Background. The explanation for the very substantial decrease in the incidence of acute rheumatic fever in the United States, particularly over the past 50 years, is unclear. It has been proposed that certain M types of group A streptococci (GAS) include strains that are particularly rheumatogenic and that others are nonrheumatogenic.
Methods. We compared the M type distribution of GAS recovered from children from Chicago, Illinois, with acute pharyngitis during 1961–1968 to that of GAS recovered from Chicago children and children from across the United States in 2000–2004, with attention to changes in M types that previously were associated with rheumatogenic strains.
Results. The rheumatogenic types 3, 5, 6, 14, 18, 19, and 29 comprised 49.7% of 468 pharyngeal isolates during 1961–1968 but only 10.6% of 450 Chicago isolates during 2000–2004 (P < .001) and 17.9% of 3969 isolates nationwide during 2000–2004 (P < .001). Significant decreases in types 3, 5, and 6 and virtual disappearance of types 14, 18, 19, and 29 occurred between the 2 study periods. No change in the proportion of type 1 isolates, a highly heterogeneous group that includes some rheumatogenic strains, was observed. The nonrheumatogenic GAS types 2, 4, 22, and 28 increased from 4.9% to ∼28% of pharyngeal isolates in Chicago and nationwide between the 2 study periods (P < .001).
Conclusions. These data support the concept of rheumatogenic strains of GAS and indicate that the marked decrease in the incidence of acute rheumatic fever in the United States over the past 4 decades is correlated with the replacement of rheumatogenic types by nonrheumatogenic types in cases of acute streptococcal pharyngitis in children. The reasons underlying the observed change in distribution of M types remain to be elucidated.
The incidence of acute rheumatic fever (ARF) in the United States and western Europe has decreased markedly over the past century [1, 2]. This decrease accelerated during the 1970s, as was well documented in Chicago and Baltimore [2, 3] (figure 1). Resurgence in some US areas in the 1980s temporarily interrupted the decrease [4–8]. Very few cases of ARF have been documented recently in most areas of the United States, with substantial activity limited to the Salt Lake City, Utah, area . The explanation for the dramatic incidence decrease is unclear, because ARF represents the intersection of microbial (group A streptococci [GAS]) and host factors (human immunogenetic background), modulated by environmental factors, such as crowding, nutrition, and access to medical care [10, 11].
On the basis of observations from decades earlier, Stollerman [12, 13] and Bisno  proposed that certain GAS serotypes are particularly rheumatogenic. Streptococcal pharyngitis outbreaks caused by certain M types were highly associated with ARF, whereas other M types failed to initiate or reactivate ARF even in highly susceptible individuals [14–16]. M types particularly associated with ARF are 1, 3, 5, 6, 14, 18, 19, 24, 27, and 29 , whereas those rarely, if ever, leading to ARF include the relatively common types 2, 4, 12, 22, and 28 [12, 14, 17–22]. Subsequent studies emphasized the heightened virulence of rheumatogenic strains characterized by large hyaluronate capsules that produce “mucoid” colonies and more M protein.
Here, we contrast the type distribution of GAS isolates recovered from children from Chicago, Illinois, with acute pharyngitis during 1961–1968 (when ARF was highly prevalent) and isolates from Chicago and nationwide in 2000–2004 from our on-going pediatric pharyngeal surveillance study [23, 24]. We report that the prevalence of almost all putative rheumatogenic types of GAS has decreased sharply over the past 4 decades, closely paralleling the decrease in the incidence of ARF.
GAS strains were collected in 1961–1968 from children aged 3–15 years with acute pharyngitis in the Streptococcal Research Clinic of Children's Memorial Hospital on the north side of Chicago, which was directed by one of the authors (G.S.). β-Hemolytic isolates were confirmed to be GAS serologically by capillary precipitin reaction and were M-typed serologically using Lancefield grouping and typing sera .
In 2000, we initiated the US Streptococcal Pharyngitis Surveillance Study with 9 sites across the United States, including a pediatric office on the north side of Chicago ∼5 km from Children's Memorial Hospital, as recently described elsewhere . A tenth US site (near Salt Lake City, UT) was added in 2001. Approximately 100–120 pharyngeal GAS isolates were collected from each site each winter-spring season from 2000 to 2004. After serologic confirmation as GAS, isolates were emm genotyped (emm encodes the M protein) at the Streptococcal Research Laboratory of the Centers for Disease Control and Prevention (Atlanta, GA) . Concordance between serologic and genotypic results has been demonstrated .
We compared the M serotype distribution of Chicago isolates from the period of 1961–1968 to the emm genotype distribution of Chicago isolates and all US isolates from the period of 2000–2004. Differences were assessed by the χ2 test.
The distributions of pediatric pharyngeal GAS serotypes/genotypes in Chicago from the period of 1961–1968 (n = 468), in Chicago from the period of 2000–2004 (n = 450), and from 10 US sites from the period of 2000–2004 (n = 3969) are shown in table 1 in descending order of frequency during 1961–1968. The relative proportions of types 12 and 1 (2 of the 3 most common pharyngeal types during both 1961–1968 and 2000–2004)  did not differ. Compared with the 1961–1968 isolates, types 3, 6, 5, and 14 were significantly less prevalent among recent Chicago isolates (P < .001, P < .001, P < .05, and P < .001, respectively) and recent US isolates (P < .001 for each). Type 28 was very rare in the 1960s but was much more prominent recently in Chicago and nationwide (P < .001 for both). Type 4 strains also were more common recently (P < .001 in Chicago and P < .01 nationwide), whereas types 18, 29, 41, and 43 were all significantly less common during 2000–2004. In the 1960s, 10 other less common GAS types (8, 11, 13, 15, 19, 31, 33, 42, 46, and 47) were identified, constituting 3.3% of isolates. These 10 types accounted for only 0.9% of recent Chicago isolates (all emm11) and 1.2% of US isolates (P < .025 and P < .001, respectively). There were 10 untypable strains (2.1%) among isolates from 1961–1968. Approximately one-third (34.2%) of recent Chicago isolates and 26.6% of recent US isolates represented 11 and 20 other types, respectively, not identified in the 1960s.
Rheumatogenic types. Table 2 contrasts the prevalence of putative rheumatogenic GAS types [10–14]. The proportions of type 1 isolates from the 1960s and 2000s are almost identical. In marked contrast, types 3, 5, and 6 were significantly less prevalent among the recent Chicago isolates (P < .001, P < .05, and P < .001, respectively) and US isolates (P < .001 for each), compared with Chicago isolates from 1961–1968, when ARF was still prevalent (figure 1). Types 3, 5, and 6 together constituted 35.1% of Chicago isolates during 1961–1968 but only 10.4% of Chicago isolates during 2000–2004 (P < .001). In addition, there was virtual disappearance of types 14, 18, 19, and 29, which together constituted 14.6% of isolates from 1961–1968, with only a single emm14 isolate and no emm18, emm19, or emm29 isolates during 2000–2004. No type 24 or 27 strains were identified during either time period. The putative rheumatogenic types 3, 5, 6, 14, 18, 19, and 29 made up 49.7% of pharyngeal isolates during the 1960s but only 10.6% of isolates 40 years later (P < .001).
Nonrheumatogenic types. Table 3 compares the prevalence of GAS types thought to be nonrheumatogenic. Outbreaks of pharyngitis due to these types in past decades were not associated with ARF, even in high-risk populations [14–16]. The prevalence of emm12 GAS has not changed over time, but types 2, 4, 22, and 28 are each significantly more prevalent in recent years; in the aggregate, they were associated with 4.9% of cases of pharyngitis in the early era and with ∼28% of cases both nationally and in Chicago during 2000–2004 (P < .001 for each).
That ARF is now distinctly uncommon in almost all areas of the United States is very well documented [1–3]. The rheumatic fever wards and hospitals of several generations ago are now virtually nonexistent. The beginning of the decrease in the rate of ARF clearly antedated antibiotics [1, 27, 28], and the decrease accelerated in the 1970s [2, 3]. Figure 1 demonstrates the sharp decrease in the incidence of ARF in Chicago in that era. In the mid-to-late 1980s, there was an unexpected upsurge in the ARF rate in several areas of the United States [4–8]. Cases continue to be reported from Salt Lake City, although there are fewer cases than in the peak years of 1985 and 1998 . No ARF resurgence was observed in Chicago in the 1980s or subsequently.
The reasons for the changes in the incidence of ARF remain obscure. Possibilities include changes in host factors of susceptibility, in environmental conditions that may modulate susceptibility, or in characteristics of circulating GAS . Host determinants of susceptibility to ARF have been assessed , but changes in human genetic determinants cannot account for changes in the ARF incidence within a few generations. Much of the decrease in the early 20th century may relate to improved living conditions, with less crowding, improved nutrition, improved access to medical care, and availability of antibiotics [2, 10, 11, 27–29]. However, neither the precipitous decrease in the incidence during the 1970s nor the resurgence during the 1980s can be explained on this basis. Moreover, there is little evidence of a substantial change in socioeconomic status of the Chicago cohorts (i.e., a lower-middle-class population at Children's Memorial Hospital in the 1960s and a blue collar, urban, private practice population in 2000–2004). Differences in antibiotic usage are likely to have occurred, but it is unclear how that may have influenced the selection of specific genotypes of GAS, which are exquisitely susceptible to virtually all antibiotics commonly used in pediatric medicine. Interest has focused most on changes in circulating strains of GAS to explain the observed temporal and geographic fluctuations in the incidence of ARF [1, 11–14, 17, 18, 27, 30].
On the basis of epidemiologic data, Stollerman [12, 13, 17] and Bisno  proposed that certain GAS strains are more likely than others to lead to ARF (i.e., certain strains are more rheumatogenic) and that certain characteristics of individual GAS isolates are associated with rheumatogenicity. These included higher M protein content and mucoid colonial morphology indicative of a large capsule [19–21] and lack of in vitro production of serum opacity factor (SOF), an enzyme that interacts with serum α1 lipoproteins to result in opacification [12, 14, 17, 31]. Epidemiologic observations made decades earlier indicated that some outbreaks of streptococcal pharyngitis in rheumatic fever hospitals or wards were followed closely by recurrent ARF, whereas other outbreaks failed to trigger such recurrences, even in these highly susceptible populations [14–16]. In 1935, Coburn and Pauli  noted the rarity of ARF after M4 or M28 GAS pharyngitis. In 1941, Kuttner and Krumweide  emphasized the absence of ARF after M4 infection, despite the fact that M4 infection caused severe pharyngitis, whereas other strains caused less severe acute pharyngitis but triggered outbreaks of ARF.
Bisno , Stollerman , and Bisno et al.  also documented the lack of ARF recurrences among rheumatic heart disease patients after acute pharyngitis and/or colonization with GAS strains that were usually isolated from cutaneous sites, and they documented a correlation between pharyngitis caused by certain M types (M6 most notably) and higher ARF rates. Reviewing reports about ARF epidemics, Bisno noted that only M3, 5, 14, 18, and 24 had been associated with >1 epidemic; that M6, 19, 27, and 29 were each associated with a single epidemic; and that almost all ARF outbreaks were associated with SOF-negative strains of GAS . In the 1960s, Maxted et al.  noted the marked increase in extractable M protein from GAS strains associated with outbreaks of pharyngitis and ARF, compared with strains of the same serotypes isolated in communities in which pharyngitis was common but ARF was rare.
In 1989, Bessen showed that GAS serotypes could be separated into 2 classes on the basis of M protein characteristics: class I M proteins (but not class II M proteins) share a specific C repeat antigenic domain . Virtually all M types (with the possible exception of M11) associated previously with ARF outbreaks in the United States, Europe, and most other regions express a class I M protein, including M1, 3, 5, 6, 14, 17, 18, 19, 24, 27, and 29 , and class I M protein types are virtually exclusively SOF negative . Whatmore et al.  demonstrated extensive genetic variation within certain individual M types and distinct emm sequence clusters generally paralleling SOF status and historical associations with ARF, concluding that SOF-positive and SOF-negative M types represent distinct evolutionary lineages.
Observations during the local ARF resurgence in the United States in the 1980s and subsequently also strongly indicate that certain M types (particularly mucoid isolates) are rheumatogenic. Congeni reported GAS recovered from 5 patients with ARF or their siblings in Akron, Ohio, in 1985–1986, including M1, 5, 6 and 18 . Marcon et al.  collected 116 mucoid GAS pharyngeal isolates from 1986–1987 in Columbus, Ohio (coinciding with an ARF upsurge), and 40 of 40 M-typed isolates were M18, including the only isolate recovered from a patient with ARF. An ARF outbreak at Ft. Leonard Wood, Missouri, in 1987–1988 was associated with mostly mucoid GAS; 74% of isolates were M18, and 20% were M3 . Additionally, 32 of 33 SOF-negative GAS isolates from patients with ARF or their siblings in the mid-1980s belonged to 5 previously suspected rheumatogenic M types (M1, 3, 5, 6, and 18) . The same group later reported that ARF-associated strains were predominantly M3 and M18 and that a very high proportion of the ARF-linked M18, 5, and 6 strains were highly mucoid, as noted by others [12, 14, 20, 21, 38]. In some tropical areas of the world, very different spectra of GAS organisms have been associated with ARF, with some being M-nontypable or representing newly identified types.
With use of emm genotyping and multilocus sequence typing, Miner et al.  analyzed random GAS isolates from the 2 ARF peak years in Utah (1985 and 1998) and during 1992, a year with little ARF activity. The emm3 and emm18 strains were present during 1985 and 1998 but not 1992, suggesting their importance in the ARF resurgence . Smoot et al.  and, later, Veasy et al.  also found that, among Utah isolates (primarily from 1985–1992 and 1997–1999), mucoid colony morphology was associated with both the 1985 and 1998 ARF peaks and that >80% of the mucoid isolates were emm18. By genome sequencing and comparative microarray analysis of 36 M18 strains from the period of 1931–2000, Smoot also showed that Utah emm18 isolates from 1985 and 1998 were genetically identical . The waning of peak ARF activity, particularly in Utah, correlated with the absence of mucoid M/emm18 and M/emm3 strains in that area [22, 39, 40]. All of these findings support the concept that there are specific rheumatogenic GAS strains and that such strains have properties, particularly mucoidicity (encapsulation), associated with increased virulence and rheumatogenicity [17–21].
The present data show that the distribution of GAS pediatric pharyngeal types in Chicago during 1961–1968 was very different from the distribution in Chicago and nationwide during 2000–2004 (tables 1–3). Specifically, the proportion of cases of streptococcal pharyngitis caused by putative rheumatogenic serotypes (with the exception of emm1) was significantly higher and the proportion due to nonrheumatogenic types was significantly lower during the 1960s. Rheumatogenic types 3, 5, 6, 14, 18, 19, and 29 constituted 49.7% of pharyngeal isolates in the 1960s and only 10.6% of recent isolates in Chicago and 17.9% of recent isolates nationally (P < .001). Conversely, the putative nonrheumatogenic types 2, 4, 12, 22, and 28 caused 23.5% of streptococcal pharyngitis cases during 1961–1968 but 45.7% of cases in Chicago and 45.6% of cases nationally in recent years (P < .001).
That putative rheumatogenic M types predominated decades ago is supported by a study by Quinn et al. , which found that 72% of M-typable pharyngeal GAS isolates recovered from schoolchildren in Nashville, Tennessee, during 1953–1955 were of rheumatogenic types 1, 3, 5, 6, 18, 19, and 24, which is similar to our findings for 1961–1968. These changes in prevalent types can substantially explain the dramatic decrease in the prevalence of ARF in the United States in recent decades. Also contributing to this decrease may be improved living conditions and possible reduction in virulence of modern strains of rheumatogenic M types, reflected by lower capsule (decreased mucoidicity) and toxin and M protein production rates. These environmental and microbial factors have not been assessed systematically.
Serotype M1 strains merit special mention, because the proportion of these isolates over time did not change. This issue is complex, because M1 are highly diverse genetically, and they are prominent causes of uncomplicated pharyngitis, acute invasive infections, ARF, toxic shock syndrome, and acute glomerulonephritis [34, 35, 43–46]. Overall, these analyses show that M1 strains represent heterogeneous chromosomal lineages rather than a single phylogenetic branch  and that some M1 strains are more closely related to organisms that express another M protein than to other M1 strains . For example, on the basis of current emm-typing criteria, neither M1.2 nor 1.4 strains would be considered to be type emm1, because their emm sequences diverge considerably from emm 1.0, and M1.2 (emm1.2) and M1.4 (emm1.4) isolates represent multilocus sequence types (STs) unrelated to the major US M1 clone currently circulating (compare ST 28, typical of emm1.0 isolates from the US, with STs from types emm1–2 and emm1–4 [ST93 and ST92, respectively] at http://www.mlst.net). Considerable differences also exist among M1 strains regarding their ability to be opsonized by human immune sera [48, 49]. RFLP analysis of M1 organisms yielded 16 patterns, confirming their extensive chromosomal diversity .
Despite the great genetic diversity among older M1 strains, PFGE and multilocus sequence typing data indicate very little genetic diversity among US M1 clinical isolates during the past decade, nearly all of which carry the emm1.0 gene or closely similar alleles (B.B., unpublished data). This does not exclude the possibility that, during the 1960s, serotype M1 primarily included clonal types that differed from the current predominant M1 clone. Alternatively, subtle genetic change(s) within the predominant M1 clone possibly resulted in emergence of a successful less rheumatogenic subclone.
The evidence presented here does not explain directly why the well-documented shift in GAS type distribution in the United States over 4 decades occurred, but it likely reflects a highly complex set of environmental, host, and microbial factors.
This study has several limitations. Chicago subjects with pharyngitis in the 1960s were 3–15 years old, whereas those from 2000–2004 were 3–18 years old. Because 16–18-year-old subjects constituted only 1.3% of subjects, it is highly unlikely that they contribute significantly to the type distribution. The site of acquisition of throat isolates in 1961–1968 was an outpatient clinic at The Children's Memorial Hospital, whereas recent Chicago isolates were obtained at a private pediatric office in Chicago ∼5 km northwest of the hospital. The ethnic distribution of subjects in both eras is not available, but the hospital traditionally has served an ethnically and socioeconomically highly diverse population, as does the pediatric office.
Unfortunately, systematic study of encapsulation and other possible virulence markers was not a feature of this study. In earlier Chicago studies, the virulence of strains recovered from untreated children with nonexudative pharyngitis was examined in detail because of the absence of ARF in this group, and the results suggested the importance of mucoidicity (encapsulation) in virulence [19–21]. The GAS isolates from the 1960s are not available for additional studies.
The methods utilized to type GAS merit discussion. In the 1960s, organisms were grouped as group A and then serotyped utilizing classic Lancefield grouping and typing sera . Isolates from 2000–2004 were also confirmed to be group A by serologic testing and were then genotyped by PCR for the hypervariable region of the emm gene (encoding M protein) . emm Genotyping has substantially replaced serotyping. In the original assessment of correlation between the 2 typing systems, there were only a few discrepancies that did not involve types relevant to the data presented here .
In conclusion, our data strongly support (1) the concept that there are rheumatogenic and nonrheumatogenic types of GAS, and (2) that the marked decrease in the prevalence of ARF in the United States over recent decades is very likely to be the consequence of a dramatic decrease in the proportion of cases of acute streptococcal pharyngitis in children caused by rheumatogenic types and, in particular, by mucoid strains within those types. Better understanding of the impact of the distribution of GAS types on poststreptococcal complications like ARF may lead to strategies, including vaccine development, to control focal outbreaks and ARF in areas of the world where it is still highly prevalent.
Potential conflicts of interest. S.T.S has received research support from and owns stock in ID Biomedical and has been paid lecture fees by Pfizer, Wyeth, Abbott, and Glaxo-Smith Kline. J.B.D. has received research support from and owns stock in ID Biomedical. R.R.T. has received research support from ID Biomedical. G.S. and B.B.: no conflicts.