How do autoimmune diseases cluster in families? A systematic review and meta-analysis

A literature search was done even though 'familial autoimmunity' is not a Medical Subject Headings (MeSH) term. Nevertheless, the search was done in the electronic databases Medline and Embase, and included articles, from 1966 for the former and 1980 for the latter, up to June 2012. The search strategy was limited to humans and included the words '(familial OR clustering OR aggregation)' AND 'autoimmunity followed by each of the diseases we have focused on: 'multiple sclerosis,' 'diabetes mellitus, Type 1,' 'arthritis, rheumatoid' and 'lupus erythematosus, systemic' using MeSH terms and key words for 'autoimmune thyroid disease'. In order not to miss potentially eligible studies we used wild cards for the words familial, clustering and aggregation in the following manner: famil*, aggrega* and cluster*. No language restrictions were used. Articles were included if they fulfilled the following conditions: ADs diagnosis was carried out according to international criteria or through international classification of diseases, articles were published as full articles and, as mentioned earlier, if ADs in first degree relatives (FDRs) were different than in the proband. Studies were excluded if they only referred to autoantibody prevalence, if a clear cut distinction between diseases was not possible, if it was not possible to distinguish between probands and FDRs, if the studies were case reports, and if they dealt with a single family. Unpublished data were also excluded. Eligibility assessment was done by a primary reviewer who screened all titles and abstracts of publications. Retrieved articles were rejected if eligibility criteria were not met and a secondary reviewer was consulted in cases in which eligibility criteria were unclear. References from the articles that seemed to be relevant for our review were hand-searched. All articles were assessed according to the Oxford Centre for Evidence-based Medicine 2011 Levels of Evidence [28]. The search returned articles in which familial autoimmunity was assessed in other ADs and they were included. From each study we extracted data including total number of FDRs, numbers of FDRs affected, prevalence of ADs and, where possible, extraction of crude and adjusted measures of association, that is, odds ratio (OR) or risk ratio (RR). With the prevalences extracted, aggregation for different ADs across the five index diseases mentioned earlier was calculated by dividing the prevalence of a given AD in FDRs by the prevalence in the general population (λ_relatives). We extracted data on prevalences from five reports [2, 29‐32]. Inclusion criteria for the meta-analyses were applied to publications that provided epidemiological data on risk factors, RR and OR with confidence intervals (CI), or that provided information that allowed us to calculate these data. If the study did not report the number of subjects in each group, either the RR or the OR with the CI, must have been reported in order for them to be included in the meta-analyses calculations.

In order to study aggregation, we determined worldwide prevalences of ADs from five studies mentioned earlier [2, 29‐32]. If a range was reported, we arbitrarily calculated the mean.

Data were analyzed using the Comprehensive Meta-Analysis Version 2 program (Biostat, Englewood, NJ, 2004). Calculations were carried out for the whole group of articles depending on the binary data available for any AD: number of subjects and risk data (OR and RR with the corresponding 95% CI). Effect size was calculated based on studies that reported an OR with its respective 95% CI and from raw data given by case-control and cohort studies. If raw data from cohort studies were available, a second effect size was calculated with studies that only showed the RR and the respective 95% CI. Different study designs were used to compute the same effect size since the effect size had the same meaning in all studies and was comparable in relevant aspects. In order to perform the analyses, the association measures were transformed to log values, and then the results were converted back to ratio values for presentation. This approach prevented the omission of studies that used an alternative measure. Two types of meta-analyses were done in order to analyze autoimmunity as a trait. First, a given AD in FDRs was analyzed through all the studies regardless of the AD of the proband. The second type of meta-analysis analyzed ADs in FDRs through all the studies from a specific AD present in the proband.

Additional meta-analyses were done for studies with complex data structures and non-cumulative results as the information for the different effects was not totally independent. This is the case for studies reporting multiple independent subgroups, that is, aggregation for son and daughter separately, within a study. A flow diagram of the current study is shown in Figure 2.

ORs were grouped by weighing individual ORs by the inverse of their variance. For each analysis, the final effect OR and 95% CI were obtained by means of the random effect model, which was preferred because it accepts distributions of true effect sizes rather than one true effect and assigns a more balanced weight to each study. It was also used because all the studies were considered to be unequal in terms of specific ADs.

Heterogeneity was calculated by means of Higgins's (I ² ) tests. The variance between studies was estimated by the DerSimonian and Laird method. The I ² test showed the proportion of observed dispersion that was real rather than spurious and was expressed as a ratio ranging from 0% to 100%. I ² values of 25%, 50%, and 75% were qualitatively classified as low, moderate, and high, respectively. Publication bias was determined using Funnel plots and Egger's regression asymmetry tests.

After discarding duplicates, the search in both databases retrieved 2,552 articles. In a first assessment we considered 61 articles to be eligible. In a second screening 17 of these articles were not eligible due to reporting inconsistencies, such as not distinguishing between probands and FDRs. As we did not identify other articles from the reference lists, only 44 articles met eligibility criteria [32‐75]. Figure 3 and Table 1 summarize the search results. Although ankylosing spondylitis (AS) is considered an auto-inflammatory more than autoimmune disease [76], it was included in the results since it was found to aggregate in families. Most of the studies found lacked controls and had a small sample size, which is reflected in low grading according to the 2011 levels of evidence from the Oxford Centre for Evidence-based Medicine [28]. Detailed information is shown in Table 2.

Three articles assessed familial autoimmunity for autoimmune thyroid disease (AITD) [33‐35]. AITD encompasses Graves' disease as well as Hashimoto's thyroiditis with the latter being the most common cause of acquired hypothyroidism [77]. Moreover, AITD is the most common AD [78]. Various studies have shown that AITD coexists with other ADs in the same subject [79, 80] and it has also been shown that there is familial clustering of AITD in FDRs, particularly in female relatives [81]. Boelaert et al. [33] described familial autoimmunity among probands with Hashimoto's thyroiditis or Graves' disease. Both ADs were significantly associated with the presence of type 1 diabetes mellitus (T1D), rheumatoid arthritis (RA), pernicious anemia (PA), systemic lupus erythematosus (SLE), celiac disease (CD), vitiligo (VIT) and multiple sclerosis (MS). Only Graves' disease was associated with Addison's disease (AdD) and inflammatory bowel disease (IBD). Compared with the general population, familial autoimmunity in Graves' disease probands disclosed PA as the strongest association (RR: 14.1; 95% CI: 11.48 to 17.03), followed by RA (RR: 13.5; 95% CI: 12.32 to 14.86).

Hemminki et al. [34] assessed familial autoimmunity only in probands with Graves' disease from Sweden. To calculate familial risk within a large community based cohort they calculated standardized incidence ratios (SIR) as the ratio between the observed and the expected frequency for each disease. A value over one indicates an increased frequency of what is expected whereas a value below one indicates a decreased frequency. The analysis was stratified according to the FDR involved. For a single parent affected, Hashimoto's disease, PA, and RA were the only diseases significantly associated, having a SIR of 2.04, 1.82 and 1.48, respectively, thus showing an increased frequency of what is expected. Significant associations for singleton siblings were found for T1D, discoid lupus and localized scleroderma, having a SIR of 2.14, 6.03 and 6.62, respectively. If a parent and a sibling were affected with the same AD, the significant association was between Hashimoto's disease with a SIR of 37.41 and SLE with a SIR of 14.33 [34].

The search returned seven articles about T1D probands [35‐41]. AITD was responsible for the familial autoimmunity found in most of the articles [36, 40], even when compared to control subjects [37]. Wagner et al. [38] replicated the results but also described the presence of CD, psoriasis (PSO), and VIT.

Hemminki et al. [39] also reported familial autoimmunity in probands with T1D. When a parent had AdD, the SIR for T1D in offspring was 2.41. It was 2.73 for CD, 1.83 for Graves' disease, 2.13 for Hashimoto's thyroiditis, 3.09 for PA, 3.63 for primary biliary cirrhosis (PBC), 2.12 for RA, 1.62 for SLE, 1.23 for ulcerative colitis (UC), and 1.23 for Wegener's granulomatosis (WG). Only the presence of AdD, CD or Graves' disease in singleton siblings was associated with T1D in probands. Likewise, when a parent and sibling had RA, the SIR for T1D was 5.34 [39].

Four articles assessed familial autoimmunity in SLE probands. Alarcon-Segovia et al. evaluated familial aggregation in the 'Grupo Latinoamericano de Estudio de Lupus' (GLADEL) [42]. They found that among all family members who had any AD, 6.7% had RA, 2% AITD and other ADs at a lesser frequency. In FDR (n = 114) with ADs, 28% (n = 32) had RA and 16% (n = 32) had AITD [42]. Likewise, an increased frequency of familial autoimmunity was found in SLE probands compared with population prevalence. Priori et al. [43] found an OR of 4.6 (95% CI 1.94 to 11.1) in a multivariate analysis of familial autoimmunity in FDR of SLE patients. They reported AITD as the most frequent disease with eight cases, followed by RA with five cases, VIT with three cases and T1D with two cases. PSO frequency was higher among non-autoimmune controls. Sjögren´s syndrome (SS) as well as AITD were described by Scofield et al. [82] while Corporaal et al. [44] found clustering of MS and RA.

In RA, familial autoimmunity was ascertained in five articles, all of which linked AITD or T1D to AR. Lin et al. [45] showed an association with AITD in 7.8% of the probands and T1D in 2.8%. Thomas et al. [46] also reported T1D as the disease responsible for familial autoimmunity. In another study, Taneja et al. [83] stated that SLE, T1D, AITD, SS, PSO and systemic sclerosis (SSc) were found in families with RA. However, they included probands within this description, thus assessing and combining polyautoimmunity or MAS with familial autoimmunity. Walker et al. [84] found an excess risk for AITD in RA multicase families compared with the general population. However, this significance was lost when RA sufferers were withdrawn from the analysis. Jawaheer et al. [48] found the presence of AITD and other ADs in siblings but, compared with siblings of non-RA probands, the difference was not significant.

Hemminki et al. [47] also reported familial autoimmunity in probands with RA. Just as described above, when a parent had AS, the SIR for RA in offspring was 2.96. It was 2.25 for SS, 2.13 for SLE, 1.65 for SSc, 1.54 for AITD, 1.53 for PA, 1.36 for PSO and 1.34 for WG. When singleton siblings had PSO, the SIR for RA of the proband was 2.01 and 2.77 for SLE.

In our search, MS was the AD with the most articles assessing familial autoimmunity with 10 articles found [32, 35, 49‐56]. Some studies suggest that FDRs and other relatives of probands with MS could be at greater risk of ADs other than MS [32, 35, 49‐55, 57] while the studies done by Ramagopalan et al. [54] and Midgard et al. [85] do not support these findings. Although Annunziata et al. [86] found an association between MS and other ADs in first and second degree relatives, the results were not significant when compared to non-AD controls. Conversely, Alonso et al. [87] and Magaña et al. [88] found a significant association between MS and other ADs in relatives of any degree.

Using 265 families from the Multiple Autoimmune Disease Genetics Consortium (MADGC), Criswell et al. [35] compared the frequency of ADs in siblings of multiplex families stratified by seven ADs: AITD, RA, MS, SLE, T1D, IBD and PSO. These diseases were pre-specified given a variety of considerations. There was no evidence of familial autoimmunity except in the case of IBD patients in whose families MS was observed among FDRs (OR: 8.1; 95% CI: 1.77 to 37.0; P value = 0.018). However, selection bias was present as families selected for inclusion were not recruited in the same manner [35].

For the first effect size, OR, 13 meta-analyses were developed. Ten analyzed the proportion of a specific AD in FDRs independent of the AD present in the proband. Of these, three showed significant association: AITD, T1D and IBD. Three included an independent AD in FDRs in a specific AD of the proband, two of them showed significant associations: RA and MS. Figures 4 and 5 show the forest plots corresponding to six meta-analyses.

A second effect size was calculated based on data from studies showing RR data. Twenty eight meta-analyses were developed. Twenty three analyzed the proportion of a specific AD in the FDR through all the studies independent of any AD of the proband. Of these, nineteen showed significant association, the most relevant results being related to VIT, PA, RA and T1D. Additional results are shown in Additional file 1. Through all the studies, four additional analyses performed included any AD present in FDRs. All these analysis disclosed significant results. The ADs in the proband were AITD, MS, RA and T1D (Figures 6 and 7).

Evidence of significant publication bias was identified using the Egger test (P-value 2-tailed: <0.05) for two meta-analyses which included studies that reported OR with its respective 95% CI (T1D in FDR (P-value 2-tailed: 0.047) and MS in probands (P-value 2-tailed: 0.007)). One meta-analysis that reported RR data showed publication bias by the Egger test (AITD in probands (P-value 2-tailed: 0.008)) (Figure 6A). The corresponding funnel plot showing the standard error or the precision on the Y axis is shown in Additional file 2. Therefore, a second analysis was run in a search for publication bias. The classic fail-safe analysis indicated a number of missing studies that would give a P-value of >0.05. Begg and Mazumdar rank correlation was not significant and the trim and fill adjustment did not suggest a lower risk than the original analysis. Based on all the analyses for publication bias, we consider the impact of bias in the three meta-analyses to be trivial.

Familial autoimmunity as an outcome was also assessed in certain articles, particularly in MS and SLE probands (Figure 4D).

Several studies retrieved only reported prevalences of ADs in FDRs. Aggregation, based on data from five studies mentioned earlier in Table 3[2, 29‐32], is shown in Table 4, which discloses information on calculated aggregation for diverse ADs, in AITD, T1D, SLE, RA and MS.

The systematic search we performed retrieved other studies that assessed familial autoimmunity besides the five ADs we were focused on. These ADs are SSc, SS, IBD, juvenile dermatomyositis (JDM), VIT, juvenile rheumatoid arthritis (JRA), juvenile SLE (JSLE), idiopathic inflammatory myositis (IIM), CD, and alopecia areata (AA).

Four studies reported diagnosis of AITD, RA and SLE in FDRs of SSc probands [57‐60]. Frech et al. [60] found a RR of 2.49 (95% C.I. 1.99 to 3.41) for familial autoimmunity in FDRs and a RR of 1.48 (95% C.I.1.34 to 2.39) for familial autoimmunity in second degree relatives.

Two studies were found on this disease, one by Reveille et al. [61] and the other done by our group [62]. Both studies agreed on the occurrence of AITD and SLE among relatives. In addition to these ADs, we described the presence of RA [62] while Reveille et al. [61] reported the occurrence of MS and SSc.

Two studies were retrieved. As mentioned earlier, Criswell et al. [35] found an increased frequency of familial autoimmunity among probands with IBD. A study conducted by Hemminki et al. [63] assessed familial autoimmunity within IBD probands. In UC patients when a parent had AS the SIR for UC in offspring was 1.6, for Crohn's disease (CrD) 2.5, for T1D 1.2, for Graves' disease 1.3, for MS 1.4, for polyarteritis nodosa (PAN) 2.0, for PSO 1.3, for RA 1.1, and for SLE 1.5 [63]. When singleton siblings had CD, the SIR for UC was 2.5, and for AS was 2.1. When a parent and a sibling had CrD the SIR for UC was 4.7 and for PSO was 4.3. In CrD patients, when a parent was diagnosed with UC, the SIR for CrD in offspring was 2.4, for AS was 1.8 and for PSO was 1.4. When singleton siblings had UC, the SIR was 2.8 and for AS was 2.1. When a parent and a sibling had UC, the SIR for CD was 5.0 and for RA was 2.2. In twins, the SIR for CrD-UC pairs was 4.9 [63].

For VIT, three studies assessed familial autoimmunity. The studies done by Alkhateeb et al. [64] and Laberge et al. [65] discovered a significant increase in the occurrence of three ADs other than VIT, namely, AITD, PA and AdD. Alkhateeb et al. also reported the occurrence of SLE, myasthenia gravis (MG) and SSc [64], while Laberge et al. found the presence of PSO and RA [65]. In Chinese patients, Zhang et al. [66] found a significant association with RA, AA and PSO.

Two studies were found on familial autoimmunity in JRA [67, 68]. Prahalad et al. [67] found that AITD accounted for the familial autoimmunity seen in these probands. Huang et al. [68] found, in addition to AITD, the presence of PSO, AS and SLE. Furthermore, Huang et al. [68] compared the prevalence of ADs in family members of probands with JRA against the prevalence in family members of probands with JSLE. Including all family members (that is, first, second and third degree relatives), JSLE probands had a greater prevalence of familial autoimmunity than probands with JRA. Nonetheless, in FDR the prevalence of ADs was not significantly different between these two diseases. Thus, familial autoimmunity is equally present in JRA and JSLE. Likewise, Pachman et al. [89] compared JRA to JDM and to healthy controls. The only statistically significant association was an increased frequency of RA and PA in FDR of JRA probands.

Two articles were found for this disease. While Huang et al. [68] found that 17% of the FDRs of JSLE probands were affected with an AD, Walters et al. [69] found a prevalence of 51%, with 35% of FDRs from JSLE probands having SLE, 30% AITD and 13% PSO.

Familial autoimmunity has also been assessed for IIM in two studies. The study by Ginn et al. [70] found that the most common disease was, once again, AITD followed by RA, T1D and PSO. In this article, OR for familial aggregation of ADs was calculated irrespective of disease status (that is, case or control). The strongest predictors were a blood relative and female gender. Niewold et al. [71] reported that FDRs of probands with JDM had a higher frequency of T1D or SLE than in FDRs of controls. However, this relationship did not reach statistical significance.

Three articles were found. Petaros et al. [72] found that the prevalence of familial autoimmunity was 4.9% among first and second degree relatives. The ADs that became manifest were AITD, PSO and T1D. In line with these results, Cataldo et al. [73] found an increased prevalence of ADs including AITD and T1D. Neuhausen et al. [74] also found a significant association with T1D and JRA. However, contrary to what was expected, they found a decreased prevalence of AITD.

An increased frequency of AITD, VIT, T1D, PSO, and CD was found among FDRs of pediatric patients with AA [75].