Does Group A β-Hemolytic Streptococcal Infection Increase Risk for Behavioral and Neuropsychiatric Symptoms in Children? FREE

Pediatric autoimmune neuropsychiatric disorders associated with streptococcal infection, or PANDAS, is hypothesized to describe a subset of children with sudden onset obsessive-compulsive disorder (OCD) and tic disorders that occur in the weeks following a group A β-hemolytic streptococcal infection ("strep throat" or GAS).¹ Swedo et al¹ at the National Institute of Mental Health, Bethesda, Md, identified 5 diagnostic criteria for PANDAS in 1998: (1) presence of OCD and/or tics; (2) pediatric onset of symptoms (from age 3 until puberty); (3) abrupt onset and/or episodic course of symptom severity; (4) exacerbations associated with GAS infection (confirmed by positive culture and/or elevated antistreptococcal titers); and (5) associated neurological abnormalities (eg, choreiform movements). The children in the PANDAS subgroup of children with childhood-onset OCD and tic disorders also have been documented to have a high frequency of comorbid neuropsychiatric symptoms. Increased frequencies of comorbid attention-deficit/hyperactivity disorder, oppositional defiant disorder, major depression, separation anxiety disorder, overanxious disorder, and enuresis have been reported.¹

Epidemiologic tenets for causality between a risk factor and disease suggest a link between GAS and PANDAS.² Pediatric autoimmune neuropsychiatric disorders associated with streptococcal infection is biologically plausible and has an appropriate analogy in another poststreptococcal illness. It resembles in some ways the well-described component of acute rheumatic fever, Sydenham chorea.³ There is consistency of reports in the PANDAS literature.^4- 7 Several case-control studies report the following relevant relationships: between antistreptococcal titers and tics^8- 9; antineuronal antibodies and movement disorders^9- 10; antistreptococcal antibody titers and basal ganglia volumes and OCD, chronic tic disorder, and attention-deficit/hyperactivity disorder¹¹; antistreptococcal antibody titers and Tourette syndrome¹⁰; and caudate, putamen, and globus pallidum size and streptococcus-associated OCD and/or tics.¹² Some studies explored the tenet of reversibility in that they found that experimental treatments of immunomodulatory agents such as intravenous immunoglobulin and plasma exchange yielded benefits.¹³ One randomized controlled trial did not demonstrate reversibility of PANDAS symptoms with penicillin prophylaxis, though this was not surprising since the penicillin failed to prevent GAS.¹⁴ More recently, Murphy and Pichichero¹⁵ reported from their observational prospective analysis of 12 children with PANDAS that PANDAS-related symptoms declined following antibiotic treatment.

The entity of PANDAS, however, remains controversial. Some postulate that the relationship between GAS infection, an extremely common illness in children, and neuropsychiatric symptoms may merely be an "epiphenomenon."^16- 18 Skeptics have pointed to limitations of some previous PANDAS research, including lack of prospective and controlled studies¹⁷; small numbers of study subjects¹⁹; and failure to identify whether subjects were preselected for having had "recent, recurrent, or chronic" GAS tonsillopharyngitis.¹¹ Furthermore, most PANDAS studies focus on patients with severe neuropsychiatric illness, and no studies have included comparison groups derived from outpatient pediatric populations.

To date, no large prospective cohort study of children with and without GAS exposure has been performed to determine the risk of GAS exposure on the development of PANDAS or its milder symptoms. Therefore, we sought to conduct a prospective cohort study following 2 groups of children (those with and without GAS infection) for 12 weeks in an outpatient pediatrics practice to determine if children with GAS infection are more likely than children without GAS (those with presumed viral infection and well children) to develop symptoms that have been reported to occur in the typical PANDAS subgroup of children with OCD and tics and their comorbid conditions. We hypothesized that although PANDAS that meets all criteria¹ may be rare, milder variants of the disorder with more transient symptoms might be captured in a large, prospective outpatient study.

We used a prospective cohort study design and followed 2 distinct cohorts of children for 12 weeks: children with GAS infection and children without GAS infection. While cohort studies are not traditionally used to investigate the risks of rare disorders such as PANDAS, we hypothesized that a "mild PANDAS variant" would be more common and would be identifiable in a cohort study. We defined a mild PANDAS variant a priori as the development of 2 or more new PANDAS-related symptoms during 12 weeks.

We recruited 4- to 11-year-old English-speaking children and their parents from the Elmwood Pediatric Group, a large suburban practice in Rochester, NY, during 1 GAS season (October 2001 through June 2002). While PANDAS criteria indicate children must be at least age 3 years, we believed that by beginning enrollment at age 4 years, we might more likely have children who could reliably report psychological symptoms. The upper limit of age 11 years was chosen because in the clinical description of the first 50 cases, the upper age limit for the onset of PANDAS was 11 years for girls and 12 years for boys.¹ Written informed consent was obtained from parents and written assent from the children. Because we were interested in incident symptoms, children were excluded from participation if they had any previously diagnosed psychiatric or neurological problem, including attention-deficit/hyperactivity disorder, autism, depression, OCD, pervasive developmental delay, mental retardation, and Tourette or other tic disorders. Children were also excluded from participation if they were taking any psychotropic medications (including medication for attention-deficit/hyperactivity disorder or selective serotonin reuptake inhibitors) or had any history of closed head injury with loss of consciousness or seizures.

Children enrolled in the GAS-infected cohort had a throat culture positive for GAS and symptoms of GAS infection including a sore throat and other symptoms such as fever or headache. The GAS-uninfected cohort of children was divided, roughly in half, with patients who were seen with illness symptoms (ie, fever, sore throat, or headache) and presumed viral infection and with patients who were seen for routine well-child care.

Of the 814 children who received throat cultures at the outset of the study, 411 had GAS infection and 403 did not have GAS infection. Among the 411 who had GAS infection, 399 had come to the practice with symptoms of illness. Because the practice is open 7 days per week for sick care and because patients who had positive results by initial rapid streptococcal testing (later confirmed by culture) were given a prescription at the time of initial examination, there were unlikely to be significant delays in onset of treatment. Symptomatic patients with GAS infection were treated according to the practice's previously established consensus on treatment of GAS. Children were treated with a twice-daily, weight-based regimen of amoxicillin. Penicillin-allergic patients, those with recurrent streptococcal illness, or those with demonstrated previous amoxicillin resistance were treated with a cephalosporin antibiotic.

Twelve (<2%) of the patients seen for well-child care who had throat cultures because of study protocol were found to be GAS positive but were symptom free. They were believed to be carriers and excluded prior to data analysis because carriers usually do not mount an immunological response to infection. The other 403 participants were GAS negative. This group consisted of 196 children seen for well-child care and 207 children with nonstreptococcal acute respiratory illnesses, which included sore throat. We refer to this latter group as the "presumed viral group" (Figure 1). Children without GAS infection had not had previous GAS infection during that streptococcal season (October 2001 to June 2002).

All parents and children were encouraged to speak with their pediatricians about symptoms of concern rather than to assume problems would be brought to attention as a result of the study. All procedures were compliant with and the study was approved by the University of North Carolina at Chapel Hill School of Medicine institutional review board.

Because no standardized instruments to assess PANDAS symptoms exist, we developed 2 instruments for our outcome measures. The first instrument was a questionnaire administered to parents that asked about the recent onset of 20 unusual behaviors reported frequently by children in the PANDAS subgroup. These behaviors included the 2 primary neuropsychiatric findings in this subgroup of children with childhood-onset OCD and tic disorders—namely, OCD symptoms (such as checking, hoarding, washing, fixation with order, or unusual religiosity and were adapted from the Children's Yale-Brown Obsessive Compulsive Scale²⁰) and tics. They also encompassed the associated comorbid neuropsychiatric symptoms of attention-related problems; separation anxiety and "clinginess"; more generalized worries and fears; negativity; obsessive toilet paper wiping and unusual bathroom routines; and unusual urinary problems such as urgency and enuresis. The symptoms of toileting and urinary problems were included because new-onset enuresis or daytime urinary frequency were found in 6 (12%) of 50 of the initial PANDAS sample described by Swedo et al,¹ and compulsive daytime urinary urgency, frequency, and wiping were noted in 7 (58%) of 12 of the PANDAS sample evaluated by Murphy and Pichichero.¹⁵

Following these questions, parents were asked: "Which of the following statements represents your feelings about your child's recent behaviors? (a) Recent behaviors have not been a problem; (b) Recent behaviors have been a little bit of a problem; (c) Recent behaviors have been a moderate problem; and, (d) Recent behaviors have been a severe problem and have meant that we have had to change our routine at school or home to compensate." They were also asked: "Which of the following statements is true? (a) My child and/or our family has had a significant life event recently; or (b) My child and/or our family has NOT had a significant life event recently." For choice "a," parents were given a series of examples of significant life events such as death of a family member or a recent move. Finally, parents answered demographic questions about their children, themselves, and their partners (age of child, sex of child, ethnicity, sex of responding parent, education of responding parent, education of spouse/significant other, marital status).

Children who were able (n = 682) privately completed a 10-item written questionnaire about internalizing or anxious thoughts and obsessive-compulsive symptoms that might not be known by parents.^20- 23 For children who were not able to read, a study nurse read questions to them and they verbally answered "yes" or "no." These replies were then marked on a separate form. Study nurses also demonstrated to parents how to administer this questionnaire in a private manner for follow-up surveys so that parents would not be aware of children's responses. Children who were not able to read the questionnaire themselves or understand directions for completing the nonliteracy-dependent version (as assessed by our study nurses) were not given questionnaires (n = 122), though their parents were still invited to participate.

To assess development of new symptoms as they emerged 2 and 12 weeks after enrollment, parents and children were sent surveys identical to the baseline instruments. If their responses to the follow-up mailings were not received within 2 weeks, another mailing was sent and study nurses telephoned to remind parents to respond. Questionnaires were considered in the analysis if they were complete or nearly complete (no more than 2 of 32 questions missing answers). A random 10% of the sample was rechecked for data entry reliability.

We assessed 4 main outcome measures. First, we examined the prevalence of individual parent- or child-endorsed symptoms at baseline (time of enrollment) in the GAS-infected and presumed viral groups vs the well-child group. We hypothesized that a higher percentage of children with GAS infection and their parents would endorse individual symptoms than those with presumed viral infections or well children. We used Fisher exact tests for categorical variables. For these symptoms' comparisons, we used an α level of .01 because of the use of multiple comparisons. We also calculated the prevalence odds ratios (ORs) with exact 95% confidence intervals (CIs) to determine if either children with GAS infection or children with presumed viral infection had an increased prevalence of symptoms when compared with the referent of well children.

Second, we assessed the incidence of individual symptoms endorsed by either parents or children across time in the 3 comparison groups. χ² tests for categorical variables were determined using α = .01. Incidence ORs were then calculated with exact 95% CIs to determine if either children with GAS infection or children with presumed viral infection had a greater incidence of symptoms than well children.

We also considered a mild PANDAS variant, defined a priori as 2 or more changed PANDAS symptoms (not endorsed at baseline but endorsed 2 or 12 weeks later) on parent or child questionnaires. We postulated that the new onset of 2 or more symptoms from one point to the next would be more clinically meaningful than a single symptom and yet prevalent enough by sample size calculations to be able to detect a difference by GAS exposure in our sample size. We determined a necessary sample size of 275 based on a power of 90%, α = .05, and an estimated 10% difference in proportions.

We calculated unadjusted and adjusted relative risks for the development of a mild PANDAS variant by exposure to GAS at 2 weeks, 12 weeks, and overall (2 or 12 weeks). We used generalized linear models with binomial distribution and log link (binomial or multiplicative risk regression) to determine the relative risk of developing a mild PANDAS variant by illness type (children with GAS infection and children with presumed viral infection were compared with well children). Adjusted models accounted for potential confounders of the relationship. The potential confounding variables were identified a priori and included the child's age (≥7 vs <7 years), sex, and race (white vs nonwhite), as well as the responding parent's sex, educational attainment (<college degree vs ≥college degree), and endorsement of a recent significant life event (yes vs no).

The final outcome measure was a dichotomous variable representing whether behavior worsened across time as reported by parents on a behavior scale across the 3 time points (from 1, "not a problem" to 4, "a severe problem"). The relative risk for developing more problematic behavior from time point 1 (baseline) to time point 2 (2 weeks following baseline) or time point 3 (12 weeks following baseline) by illness type (those with GAS infection, those with presumed viral syndrome, and those who were well) was calculated. We used a generalized linear model with binomial distribution and log link (binomial regression or multiplicative risk regression) to determine the relative risk relationship between worsening of behavior and GAS exposure status, adjusted for the potential confounders listed earlier, in 3 possible combinations: (1) "transiently" worse (behavior worsened from time point 1 to time point 2 but returned to baseline by time point 3); (2) "lastingly" worse (behavior at time point 3 worse than behavior at time point 1); and (3) "ever" worse (combines those whose behavior at any 1 of the time points was worse than at the other time points).

Most study participants in this upstate, suburban practice were non-Hispanic white and their parents were highly educated; only 24% had less than a college degree, and 33% completed some graduate education (Table 1). The GAS-infected, presumed viral, and well-child cohorts differed by percentage of children younger than or equal to age 7 years and percentage of parental respondents who were women. Mean age at enrollment, child's sex, educational level of the responding parent, educational level of the responding parent's significant other or spouse, and marital status of the responding parent were similar between cohorts.

At 2 weeks, 676 (84%) of 803 enrollment participant parents returned complete questionnaires. At 12 weeks, 625 (78%) returned complete questionnaires. Child response rates were as follows: at 2 weeks, 548 (80%) of 682 (able to complete questionnaires) returned complete questionnaires, and at 12 weeks, 508 (74%) of 682 returned complete questionnaires.

At baseline, parents of ill children vs parents of well children reported the recent onset of several PANDAS symptoms more frequently, including harder time paying attention, increased fidgetiness/restlessness, unusual clinginess/harder time with separation, unusual noises such as throat clearing or barking, trying to stop a behavior the child always repeats, unusual urinary problems, and seeming more negative than usual (Table 2). However, none of these symptoms was significantly more commonly reported for those with GAS exposure than for those with presumed viral infection. Being more sick with worry and having upsetting, intruding thoughts were more commonly reported for those with viral syndrome compared with well children but not for those with GAS exposure (Table 2). Considering symptoms reported by the children, no PANDAS symptoms were more commonly reported by children with GAS infection or with presumed viral infections when compared with the well cohort. The obsessive-compulsive symptom of needing to dress in a certain way each day was significantly less frequently endorsed by children with GAS infection (Table 3).

We also examined change in symptom endorsement across time by illness status. No differences in the onset of incident symptoms were reported among the GAS, viral, and well cohorts from enrollment to 2 or 12 weeks at the P<.01 significance level for either parents or children (Table 4 and Table 5). However, parents of children with GAS infection were 4 times as likely to report that their children were "more sick with worry" than parents of well children (OR, 4.2 [95% CI, 1.1-undefined]), and parents of children with presumed viral infection were nearly 9 times as likely to report unusual fears than parents of children who were well (OR, 8.7 [95% CI, 1.4-undefined]).

Children with GAS infection were not at increased risk for development of a mild PANDAS variant based on parental or child report when compared with well children across 2 or 12 weeks or overall (Table 6 and Table 7).

Finally, parents of the GAS-infected cohort did not report significantly more transient, lasting, or ever worse behavior than parents of children without GAS infection in multivariate analyses. The adjusted relative risk for reporting behavior that worsened transiently was 1.76 (95% CI, 0.47-6.51) for children with GAS infection and 2.00 (95% CI, 0.50-8.02) for children with presumed viral infection compared with well children. For reporting behavior that worsened and persisted throughout time, the relative risk was 1.67 (95% CI, 0.71-3.93) for children with GAS infection and 0.92 (95% CI, 0.33-2.55) for children with presumed viral infection compared with well children. Similar findings were observed for reporting behavior that was worse at any point in the study (adjusted relative risk for children with GAS infection, −1.16 [95% CI, 0.76-1.78]); with presumed viral infection, 0.99 (95% CI, 0.62-1.63); referent, well children).

More parents of ill children who were both GAS positive and GAS negative reported inattentiveness, fidgetiness, clinginess, and unusual noises (barking, throat clearing) among their children at baseline. These findings are consistent with reports noting associations between illness (even in nonhospitalized children) and behavioral problems and tic exacerbations.^24- 27 However, from this prospective cohort study of 802 patients in an outpatient pediatrics practice, using parent- and child-reported data, children infected with and treated for GAS do not appear to be at increased risk over children with presumed viral syndrome for development of PANDAS symptoms or a mild PANDAS variant, defined in our study as the new onset (<2 weeks) of 2 or more symptoms that characterize PANDAS. Because all participants with GAS infection were treated with appropriate antibiotics, intervention may have prevented or altered the natural history of mild PANDAS symptoms as at least 1 report suggests¹⁵ and as is true for the analogous rheumatic fever.

This is the largest prospective study of PANDAS symptoms to date. Nevertheless, sample size and ascertainment bias may have played a role in our negative results. The prospective study by Murphy and Pichichero¹⁵ identified only 12 children manifesting PANDAS among approximately 4000 patients infected with GAS across a 3-year period in the same practice, suggesting the rarity of the entity in its classical manifestation. While based on these numbers we would only expect to detect 2 to 3 patients with PANDAS by classic, explosive manifestation in this prospective study of 800 patients, we expected to identify individual incident symptoms and subclinical manifestation with more frequent incidence. We were cautious throughout to label our findings of greater than 2 new-onset symptoms by parents or children a mild PANDAS variant. Because of our limited sample size and symptom measures, our study cannot provide relative risks or incidence figures for PANDAS itself in this population. Our study was powered to detect a 10% difference in the proportions of number of symptoms endorsed by GAS exposure (with 90% power and a 2-tailed α of .05). If there was a difference in proportions of even subclinical manifestations below this level, our study might not have detected that.

We relied on parent and child subjective responses to serve as a proxy for symptom development, but this approach has limitations. Previous studies, for example, have demonstrated that parents may underreport tics in children when compared with direct observation,²⁸ and children may hide obsessive-compulsive symptoms from their parents too.²¹ Our child queries revealed high rates of symptom endorsement. These rates were unexpected given the known rates of obsessive-compulsive symptoms and may have obscured between-group differences. The questionnaire may have been worded in such a way that the "symptoms" endorsed by parents or children did not reflect problematic behaviors. In fact, the lack of impairment related to these symptoms suggests that the frequency of endorsement on this questionnaire may have been a reflection of normal childhood concerns rather than symptoms of psychopathology. To determine the validity of our questionnaire will require further research.

Because some patients with sore throat and a positive GAS culture may have had viruses and merely be carriers of GAS, there may have been misclassification bias. This would have biased our results to the null hypothesis. In some patients with pharyngitis, a culture positive for GAS represents a carrier state in which the oropharynx is colonized with GAS. There is not a true GAS infection since there is not an immune response. In those instances, the positive culture is an incidental finding and not a reflection of a GAS pharyngitis. Since there is not an immune response to the GAS colonization, the carrier cannot have poststreptococcal autoimmune sequelae. Twelve children were considered GAS carriers in the present study and were not included in the original analyses to avoid misclassification bias. When the 12 carriers were included in a reanalysis of the data, the results did not differ.

Of more importance, the present study did not capture children with asymptomatic GAS infections. These children were not included in the GAS-positive group as they were not seen at the clinic for evaluation of pharyngitis; if they happened to be there for a well-child visit, they would have been excluded from the healthy control group because of their positive GAS culture. The literature has shown that asymptomatic GAS infections are a frequent cause of post-GAS sequelae. For example, the recent rheumatic fever epidemics in Utah demonstrated that 83% of children with rheumatic carditis did not have a history of sore throat or other symptoms of GAS pharyngitis.²⁹ Asymptomatic GAS infections also appear to play an important role in triggering neuropsychiatric symptoms, such as OCD and tics. Murphy et al³⁰ found that the majority of the children meeting criteria for the PANDAS subgroup were initially seen for "behavior conferences" and did not have complaints of sore throat or fever, although their throat cultures were positive for GAS and they demonstrated rising antibody titers. Future research studies should be designed to capture children with asymptomatic infections, as this may be the group with the highest yield for GAS-related neuropsychiatric symptoms.

The study occurred during only 1 GAS season. It is possible that, just as there are rheumatogenic GAS strains, there may also be "PANDAS-genic" GAS strains. Looking at only 1 year of streptococcal data might have missed such strain-specific occurrences.³¹ Since we asked parents at baseline for endorsement of symptoms with recent onset and then used strict epidemiologic methods to base relative risks only on change from baseline to 2 or 12 weeks, our study design might have underestimated the relative risk of GAS-related acute behavioral/neuropsychiatric problems in our population. This would be the case if symptoms were present at baseline but only had been present for a brief period. We have no data about child or parental perceptions of behavior or neuropsychiatric symptoms prior to enrollment, and it is possible that we started our study effectively too late. However, we did exclude patients with known psychiatric disease at enrollment.

In conclusion, this is the first prospective cohort study to examine the risk of behavioral symptoms associated with GAS infections. We found no evidence to support the notion that symptomatic, treated GAS tonsillopharyngitis confers significant risk toward the development of most neuropsychiatric symptoms or mild PANDAS variants. While it is possible that there is no relationship between GAS and OCD and tic disorders, other explanations for our negative findings are that mild PANDAS variants may be an uncommon manifestation of GAS tonsillopharyngitis, related to specific GAS strains, or occur only in untreated or predisposed hosts.

Pediatric autoimmune neuropsychiatric disorders associated with streptococcal infections is the diagnosis applied to a subgroup of children with new-onset neuropsychiatric and behavioral problems following GAS infection. While much evidence indicates an association between GAS infections and PANDAS symptoms, no large prospective cohort studies exist to determine if GAS increases the risk for PANDAS symptoms.

Our study determined that GAS infection was not associated with the development of a mild PANDAS variant (which we had defined a priori as 2 or more new-onset PANDAS symptoms). Ours is the first large prospective study of PANDAS symptoms. Since the entity of PANDAS is controversial, this negative study is important. While there was no relationship between GAS and OCD or tic disorders in this study, it remains possible that antibiotic treatment (which patients in our study received for symptomatic GAS infections) prevents or treats PANDAS symptoms, that PANDAS occurs more frequently in patients with asymptomatic yet infectious GAS, or that even mild PANDAS variants occur at a rate too rare to identify by our sample size.

Correspondence: Eliana Miller Perrin, MD, MPH, Division of General Pediatrics and Adolescent Medicine, Department of Pediatrics, University of North Carolina at Chapel Hill, CB 7220, 200 Mason Farm Rd, 5th floor, Chapel Hill, NC 27599-7220 (eliana_perrin@med.unc.edu).

Accepted for publication April 14, 2004.

This study was funded by The Robert Wood Johnson Clinical Scholars Program, Princeton, NJ (Dr Perrin), and grant HD01441 from the National Institutes of Health K12 University of Carolina Building Interdisciplinary Research Careers in Women's Health Career Development Program, Bethesda, Md (Dr Perrin).

Dr Perrin was a research fellow of The Robert Wood Johnson Clinical Scholars Program during the conception and analysis of this research.

We thank Anne Francis, MD; Steven Marsocci, MD; William Hoeger, MD; Carolyn Cleary, MD; Ann Sorrento, MSN, PNP; Kathleen White-Ryan, MSN, PNP; Juli Miller, RN, CCRC; Traci Stanton, AAS; Doreen Francis, RN, CCRC; Sally Thomas, LPN, CCRC; and Melissa Maschoff, RN, of the Elmwood Pediatric Group, Rochester, NY, without whom this study would not have been possible. We thank Nikki Rogers, BS, and Halle Amick, BFA, for tireless help with data entry and administrative support and Carol Porter, BS, and Andrew Perrin, PhD, for help with data management. We appreciate the parents and children who participated in this study. We thank Joanne Garrett, PhD, and Robert Hamer, PhD, for advice on statistical analysis and David Ransohoff, MD, for advice on epidemiological methods.

This study was presented in part at the Pediatric Academic Societies Annual Meeting; May 5, 2003; Seattle, Wash, and at The Robert Wood Johnson Clinical Scholars Program National Meeting; November 7, 2002; Ft Lauderdale, Fla.