antiphospholipid antibody syndrome; IIM, idiopathic inflammatory myopathy; MS, multiple sclerosis; PBC, primary biliary cirrhosis; PM/DM, polymyositis/dermatomyositis; RA, rheumatoid arthritis; SSc, systemic sclerosis (scleroderma); SLE, systemic lupus erythematosus; T1DM, type 1 (insulin-dependent) diabetes mellitus ; UC, ulcerative colitis

Disease Co-occurrence Within Individuals

Studies examining the coexistence among autoimmune diseases typically begin with patients with an index disease, and assess the occurrence of other comorbid conditions within the index population. As outlined in Table 2, the key studies on coexistence have focused in particular on the following diseases as index conditions: multiple sclerosis, rheumatoid arthritis, autoimmune thyroiditis (hypothyroidism), type 1 diabetes, inflammatory bowel disease, and vitiligo. The most comprehensive of such studies was based on the United Kingdom (UK) General Practice Research Database (GPRD) [146]. This population-based study included four index conditions, and was unique in that it incorporated the sequence of diagnoses in its design so that diagnosis of the index condition preceded diagnosis of the comorbid condition. In contrast, classification of the index condition in other studies was based on special interest or availability of a particular patient population, and did not imply that the index condition occurred before the comorbid conditions. After controlling for age and calendar year in the UK GPRD study, Somers et al. found increased coexistence of rheumatoid arthritis, thyroiditis, and type 1 diabetes versus that expected based on population-based incidence rates during the same time period. The most striking association was for thyroiditis among type 1 diabetes patients, where there was a greater than fourfold excess risk of thyroiditis than expected. However, this study documented an inverse association between rheumatoid arthritis and multiple sclerosis, regardless of diagnostic sequence. Sex-specific results were similar to those for both sexes combined.

Multiple sclerosis as an index disease was also examined in other population-based studies. In two Danish record linkage studies by Nielsen et al., increased incidence of type 1 diabetes [147], ulcerative colitis, pemphigoid and pemphigus foliaceus, but decreased incidence of rheumatoid arthritis and temporal arteritis [148], were found compared to the general Danish population. Ramagopalan et al. found increased occurrence of pernicious anemia and autoimmune thyroid disease (hypo- or hyper-) in a Canadian case-control study using spousal controls, but did not detect an overall increase in autoimmune disease based on ten diseases studied [149].

Three studies examined index inflammatory bowel disease and comorbid autoimmune conditions. In an index population of ulcerative colitis patients in Crete, Koulentaki et al found an 30 fold increase of primary biliary cirrhosis [150]. Weng et al. [151] studied a number of comorbid autoimmune diseases among inflammatory bowel disease patients and matched controls in the Kaiser Permanente Medical Care Plan (USA), and found significantly increased occurrence of psoriasis, rheumatoid arthritis, multiple sclerosis and combined category of 6 other diseases (Addison disease, hemolytic anemia, primary biliary cirrhosis, immune thrombocytopenia purpura, Sjögren disease and systemic sclerosis). They did not detect an association for the following comorbid conditions: type 1 diabetes, systemic lupus erythematosus, vitiligo, autoimmune thyroid disease (Grave's & Hashimoto's combined), or chronic glomerulonephritis. Cohen et al. [152] analyzed data from two medical claims databases from the United States, and likewise found significantly increased occurrence of rheumatoid arthritis, multiple sclerosis and psoriasis. Further, they detected a positive association with ankylosing spondylitis.

Eaton et al. examined the associations between 31 autoimmune diseases based on Danish hospital data [2]. Since this enabled the estimation of 465 pairwise comorbidities, these data are not summarized in Table 2. However, as synthesized by the authors, a few patterns emerged, including a tendency for positive associations between pairs (only 12 negative associations were found, i.e., ORs < 1.0), and the connective tissue diseases in general had higher comorbidities than other types of autoimmune diseases. A simplified tabulation of the intra-individual associations observed Eaton et al. [2] and the other studies described above is presented in Table 3.

Disease Co-occurrence Within Families

Studies of autoimmune diseases within families have often been restricted to assessing the occurrence of a proband disease in pedigrees, but a growing number have examined the aggregation of various autoimmune diseases within families of case and control probands. As summarized elsewhere, such research tends to indicate familial predilection for both proband and additional autoimmune diseases [153]. However, many studies have relied on self-reported family history, and reliability of such data may be questionable. For instance, in a study of relatives of systemic lupus erythematosus patients and population controls, the total confirmation rate of self-reported family history of autoimmune diagnoses excluding cases with unavailable medical records was 76%; the rate was 44% when all reported cases were included [154]. Another study of familial autoimmunity found that females tend to be more aware of family medical history versus males; moreover, males reported significantly higher rates of autoimmune disease in the presence of their spouse versus when their spouse was absent [149]. Given these considerations, we have restricted our summary of family studies to those with physician or medical record verification of autoimmune disease diagnoses in relatives of proband cases and controls (Table 4). However, aside from studies based on record linkage (e.g., the Danish studies utilizing national registers), most family studies were only able to confirm positive reports of autoimmune disease. Thus underascertainment of familial cases remains a concern, particularly if recognition of family history of autoimmune disease is differential between case and control probands.

In the Danish study by Eaton et al. examining 31 autoimmune diseases, the authors observed high familial autoimmunity for individual diseases, i.e., the occurrence of the proband (index) disease within family members, with a tendency for a higher magnitude of association within siblings versus parent-offspring pairs [2]. However, they found little evidence to support the aggregation of other (non-index) autoimmune diseases within families. The authors interpret these observations to suggest that the genetic origins for autoimmune diseases are more specific than general, and that the finding of stronger associations within sib versus parent-offspring pairs further emphasizes the role of the environment.

In contrast, other family studies have documented significant increases in various autoimmune diseases among first degree relatives of case versus control probands. In a Danish study of multiple sclerosis probands by Nielsen et al. which assessed rates of 42 autoimmune diseases, increased rates were found for polyarteritis nodosa, Addison's, and Crohn's, as well as autoimmune disease in general (based on the composite category for the 42 diseases) [148]. In a separate study, Nielsen et al. found increased type 1 diabetes in families of multiple sclerosis probands [147]. A case control study using spousal controls by Broadley et al. found increases in autoimmune disease based on a composite of 9 diseases, as well as increases in Hashimoto's and Graves disease [155].

A study of systemic lupus erythematosus probands by Cooper et al. found increased Hashimoto's and autoimmune disease in general (based on 11 diseases) in the relatives of lupus patients versus control relatives [154] Anaya et al. found an increase in autoimmune thyroid disease (hypo- or hyper-) in type 1 diabetes patients [156], and in another study by Anaya et al. found evidence for increases in systemic lupus erythematosus and general autoimmune disease (based on 9 diseases) in the relatives of probands with primary Sjögren disease [157]. Finally, a study of idiopathic inflammatory myopathy patients found increases in general autoimmune disease (based on 25 diseases) and autoimmune thyroid disease (hypo- or hyper-) [158]. It may be more feasible to detect statistically significant associations with thyroid diseases than with less common autoimmune diseases.

4. Discussion

Jacobsen et al.'s review of studies published from 1965 to 1995 resulted in an estimated prevalence of a broad group of 24 autoimmune diseases of 3.2% [1]. However, this estimated was limited in terms of the number of diseases included and the lack of recent data for many diseases. This approach also does not taken into account the co-occurrence of diseases within an individual, and thus is an estimate of the number of individual diseases diagnosed within a population rather than the number of people affected by any of group of autoimmune diseases. The approach based on hospitalization registry data in Denmark from 1977 to 2001 presented by Eaton et al. allows for the direct estimation of the prevalence of a group of disease in a population, without double-counting of individuals. The estimated prevalence of 31 diseases in this study was 5.3% [2]. Applying a correction to Eaton et al.'s estimates for 6 diseases for which reliance on hospitalization data produced significant underascertainment (alopecia, celiac disease, hyperthyroidism, hypothyroidism, psoriasis and vitiligo), we calculated a corrected prevalence of 7.6 – 9.4% (depending on the size of the correction factor). Although considerable variation in the rates for some diseases has been seen (e.g., multiple sclerosis), the rates for most diseases for which data are available from many geographic regions span overlapping ranges. Thus although this estimate is primarily based on a population-based study from Denmark, it may apply to countries in other parts of the world.

Estimates of the prevalence of individual diseases and of individuals with any of a group of diseases provide data needed for health policy discussions and in setting research priorities, and can aid in health services planning. Expansion and enhancement of the database of disease incidence and prevalence studies within and across specific geographic areas would also allow for the assessment of temporal trends in disease rates. There are many challenges, however, to conducting population-based epidemiologic studies of the prevalence or incidence of specific autoimmune diseases. Many of these diseases are relatively rare and are characterized by heterogeneous clinical presentations and complex case definitions. Few are routinely diagnosed based on a biopsy. For some diseases, the International Classification of Diseases (ICD) system does not provide a specific code. For example, we have found at least four ICD-9-CM codes used for antiphospholipid antibody syndrome, including: other and unspecified nonspecific immunological findings (795.79), primary hypercoagulable state, (289.81), hemorrhagic disorder due to circulating anticoagulants (286.5), and other and unspecified coagulation defects (286.9). Another example of the limitation of ICD coding is diabetes. The most recent revision of the ICD classification system, ICD-10, was the first to distinguish between type 1 and type 2 diabetes, and it will take some time to determine the accuracy of this new coding scheme. Improving the specificity of disease codes, and facilitating their adaptation across medical specialties, would enable better estimates of these diseases.

Hospitalization data registries, where available for a population, can provide a valuable resource for epidemiologic studies. Because of the underascertainment that is inherent in a registry that only includes hospitalized patients, however, the usefulness of these registries would be greatly enhanced by an evidence-based answer to specific questions regarding the sensitivity and specificity of the hospitalization data used for disease classification. For example, a clinical study that is based on an inception cohort of patients with a specific disease could be analyzed to determine the proportion of patients who are hospitalized within a period of the diagnosis, and for these patients, the proportion who would have been correctly identified as having the disease based on the hospitalization coding data. For some diseases, additional sources of case ascertainment, such as laboratory or biopsy records and pharmacy records may be needed to accurately estimate disease incidence or prevalence. Inclusion of multiple sources allows for capture-recapture analytic methods, which may further improve the accuracy of the resulting estimates [159]

Death certificate data have also been used to estimate the relative ranking of autoimmune diseases among causes of death [3]. Studies of several autoimmune diseases have reported considerable under-reporting of these diseases on death certificates, however, even when contributing causes of death are included in the analysis. For example, in studies of multiple sclerosis patients in the United Kingdom [160] and type 1 diabetes patients in Germany [161], 27% and 29%, respectively of the death certificates did not include any mention of these respective diseases. Underascertainment rates were higher (40% and 42%, respectively) in studies of systemic lupus erythematosus [162] and rheumatoid arthritis [163]

Aggregate findings from the literature pertaining to co-occurrence of diseases indicate that certain autoimmune diseases co-occur at greater than expected rates within patients and their families, though a small number of disease pairs (e.g., multiple sclerosis and rheumatoid arthritis) appear to have a decreased chance of coexistence. Disease associations are more evident within individuals, and in families stronger associations have been found among sibling versus parent-offspring pairs. The precise measurement of the frequency of multiply affected individuals and how this varies by disease, time, and place is likely to reflect gene-environment interaction.

Rapid expansion of knowledge related to the genetics of autoimmune disease is currently underway. Data from twin, linkage and association studies point to a complex mode of inheritance, and indicate that genes involved in autoimmune disorders are pleiotropic rather than disease specific. Becker describes the “common variant/multiple disease” hypothesis stating that common alleles manifesting in a given disorder under particular genetic/environmental conditions may have the potential to give rise to alternate clinical phenotypes when combined with a different set of genetic and environmental factors [164]. Accumulating data support the premise that clinically distinct autoimmune diseases may have common susceptibility genes. For instance, a major United Kingdom–based genome-wide association study published in 2007 identified several loci with associations to more than one autoimmune disease {, 2007 #432}. As summarized by Lettre and Rioux, at least 68 genetic risk variants have now been associated with various autoimmune diseases, in contrast to only 15 that were recognized prior to 2006 [166].

Both genetic and epidemiologic data indicate that predisposition likely exists for particular combinations of autoimmune diseases, but not in a uniform fashion across all autoimmune diseases. For instance, the PTPN22 risk allele has been strongly associated with type 1 diabetes, rheumatoid arthritis and thyroiditis, but not multiple sclerosis [167]. This finding is compatible with clinical results Somers et al.'s population-based UK study encompassing 33.5 million person-years of data, which documented intra-individual coexistence of type 1 diabetes, rheumatoid arthritis and thyroiditis beyond that expected, but suggestion of reduced risk between multiple sclerosis and rheumatoid arthritis for either diagnostic sequence (standardized incidence ratios 79.8 and 73.2 and for index diagnosis of multiple sclerosis and rheumatoid arthritis, respectively) [146]. An inverse association between multiple sclerosis and rheumatoid arthritis was likewise found in two population-based Danish studies by Eaton et al. [2] and Nielsen et al. [148] (odds ratios 0.8 and 0.5, respectively), and a systematic review of the literature published in 2006 [168]. Coupling results from such studies will aid in the understanding of the clinical relevance of proposed autoimmune susceptibility genes. Further characterization of the concordance between genetic and epidemiologic evidence will enhance our understanding of autoimmune disease pathways.

Improved delineation of the prevalence, incidence and coexistence of autoimmune diseases is also important for the interpretation of pharmacoepidemiologic data. For example, post-marketing surveillance of TNF inhibitors signaled the possible development of multiple sclerosis, though without prior availability of data on the coexistence of rheumatoid arthritis and multiple sclerosis, it could be argued that rheumatoid arthritis patients have an inherent predilection for the development of multiple sclerosis. As discussed by Somers et al., in light of recent data indicating an inverse association between rheumatoid arthritis and multiple sclerosis, the development of multiple sclerosis in patients treated with TNF inhibitors can reasonably be ascribed as a risk of treatment [146].

Continued and improved surveillance of autoimmune diseases around the world will improve our understanding of disease burden and temporal trends. Since autoimmune diseases are conventionally treated by separate medical specialties according to type of organ involvement, there have been missed opportunities to study these diseases as a group. Further studies of the relationships among autoimmune diseases are indicated in order to enhance our understanding of the etiology of this set of diseases. We are pleased to contribute to this special issue of the Journal of Autoimmunity in dedication of Dr. Noel Rose's outstanding contributions to autoimmunology, including the epidemiology of autoimmune disease [170–176], the establishment of the American Autoimmune Related Disease Association (AARDA) [177] and we note that this issue is part of the Journal of Autoimmunity's recognition of outstanding figures in immunology [178–180].