Streptococcal infection, Tourette syndrome, and OCD

Streptococcal infection (SI) can induce autoimmune neuropsychiatric disorders, the classic disorder being Sydenham chorea. In a number of small clinic-based studies, it has recently been shown that the phenotype of poststreptococcal neuropsychiatric disorders may be wider than just Sydenham chorea, perhaps also including tic disorders, obsessive-compulsive disorder (OCD), and other neuropsychiatric disorders with onset in childhood.^1–5 It has been proposed that these neuropsychiatric disorders develop following SI by the process of molecular mimicry, whereby antibodies directed against bacterial antigens crossreact with brain targets. Anti–basal ganglia antibodies have been implicated in this process, though this is not a universal observation.^6,7 A direct causal relationship with SI has therefore been postulated. Tic disorders and associated neuropsychiatric behavioral disorders are common and typically start in childhood. However, it is unknown how commonly Tourette syndrome (TS) or OCD occurs after SI, and whether these play an important role in the pathogenesis of these disorders at the population level. Careful epidemiologic studies are needed to assess the association between SI and these disorders.^8–10 To date, only one epidemiologic study has been undertaken. This case-control study, in children aged 4–13 years, found that patients receiving their first diagnosis of TS, OCD, or tic disorder were over twice as likely (odds ratio [OR] 2.2) to have had prior SI in the 3 months before onset.¹¹ The risk was higher in those with multiple SIs within 12 months (OR 3.1). However, this study had a wide confidence interval (CI), with the lower estimate being consistent with merely a 5% relative increased odds. We therefore examined the association between throat infection among children and young adults and the onset of TS/OCD using a large primary care database. We hypothesized that throat infection among children and young adults is associated with the onset of TS/OCD.

METHODS

RESULTS

We identified 255 cases, 129 (51%) with a diagnosis of OCD, 108 (42%) of TS, and 18 (7%) of tics (table 1). As the number of cases with tics was small, we combined them with the TS group, resulting in a combined group of 126 cases (referred to as TS/tics for simplicity). These were matched to 4,519 controls (2,211 for OCD cases, 2,308 for TS/tics cases; table 2).

Table 1 presents the sociodemographic characteristics of cases and controls. As expected, cases and controls were comparable in terms of age and sex, 2 of the matching variables. There was a small majority of males among the cases of OCD, and an expected, larger majority amongst cases of TS/tics. The OCD cases were older, with a median age at onset of 15.8 years, compared to a median age of 9.3 years in the TS/tics cases. There was no evidence that cases with either condition were more likely to come from more affluent areas or urban areas. There was some evidence that cases of TS/tics were more likely to live in areas where a higher proportion of the population was white (p for trend = 0.05; table 1). A similar association for cases of OCD was apparent, but supported by weaker evidence (table 1).

Twenty (15.5%) cases of OCD had been exposed to a possible SI in the 2 years prior to diagnosis (table 2). There was a very similar rate of infection among the controls, and consequently no evidence of an association between OCD and SI (p = 0.69). There were very few instances of repeated consultation for a possible SI within 3 weeks of the first, with no apparent difference between cases of OCD and controls in exposure to these potentially more severe infection episodes. The expected association between possible SIs treated with antibiotics and OCD was not observed. Rather than antibiotic prescription distinguishing those exposure episodes where the treating GP was more confident of a bacterial cause, the data were more consistent with the antibiotic preventing complications of infection. While numbers were small, OCD cases were more likely to have had a possible SI without antibiotic treatment compared to controls (p = 0.02). Repeating these analyses with exposure to possible SIs within 5 years prior to the diagnosis of OCD being considered gave very similar results (table 2).

Thirteen (10.3%) of the cases of TS/tics had been exposed to possible SI in the 2 years preceding diagnosis, with no evidence of a higher rate of infection in theses cases compared to controls (p = 0.15; table 3). None of the TS/tics cases had been exposed to a more serious infection, as indicated by a repeated consultation within 3 weeks of the first. No associations became apparent once the infections treated with antibiotics were separated from those that were not. Considering possible SIs within 5 years prior to the diagnosis of TS/tics also failed to support a positive association, though the upper CI was consistent with a modest increased risk. Analysis restricted to TS alone (rather than TS and tics combined) gave similar results. The sensitivity analysis using the restricted range of infections yielded similar results (crude results for SI within 2 years for OCD were OR = 3.62, 95% CI 0.77, 17.0, p = 0.10, and within 5 years OR = 3.14, 95% CI 0.90, 11.0, p = 0.07). There were not enough data to calculate ORs for TS/tics using these time windows but with any lifetime exposure the OR was 0.92 (95% CI 0.19, 4.40, p = 0.92).

DISCUSSION

This study could not confirm the association of neuropsychiatric disorders with SIs in a large community-based sample of children and young adults between 2 and 25 years. Cases with TS, tics, or OCD were no more likely to have had possible SIs, as diagnosed clinically. The only association found was that cases with OCD were more likely to have had possible SIs not treated with antibiotics in the 2 years prior to diagnosis than controls. However, given the number of statistical comparisons and the opposite direction of this finding to prediction, this finding must be treated with caution as it may reflect a chance type I error.

In the previously reported population-based results from the United States,¹¹ cases were more than twice as likely than controls to have had a SI preceding OCD, TS, or tics. The results of the current study are based on a larger sample with 255 cases and almost 20 times as many controls, and hence had greater statistical power. The data on exposure were collected prospectively on a completely unselected and comprehensive sample of the general population, representative of the UK population (http://www.epic-uk.org),¹² and hence recall and selection bias could not operate. However, there are differences in study design that may have contributed to different outcomes in the 2 studies: the smaller population studied in the Mell et al.¹¹ article may have contributed to a type I error or, alternatively, methodologic issues in this study (see below) may have led to a type II error in this study. In particular, due to the low number of laboratory-confirmed cases, this study relied on clinical diagnosis as opposed to laboratory confirmation of SI and neuropsychiatric diagnoses relied on GP records. The crude incidence rates observed in our population were 4.0 for OCD and 3.6 for TS/tics per 100,000 person-years. This compares similarly to the rates from the Mell et al.¹¹ study, where 33 and 47 cases of OCD and TS were identified over an 8-year period from a population of around 75,000 children aged between 4 and 13 years (rates 5.5 and 7.8 per 100,000 person-years).¹¹ A previous birth cohort analysis from Denmark also ascertained 252 and 95 cases of OCD and TS from 127,782 children who were born either in 1990 or 1991 and followed up to December 2004. Assuming little loss to follow-up or censoring, this would result in 1,788,948 person-years of observation (over 14 years of follow-up) and an average annual incidence rate of 14.1 and 5.3 per 100,000 person-years for OCD and TS.¹³

Our analysis used wider time windows of exposure (2 years and 5 years) from the previous study (3 months and 1 year). We choose these longer gaps because of the inevitable delays among symptom onset, recognition of symptoms, health care seeking behavior, and general practitioner or specialist diagnosis.

We were also able to adjust for potential sociodemographic factors which may confound the risk of SIs and neuropsychiatric disorder or its clinical diagnosis. The diagnosis of TS or tics was more frequent in areas with a large white population, suggesting either that the incidence of these disorders is lower in ethnic minority groups or that children from ethnic minorities are less likely to present to their doctor or be referred and diagnosed than white children, though our variable was ecologic rather than based on an individual measure of ethnicity. It is concordant with previous observations that white individuals have higher rates of TS as opposed to African black children and possibly children from the Far East.^14,15

There are several methodologic factors that need to be considered before accepting our negative findings as conclusive evidence of no causal association. 1) The outcome was based on GP records and hence the validity of diagnosis could not be confirmed. Most general practitioners are unlikely to make these diagnoses and would refer patients for a specialist diagnosis. A previous study using the same method for a diagnosis of autism found that 93% of diagnoses were confirmed.¹⁶ However, it is well-recognized from prevalence studies that screen-detected rates are far higher than those based on existing clinical diagnosis.^17–24 We are likely to have missed milder cases of disease, though if anything their inclusion would have further attenuated the results if we assume that the association with prior SI is stronger with more florid or severe cases. 2) The onset of TS, tics, or OCD relied on the retrospective date of onset. 3) The onset of TS, tics, and OCD is usually insidious and it is therefore difficult to establish the correct time of onset, and the relatively late age at onset suggests an earlier than reported age at onset. In both cases, our sensitivity analyses with wider time windows still failed to find any associations. 4) Diagnosis of SI was based on clinical impression, which is known to be have low accuracy (sensitivity between 39% and 87%),^25,26 although use of guidelines improves this,²⁷ rather than streptococcal cultures or serology, which appeared to have been undertaken or recorded for a tiny minority of infection episodes. We used a fairly wide range of diagnostic codes, which would increase sensitivity, but at the expense of specificity, and hence it is possible that nondifferential misclassification would have attenuated any true association (false negative). We have looked at this in 2 different ways. The incidence of possible SIs in controls in this study (15%–16% over 2 years) was actually less than that reported in the Mell et al.¹¹ study (11%–16% over 1 year), which would be surprising if our false positive rate was higher than theirs. However, given differences in population demography, true population risk of infection, and different cultural and financial barriers in seeking a general practitioner diagnosis between the United Kingdom and United States, it is difficult to know how comparable these results are. Secondly, our repeated analysis using a much smaller and more specific set of codes where SI was explicitly mentioned still failed to find any association.

Only a prospective study with high diagnostic accuracy and laboratory confirmation of SI will be able to overcome these problems, but such a study would require a very large sample size and may be prohibitively expensive. However, a recent prospective study²⁸ examining the relationship between laboratory- confirmed β-hemolytic SIs and exacerbations of childhood tics and obsessive-compulsive symptoms in patients with pediatric autoimmune neuropsychiatric disorders associated with streptococcal infections (PANDAS) and with chronic tic disorders or OCD found that only a minority of exacerbations in the PANDAS group and none in the other group were associated with these infections. While these cases were already diagnosed and an association with SI and onset of symptoms could therefore not be examined, the results argue against a strong causal relationship between group A SIs and clinical symptoms of tics or OCD at least in the majority of cases. As there is at present insufficient supportive evidence for a causal relationship between PANDAS and group A SIs, the American Heart Association²⁹ does not currently recommend routine laboratory testing for group A SI to diagnose, long-term antistreptococcal prophylaxis to prevent, or immunoregulatory therapy to treat exacerbations of this disorder.

ACKNOWLEDGMENT

The authors thank Mary Thompson and the staff of CSD EPIC for their support and Dr. Hjördis Osk Atladottir for help with the interpretation of the Danish register study.

DISCLOSURE

Dr. Schrag serves on scientific advisory boards for Osmotica Pharmaceutical Corp. and Boehringer Ingelheim; received funding for travel and speaker honoraria from Boehringer Ingelheim; serves on the editorial board of Movement Disorders; and receives research support from the Parkinson's Disease Society UK and Amgen. R. Gilbert and Dr. Giovannoni report no disclosures. Dr. Robertson received royalties from publishing Psychiatry at a Glance (Wiley-Blackwell, 2008), Why Do You Do That (Jessica Kingsley Publishers, 2006), Tourette Syndrome: The Facts (Oxford University Press, 2008), and Tourette Syndrome for Teachers, Parents and Carers (David Fulton Publishers, 2000). Dr. Metcalfe receives honoraria and funding for travel from Syngenta AG as a member of an independent data monitoring committee; serves on the editorial board of Statistical Methods in Medical Research; and receives research support from the UK National Health Service Screening Programme (PI), the UK Home Office (co-I), the UK Medical Research Council (theme lead for ConDuCT trials methodology hub), the UK National Health Service Research for Patient Benefit [(PB-PG-0807-13387 (co-I) and PB-PG-0906-11179 (co-I)], Medicines and Healthcare products Regulatory Agency [(SDS-003 (co-I)], the National School for Primary Care Research, Department of Health (co-I), the Tourette Syndrome Association (co-I), Bristol Research into Alzheimer's Care of the Elderly (BRACE) (supervisor of funded studentship), Cancer Research UK [C18281/A8145 (co-I); C11046/A10052 (PI)], and the World Cancer Research Fund [2006/15 (co-I)]. Dr. Ben-Shlomo received royalties from publishing A Life Course Approach to Chronic Disease Epidemiology (2nd edition) ( Oxford University Press, 2004); and receives research support from the Cancer Research UK [ C18281/A11326 (coapplicant)], the Tourette Syndrome Association USA (coapplicant), British Heart Foundation (coapplicant), Medical Research Council, Wellcome Trust (coapplicant), Economic and Social Research Council (coapplicant), and BRACE charity (PI).