Stevens-Johnson syndrome (SJS) is an immune-mediated blistering disorder of the skin and mucous membranes; complications include blindness, urethral strictures, sepsis and death.13,14
M. pneumoniae has been recognized as the leading infectious trigger of SJS in children and adolescents since the 1960s.15 Children with M. pneumoniae-associated SJS often have prominent mucositis and limited skin involvement.16
In November 2013, clinicians at Children’s Hospital Colorado (CHCO) noted an unusual number of children diagnosed with SJS. An investigation into this cluster of cases revealed that of the 8 SJS case-patients (8–16 years of age) identified between September 1 and November 30, 2013, 5 had confirmed [M. pneumoniae polymerase chain reaction (PCR) positive] and 3 had possible (concurrent radiographic pneumonia or fever plus cough, shortness of breath or hypoxia) infection with M. pneumoniae, making this one of the largest reported outbreaks of M. pneumoniae-associated SJS to date.17
METHODSOutbreak SettingThe CHCO is the primary children’s referral hospital for the state of Colorado, with a catchment population of more than 1.2 million children. Children with severe illness requiring specialty care (such as SJS) are commonly referred from smaller hospitals throughout Colorado and the surrounding states. In the 5 years prior to September 1, 2013, CHCO treated an average of 7.5 children with SJS per year, of whom 32% had confirmed or probable M. pneumoniae infection on chart review.17 Therefore, the 8 cases that occurred between September 1 and November 30, 2013 represented a significant rise above the expected number of cases by cumulative sum analysis,17 prompting further investigation into the possible reasons for this increase.
Pneumonia Case FindingA case of pneumonia was defined as a pneumonia diagnosis indicated by an International Classification of Diseases, 9th revision (ICD-9) code of 483.0 (pneumonia caused by M. pneumoniae), 482.9 (bacterial pneumonia, unspecified) or 486 (pneumonia, unspecified organism) in a patient 5–21 years of age from January 1, 2009 to March 31, 2014. We selected this age range to reflect the population of children and young adults most at risk for developing M. pneumoniae pneumonia (and subsequent SJS) as well as to exclude infants and young children in whom pneumonia is both more common and more likely to be because of a viral etiology.1,18,19 Only the first occurrence of pneumonia within a 12-month period was included. We selected the time period to parallel the time for which independent ICD-9 codes were available for SJS.
We contacted 5 hospitals in the Denver metropolitan area, 1 regional Colorado health management organization, the Colorado Hospital Association (CHA) and the schools and day-care facilities of M. pneumoniae-associated SJS case-patients and their household contacts 21 years of age or younger to assess the trend in pneumonia cases among children 5–21 years of age. Here, we report data from the most complete and relevant data sources: CHCO, the largest children’s hospital in Colorado, and the CHA hospital discharge dataset, which contains patient demographics and discharge diagnoses from approximately 85 hospitals in the state of Colorado. Pneumonia cases at CHCO were identified from the hospital-wide network, including inpatient, emergency department and outpatient visits. Cases from the CHA discharge database are for inpatients only.
Laboratory InvestigationWe requested frozen residuals of M. pneumoniae PCR-positive respiratory specimens collected from Colorado residents between January 1 and December 10, 2013 from the 5 hospitals and their 4 affiliated referral laboratories. These specimens were shipped to the Pneumonia Response and Surveillance Laboratory at the Centers for Disease Control and Prevention (CDC) for further analysis. We collected all M. pneumoniae PCR-positive specimens prospectively identified by CHCO and its affiliated referral laboratory from December 10, 2013 to January 31, 2014. All accessible specimens were accepted for testing, including those from patients 5 years of age or younger or 21 years of age or older.
All specimens received at CDC were re-tested for M. pneumoniae in triplicate using a validated multiplex real-time PCR assay as previously described.20 Culture was attempted on all PCR-positive specimens using previously described methods8,21 to obtain isolates for further molecular characterization. All isolates were confirmed by M. pneumoniae-specific PCR.
Multilocus variable-number tandem-repeat analysis (MLVA) was performed on nucleic acid extracted from all PCR-positive specimens and isolates as previously described.22,23Typing of the major adhesion molecule P1 was performed on all isolates using a high-resolution melt assay for determination of types 1 and 2 and identification of genetic variants.24 Classification as types 1, 2 or variant was based upon comparison of melting profiles to reference strains M129 (type 1) and FH (type 2) included in each run.
All PCR-positive specimens were assessed for macrolide susceptibility using an high-resolution melt assay to detect single base transitions (A to G) at position 2063 or 2064 within the 23S rRNA gene, which are known to confer resistance to macrolide antibiotics.25 Specimens were classified as susceptible or resistant based upon comparison of melting profiles to reference strains as previously described.26 In some cases, sequencing of the 23S rRNA gene was also performed to confirm results or identify the specific mutation present.
Household Contact Survey to Assess <italic>M. pneumoniae</italic> TransmissionHouseholds of 3 SJS case-patients with confirmed M. pneumoniae infection were considered eligible for survey, as these case-patients were diagnosed within 6 weeks of the onset of the outbreak investigation.17 We surveyed household contacts, defined as any person living in the same residential unit as an SJS case-patient with confirmed M. pneumoniae infection, for each of these 3 case-patients. Consenting adults and assenting children were interviewed about the presence and timing of respiratory or skin symptoms as well as any clinic visits, hospitalizations, diagnoses of pneumonia or medication use over the preceding 2 months. Parents were asked to respond to all interview questions on behalf of children 6 years of age or younger and to participate in interviews of children 7 years of age or older to assist with recall. Combined nasopharyngeal/oropharyngeal (NP/OP) swabs were obtained from all consenting/assenting household contacts and submitted to CDC for M. pneumoniae PCR testing, culture, MLVA typing, P1 typing and macrolide susceptibility testing as previously described. The potential incubation period for M. pneumoniae was considered to be 1–4 weeks.27
Data AnalysisWe analyzed trends in pneumonia case counts over time from the CHCO hospital network (inpatient and outpatient) and the Colorado state CHA hospital discharge dataset (inpatient only). To facilitate comparison of the 2013–2014 M. pneumoniae season to prior years, we averaged cases by month from the previous 3 years (we did not include the season between 2009 and 2010 because of the difficulty in adjusting for the rise in pneumonia cases caused by H1N1 during that season). We compared the MLVA types, P1 types and macrolide resistance of the M. pneumoniae strains from patients with and without SJS, and from SJS case-patients and their household contacts. All analyses were performed using Microsoft Excel and SAS version 9.3.
Outbreak Response and Ethical ReviewThis investigation was conducted as a collaborative effort between hospital staff at CHCO, the Colorado Department of Public Health and Environment and the CDC. Investigation methods were reviewed by the CDC and determined to be nonresearch because this investigation was part of a public health outbreak response. For the household contact survey, written consent for the interview and collection of an NP/OP specimen were obtained from parents/guardians for their participation and that of all children 18 years of age or younger; in addition, written assent was obtained from all children 7 years of age or older.
DISCUSSIONThis investigation was among the first to involve real-time molecular typing of M. pneumoniae by P1 and MLVA during an outbreak, providing a unique opportunity to correlate laboratory and epidemiologic findings. The results of our epidemiologic investigation suggest that an epidemic of M. pneumoniae occurred in Colorado among children 5–21 years of age from September of 2013 to March of 2014. Furthermore, the onset of this epidemic coincided both with the onset of M. pneumoniae-associated SJS cases and with the introduction of secondary M. pneumoniae strains of MLVA type 3-X-6-2, joining the dominant circulating strain, MLVA type 4-5-7-2. The use of MLVA typing also established molecular evidence for intra-household transmission of M. pneumoniae.
We believe that the unusually heavy burden of M. pneumoniae in Colorado during the fall of 2013 likely contributed to the concurrent increase in M. pneumoniae-associated SJS cases. However, no cases of SJS were observed at CHCO during either December 2013 or January 2014, despite the fact that concentrations of M. pneumoniae remained elevated above baseline and active surveillance for SJS cases was continued during this time (Active surveillance for SJS cases at CHCO was continued through January 31, 2014 and consisted of screening for SJS cases by ICD-9 code [695.13 (SJS), 695.14 (SJS-toxic epidermal necrolysis, or SJS-TEN) and 695.15 (TEN)] and monitoring for cases by specialist services typically called to consult on patients with SJS (including pediatric infectious disease, ophthalmology and critical care services); pediatric hospitalists were also notified of the investigation and asked to report cases). Therefore, we considered whether a certain M. pneumoniae strain (distinguished by MLVA type) may have been more likely to trigger SJS in children. We observed a trend toward MLVA type 3-X-6-2 among SJS case-patients (60% vs. 29% of patients with pneumonia only) and a decrease in MLVA type 3-X-6-2 among pneumonia cases in January 2014, the last month for which data were collected. This decrease roughly paralleled the drop off in SJS cases. However, the different MLVA types observed among SJS cases and the fact that cases with similar MLVA types were not closely clustered geographically suggest that MLVA type may not be related to SJS. Ultimately, the numbers were too small to draw conclusions about the impact of MLVA type on development of SJS. It is possible that more advanced techniques such as whole genome sequencing could identify relevant differences between strains; alternatively, other host features (genetics, co-infections, concurrent medications), which were not evaluated in this investigation, may play a role in determining which children with M. pneumoniae are more likely to develop SJS.
M. pneumoniae case counts in Colorado reached a 5-year peak between 2013 and 2014. Such “epidemic seasons” of M. pneumoniae have been observed to occur at 4- to 7-year intervals and may reflect shifts in herd immunity as the population is exposed to different strains over time.28–30 The presence of multiple MLVA types in epidemic seasons of M. pneumoniae, observed in this investigation, has been previously documented.23,31,32
Our investigation found an overall prevalence of macrolide resistance of 7%, which is consistent with recent estimates elsewhere in the United States.31 While rising global rates of macrolide resistance, particularly in Asia,33–36 have sparked discussion about best first-line antimicrobial treatment for M. pneumoniae disease in children, our findings support continued use of azithromycin or an alternative macrolide as first-line antimicrobial agents in the United States.6 However, macrolide use may result in the rapid development of resistance34,37 and its efficacy in the treatment of lower respiratory infections in children has been questioned.38
The household survey represents a rare opportunity to examine molecular characteristics of M. pneumoniae strains transmitted within a household and highlights several notable findings. First, the majority of household contacts (87%) reported respiratory symptoms, and one-third tested positive for M. pneumoniae despite a time lapse of several weeks between symptoms and specimen collection in most cases. P1 and MLVA types between the case patient and associated household contacts were identical in all cases. The M. pneumoniae-positive specimen obtained from the 20-year-old sibling of the second case-patient showed a macrolide-resistant genotype, while the specimen from the case-patient was found to be susceptible. Interestingly, the sibling had been diagnosed with pneumonia and was treated with a 5-day course of azithromycin roughly 1 month before we obtained the sample, raising the possibility that the sibling’s recent exposure to azithromycin may have induced resistance.39,40 The original specimen from the case-patient was obtained before she received any treatment. The potentially long incubation period of M. pneumoniae and the lack of symptom specificity make it difficult to accurately trace transmission within households, but the timing of the PCR-positive cases’ symptom onset is consistent with transmission either from 1 household member to another or with infection from a common source (as might be suspected in household 1, where the 2 PCR-positive household members had symptom onset on the same day). Together, these findings support previous studies documenting children as the primary reservoir of M. pneumoniae,7 a long incubation period of up to 4 weeks,9,10,41,42 asymptomatic carriage for several weeks following infection43,44 and the easy spread of M. pneumoniae within households.8–10
This investigation calls attention to the fact that our knowledge of the epidemiology of M. pneumoniae remains rudimentary. This probable epidemic of M. pneumoniae, which potentially affected a large number of individuals in Colorado, went undetected until the unusual occurrence of this cluster of SJS cases prompted further exploration. In addition, it is unlikely that this epidemic was limited to 1 state. The ability to detect epidemic M. pneumoniae has implications for public health interventions such as timely outreach to clinicians, health-care facilities and schools; despite controversy over efficacy of macrolide therapy in an individual child, treatment with azithromycin during an epidemic may help to disrupt transmission to others,45 and in rare instances, targeted prophylaxis may be indicated.8 There are also important clinical implications for the early detection of epidemic M. pneumoniae, including ensuring appropriate treatment for children with CAP, limiting inappropriate antibiotic use, which may lead to resistance, and alerting clinicians to the possibility of M. pneumoniae-associated sequelae. Currently, children with CAP are likely to receive amoxicillin as first-line empiric therapy, which is ineffective against M. pneumoniae,6 particularly if their symptoms are not compatible with the common clinical perception of “walking pneumonia.”
A major obstacle to detecting M. pneumoniae epidemics is the lack of routine testing for this organism. Several factors contribute to low rates of diagnostic testing. First, current guidelines instruct clinicians to order diagnostic testing for children only when signs and symptoms suspicious for M. pneumoniae are present.6 However, emerging evidence suggests that CAP caused by M. pneumoniae may be clinically indistinguishable from that because of other etiologies.1,46 Furthermore, in many settings, PCR-based testing is preferred over serology for its superior specificity and lack of required paired specimens44,47; however, PCR is costly, may have slow turnaround time and may be unavailable to clinicians. Finally, because children may test PCR positive for several weeks after an infection,43,44 positive results may not reflect acute infection, complicating the clinical interpretation and contributing to clinician dissatisfaction with M. pneumoniae testing. Increased adoption by clinical labs of new Food and Drug Administration-cleared multipathogen molecular technology, which includes M. pneumoniae in addition to other common bacterial and viral respiratory pathogens,48 will address some of these issues. Indeed, rapid detection of increasing numbers of M. pneumoniae-positive respiratory specimens by such a panel at CHCO heralded the onset of the outbreak and quickly established the association between this organism and SJS.
There are several limitations to this investigation. In assessing the community trends in M. pneumoniae infection over time, we were unable to rely on laboratory-confirmed cases of M. pneumoniae because of inconsistent diagnostic testing in this population. Instead, we used ICD-9 codes for pneumonia (including unspecified etiology) as proxy variables for M. pneumoniae infection, which may have been less specific, and missed cases of M. pneumoniae, which resulted in a presentation besides pneumonia. Furthermore, it is controversial whether M. pneumoniae may be carried in the nasopharynx of a child without causing illness. A recent study43 suggested that detection rates of M. pneumoniae were almost equivalent between asymptomatic children and those with respiratory symptoms, while other studies have found little to no carriage in healthy children1,49,50; colonization with M. pneumoniae in the children studied could lead to overestimation of M. pneumoniae-associated disease. In assessing the molecular features of M. pneumoniae in cases of M. pneumoniae-associated SJS, we are limited by the small number of PCR-positive M. pneumoniae specimens available. Because the molecular characteristics of M. pneumoniae in SJS have not been previously reported, we are unable to correlate our findings with prior studies. In addition, current molecular typing methods (MLVA and P1) are limited in their ability to distinguish strains and may miss important variation. Finally, the number of household contacts surveyed is small and may not reflect trends in the broader population.
In summary, this investigation into one of the largest documented outbreaks of M. pneumoniae-associated SJS revealed an epidemic of M. pneumoniae pneumonia. Household transmission of M. pneumoniae was common within the households investigated, and the novel use of MLVA and P1 typing revealed that molecular characteristics were conserved among household contacts of case-patients. Improved availability and affordability of reliable and rapid M. pneumoniae diagnostic testing are needed to increase testing rates in the pediatric population. This will allow for better appreciation of the incidence and spectrum of M. pneumoniae infections as well as facilitate appropriate treatment and earlier detection of epidemics. Whole genome sequencing may yield superior strain characterization and provide further insight into the clinical implications of strain features. Human genetic analysis may also be needed to reveal the biologic reasons for the development of M. pneumoniae-associated SJS.