Enterovirus infection and type 1 diabetes mellitus: systematic review and meta-analysis of observational molecular studiesBMJ 2011; 342 doi: http://dx.doi.org/10.1136/bmj.d35 (Published 03 February 2011) Cite this as: BMJ 2011;342:d35
- Wing-Chi G Yeung, medical student1,
- William D Rawlinson, professor12,
- Maria E Craig, associate professor234
- 1Faculty of Medicine, University of New South Wales, Sydney, NSW 2052, Australia
- 2Virology Research, POWH and UNSW Research Laboratories, Prince of Wales Hospital, Randwick, NSW 2031
- 3The Children’s Hospital at Westmead, Institute of Endocrinology and Diabetes, Sydney
- 4Discipline of Paediatrics and Child Health, University of Sydney, Sydney
- Correspondence to: M Craig, Institute of Endocrinology and Diabetes, The Children’s Hospital at Westmead, Locked Bag 4001, Westmead NSW 2145, Australia
- Accepted 25 October 2010
Objective To review the association between current enterovirus infection diagnosed with molecular testing and development of autoimmunity or type 1 diabetes.
Design Systematic review and meta-analysis of observational studies, analysed with random effects models.
Data sources PubMed (until May 2010) and Embase (until May 2010), no language restrictions, studies in humans only; reference lists of identified articles; and contact with authors.
Study eligibility criteria Cohort or case-control studies measuring enterovirus RNA or viral protein in blood, stool, or tissue of patients with pre-diabetes and diabetes, with adequate data to calculate an odds ratio and 95% confidence intervals.
Results The 24 papers and two abstracts (all case-control studies) that met the eligibility criteria included 4448 participants. Study design varied greatly, with a high level of statistical heterogeneity. The two separate outcomes were diabetes related autoimmunity or type 1 diabetes. Meta-analysis showed a significant association between enterovirus infection and type 1 diabetes related autoimmunity (odds ratio 3.7, 95% confidence interval 2.1 to 6.8; heterogeneity χ2/df=1.3) and clinical type 1 diabetes (9.8, 5.5 to 17.4; χ2/df=3.2).
Conclusions There is a clinically significant association between enterovirus infection, detected with molecular methods, and autoimmunity/type 1 diabetes. Larger prospective studies would be needed to establish a clear temporal relation between enterovirus infection and the development of autoimmunity and type 1 diabetes.
Type 1 diabetes is believed to result from a complex interplay between genetic predisposition, the immune system, and environmental factors.1 In recent decades there has been a rapid rise in the incidence of childhood type 1 diabetes worldwide, especially in those under the age of 5.2 3 4 5 6 In Europe, from 1989-2003 the average annual increase was 3.9%, too fast to be accounted for by genetics alone.4 Evidence in support of a putative role for viral infections in the development of type 1 diabetes comes from epidemiological studies that have shown a significant geographical variation in incidence, a seasonal pattern to disease presentation,2 3 7 8 and an increased incidence of diabetes after enterovirus epidemics.9
Enteroviruses are perhaps the most well studied environmental factor in relation to type 1 diabetes. A possible link was first reported by Gamble et al in 1969,10 with many subsequent studies, in humans and animal models of diabetes, showing an association, particularly with coxsackievirus B-4. Higher rates of enterovirus infection, defined by detection of enterovirus IgM or IgG, or both, viral RNA with reverse transcription polymerase chain reaction (RT PCR), and viral capsid protein, have been found in patients with diabetes at diagnosis compared with controls.11 12 13 14 15 16 17 Prospective studies have also shown more enterovirus infections in children who developed islet autoantibodies or subsequent diabetes, or both; as well as a temporal relation between infection and autoimmunity.13 18 19 20
The relation between enterovirus infection and diabetes is not consistent across all studies,21 22 23 24 however, and the subject remains controversial.25 Furthermore, in animal models viral infections might also protect from diabetes.25 A systematic review of coxsackie B virus serological studies did not show an association with type 1 diabetes,26 but to date there has been no systematic review of molecular studies. Based on the hypothesis that enterovirus infection increases the risk of pancreatic islet autoimmunity or type 1 diabetes, or both, we carried out a systematic review of controlled studies that used molecular virological methods to investigate the association between enteroviruses and type 1 diabetes.
Two reviewers (WGY and MEC) independently conducted a systematic search for controlled observational studies of enterovirus and type 1 diabetes mellitus. Databases searched were PubMed (from 1965 to May 2010) and Embase (from 1974 to May 2010). Search terms (exploded, all subheadings) used were: ‘diabetes mellitus’, ‘enterovirus’, ‘coxsackievirus’, ‘ECHOvirus’, ‘polymerase chain reaction’, ‘PCR’, ‘RNA’, ‘DNA’, ‘nucleic acid’, and ‘capsid protein’. The search was limited to studies in humans in any language and was supplemented by hand searching reference lists in the identified papers and by direct contact with authors.
Studies were eligible for inclusion if they were case-control or cohort studies (including those published as letters or abstracts); measured enterovirus RNA or viral capsid protein in blood, stool, or tissue of patients with pre-diabetes and diabetes; and provided adequate data to enable calculation of odds ratios and 95% confidence intervals. No restrictions were placed on the study population. We included only those studies that used molecular methods for viral detection (such as RT-PCR (reverse transcription-polymerase chain reaction), in situ hybridisation, or immunostaining for detection of viral capsid protein) to identify current or recent infection and because molecular testing is now standard for diagnosis of acute enterovirus infection.
The results of identified studies were classified into two groups, pre-diabetes and diabetes, depending on whether autoimmunity or type 1 diabetes was the outcome. There were four main categories of cases: autoantibody positive, newly diagnosed type 1 diabetes, established type 1 diabetes, and eventual type 1 diabetes. The latter three were combined into the diabetes group.
We calculated unadjusted odds ratios with 95% confidence intervals and P values for enterovirus identification in patients with pre-diabetes versus no diabetes and patients with diabetes versus no diabetes from the published figures using the Mantel-Haenszel method. The analysis was performed with both fixed and random effects models. Because of the presence of significant heterogeneity we have presented only the results from random effects models. Combined odds ratios were also calculated for different subgroups of studies according to study design. Statistical heterogeneity was explored with Cochrane’s Q test and the I2 statistic, which provides the relative amount of variance of the summary effect caused by heterogeneity between studies.
We assessed study quality using the Newcastle-Ottawa quality assessment scale (NOS) for case-control studies, as recommended by Cochrane collaboration.27 Three areas were evaluated—selection, comparability, and exposure—giving a possible total score 9, with 5 or more classed as good methods. In the comparability category, studies were assessed as to whether they controlled for age and sampling time as these are the two factors most likely to affect the incidence of enterovirus infection.
Our search returned a total of 114 publications and abstracts. After review of titles and abstracts, we identified and included 25 relevant papers—two letters and 23 articles. We also included data from two studies published as abstracts only. All were case-control studies (six were nested case-control studies that used samples collected prospectively20 22 28 29 30). One was excluded because it was a pilot study13 analysing the same data as a duplicate publication.20 Of the 26 remaining studies, eight contained more than one case group14 16 22 31 32 33 34 35 and these were analysed separately, giving a total of 34 studies. Of these, nine were studies of pre-diabetes (198 cases and 733 controls) and 25 were studies of diabetes (1733 cases and 1784 controls).
Characteristics of included studies
Thirty studies used RT-PCR or in situ hybridisation to detect enterovirus RNA, while four performed immunostaining for the enterovirus capsid protein vp1 on autopsy pancreas specimens (tables 1⇓ and 2⇓). Within the pre-diabetes group, all except two of the studies defined autoimmunity as positivity for at least one autoantibody associated with type 1 diabetes (table 1⇓). Study populations varied in age distribution. While most studies investigated children and adolescents (aged 16 and below), some included adults up to age 53.
Quality of evidence
The Newcastle-Ottawa scores ranged from 3 to 8, with 24 studies scoring 5 or more (table 3⇓), indicating reasonably good methodological quality overall, with no studies reporting a non-response rate.
Figure 1 presents the individual and summary odds ratio of the nine pre-diabetes studies ⇓. Odds ratios ranged from 0.1 to 483, with a summary odds ratio of 3.7 (95% confidence interval 2.1 to 6.8; P<0.001). There was some evidence for heterogeneity across the studies (χ2/df=1.34), but this value did not reach significance (P=0.22). When we analysed the results from the six nested case-control studies separately, the summary odds ratio was 3.0 (1.5 to 6.0; P=0.002) (table 4⇓).
Three of the nested case-control studies also separately examined the six or 12 month period preceding the first appearance of autoantibodies.20 22 The summary odds ratio was 3.6 (1.3 to 9.8; P=0.01). Five studies also sequenced the HLA haplotypes of their participants and two included those with low risk HLA genotypes. For those with high HLA risk haplotypes (five studies, 112 cases, 551 controls), the combined odds ratio was 3.5 (1.7 to 7.1; P<0.001).20 22 28 29 30 Only two studies (21 cases, 158 controls) included participants with low risk HLA genotypes, with conflicting results (0.4, (0.04 to 4.8)22 and 9.3 (1.9 to 45)30), but the combined odds ratio was not significant (2.3, 0.1 to 56; P=0.62).
Type 1 diabetes
Figure 2 shows the individual and summary odds ratios of the 25 studies of patients with type 1 diabetes ⇓. All studies except one32 showed an odds ratio over 1 for enterovirus positivity in patients with diabetes. Odds ratios ranged from 0.24 to 129, with a summary odds ratio of 10 (5.5 to 17; P<0.001). There was significant heterogeneity across the studies (χ2/df=3.21; P<0.001).
We carried out a subgroup analysis with respect to method of enterovirus detection (RNA or capsid protein) and case selection (newly diagnosed v established v eventual diabetes; table 5⇓, fig 2⇑). The combined odds ratios for newly diagnosed, established, and eventual diabetes were 13 (6 to 25), 11 (4 to 29), and 1.25 (0.2 to 7), respectively. The combined odds ratio of studies that used RNA detection was 8.8 (4.7 to 17; P<0.001), while for studies that performed immunostaining for enterovirus capsid protein, the odds ratio was 15 (7.5 to 31). There was no significant heterogeneity across studies that measured enteroviral vp1 protein, probably because of the similarity in study design.
We used sensitivity analyses to test the robustness of the results by country and study quality. For the 19 studies conducted in Europe,12 14 17 31 32 33 35 36 37 38 39 40 41 42 43 44 the combined odds ratio was 8.6 (4.3 to 17; P<0.001), with significant heterogeneity (χ2/df=3.75, P<0.001). The odds ratio was comparatively higher for the non-European studies (13.5, 7.1 to 26), with low heterogeneity, though there was considerable overlap of the confidence intervals between the two groups. When we excluded the studies with poor methodological quality (Newcastle-Ottawa score <5), the combined odds ratio was similar (8.9, 4.6 to 17; P<0.001). Subgroup analysis by HLA genotype was not performed because none of the studies performed HLA genotyping on all cases and controls.
This systematic review of 33 prevalence studies, involving 1931 cases and 2517 controls, shows a clinically significant association between enterovirus infection and islet autoimmunity or type 1 diabetes. The association between enterovirus infection, detected with molecular methods, and diabetes was strong, with almost 10 times the odds of enterovirus infection in children at diagnosis of type 1 diabetes compared with controls (9.8, 5.5 to 17.4), while the odds of infection was also higher in children with pre-diabetes than in controls (3.7, 2.1 to 6.8). There was some evidence for geographical differences; in non-European studies the odds ratio was 13.5 (7.1 to 25.8) compared with 8.6 (4.3 to 17.3) in European studies, though there was considerably overlap in the confidence intervals. While the findings from this meta-analysis of observational studies cannot prove that enterovirus infection has a causal role in pathogenesis of diabetes, the results provide additional support to the direct evidence of enterovirus infection in pancreatic tissue of individuals with type 1 diabetes.45 46
Strengths and weaknesses
We made every effort to reduce potential bias in this review, through use of pre-defined inclusion criteria, independent searches by two reviewers, no language restriction, and searching of references lists and conference proceedings. We included studies in children and adults, reducing the risk of bias resulting from high rates background infection in children.47 48 Studies from throughout the world were included, reducing the risk of geographical bias related to infection rates. Most studies, however, were from European countries, where the incidence of type 1 diabetes is higher. Given the heterogeneity of the study populations, we used random effects models, providing more conservative effect estimates.
Several limitations could have influenced our findings, including factors inherent in a meta-analysis of observational studies. There was significant heterogeneity in study design and methods used. Only 10 studies matched for three or more potential confounding factors (age, genetic risk, geographical location, and sampling time). Most of the included studies used children without diabetes or who were negative for antibodies as controls, but there could have been unmeasured factors influencing their risk of developing diabetes. Other environmental factors might modify the risk of type 1 diabetes, such as cows’ milk,49 vitamin D,50 and weight gain in infancy,51 but it is not possible to control for all of these potential confounders in case-control studies. Finally, enterovirus PCR primers had varying sensitivity and specificity, and not all studies reported the validation and limits of detection of their PCR method. Samples were obtained from various sites (serum, stool, throat swabs, etc) and because enteroviruses invade and replicate at mucosal surfaces, detection rates are likely to be higher in samples obtained from the gastrointestinal tract.52 53
The overall methodological quality of the studies of the studies was relatively good, with 26 publications scoring 5 or more of 9 on the Newcastle-Ottawa scale. Eleven studies included fewer than 50 participants, giving rise to the possibility of small study effects. The four largest studies of diabetes (involving more than 1000 cases and controls), however, showed a clear association between enterovirus infection and clinical diabetes.
Strengths and weaknesses in relation to other studies
A previous meta-analysis of coxsackie B virus serological studies found no significant association between type 1 diabetes and serology positivity,26 though summary estimates were not calculated because of significant heterogeneity between studies. Several major differences between the two meta-analyses could explain the discrepant findings. Firstly, most studies included in our review detected most enteroviruses by using PCR primers targeting the highly conserved 5′ untranslated region of the enterovirus genome, whereas serological studies examined only certain serotypes. Secondly, molecular methods for detection of enteroviruses are significantly more sensitive than serology.54 55 Thirdly, the detection of enterovirus RNA or vp1 identifies only current or recent infection. The latter is also a limitation of molecular methods, though this would probably cause bias towards under-reporting of infection rates and estimation of a lower than actual effect size. We could not examine whether participants had multiple enterovirus infections or the same persistent infection before the development of autoimmunity or type 1 diabetes.
Autoimmunity was mostly defined as a positive result for at least one autoantibody associated with type 1 diabetes, and the presence of a single antibody does not confer a high lifetime risk of clinical diabetes compared with positive results for multiple antibodies.56 57 Prospective studies are also limited by the frequency of sample collection, which might be only six or 12 months apart, and it is noteworthy that the only prospective study reporting an odds ratio under 1 had the longest sampling intervals.22 A temporal association between seroconversion to autoimmunity and infection could be under-reported because of lack of sampling at the time of infection or seroconversion, or both, in some individuals.
Maternal enterovirus infection might also be a risk factor for autoimmunity and type 1 diabetes. We did not specify maternal infection in our inclusion criteria, though among the “eventual diabetes” group enterovirus RNA was more commonly detected in dried blood spots from newborn infants who subsequently developed type 1 diabetes. Two of the included studies in the pre-diabetes group examined maternal enterovirus infection by using serology and showed little or no association between infection and subsequent development of autoimmunity in their offspring.20 28
There is conflicting evidence as to whether the presence or absence of high risk HLA genotypes modifies the association between enterovirus infection and type 1 diabetes. Several groups have reported higher rates of enterovirus infection in children with low risk HLA genotypes.16 58 Unfortunately, we could not do a subgroup analysis by HLA genotype in the diabetes studies because most studies did not do HLA genotyping in control participants. In the pre-diabetes group, the odds ratio of enterovirus infection in high risk HLA participants (3.5) was not different from the overall odds ratio (3.4), and the conflicting results from the two studies with low risk participants do not support an association between enterovirus infection and autoimmunity. Ideally, future studies should include individuals with low risk HLA genotypes to explore whether genetic risk modifies the effect of enterovirus infection on the risk of type 1 diabetes.
Our results show an association between type 1 diabetes and enterovirus infection, with a more than nine times the risk of infection in cases of diabetes and three times the risk in children with autoimmunity. The odds of having an enterovirus infection in people with established diabetes (odds ratio 11) suggest that persistent enterovirus infection is also common among patients with type 1 diabetes. While it is not possible to determine a causal relation between infection and type 1 diabetes with a randomised controlled trial, larger multicentre international prospective studies could examine interactions between type 1 diabetes and various environmental, geographical, and genetic factors.
What is already known on this topic
Observational studies have shown an association between enteroviruses and type 1 diabetes
A systematic review of serological studies only found no association
What this study adds
A review of molecular studies showed an association between current enterovirus infection and type 1 diabetes
Cite this as: BMJ 2011;342:d35
Contributors: WGY performed the literature search, extracted, analysed, and interpreted the data, and drafted the article. MEC conceived and designed the review, helped extract and interpret the data, and revised the article critically for intellectual content. WDR revised the article critically for intellectual content and approved the final draft for publication. WDR is guarantor.
Funding: This research received no specific grant from any funding agency in the public, commercial, or not-for-profit sectors.
Competing interests: All authors have completed the Unified Competing Interest form at www.icmje.org/coi_disclosure.pdf (available on request from the corresponding author) and declare: no support from any organisation for the submitted work; no financial relationships with any organisations that might have an interest in the submitted work in the previous 3 years; no other relationships or activities that could appear to have influenced the submitted work.
Ethical approval: Not required.
Data sharing: No additional data available.
This is an open-access article distributed under the terms of the Creative Commons Attribution Non-commercial License, which permits use, distribution, and reproduction in any medium, provided the original work is properly cited, the use is non commercial and is otherwise in compliance with the license. See: http://creativecommons.org/licenses/by-nc/2.0/ and http://creativecommons.org/licenses/by-nc/2.0/legalcode.