In this issue of Annals, Berrios and colleagues [1] report a prospective study conducted in Chile in which prophylaxis for rheumatic fever recurrences was discontinued and close medical surveillance was maintained. The subsequent rheumatic recurrence rate was only 0.7 per 100 patient-years. Moreover, no recurrences were observed during an outbreak of 52 cases of acute rheumatic fever in the study area in 1986.

How can one explain focal outbreaks of acute rheumatic fever or dramatic changes in rheumatic fever recurrence rates in populations that differ little or not at all in racial, ethnic, or socioeconomic circumstances? Is the explanation simply the relative frequency of group A streptococcal infection and its appropriate treatment or are there, in addition, marked differences among group A streptococcal strains that make some strains far more rheumatogenic than others? Those of us who have pondered these issues for several decades recall most vividly that acute rheumatic fever was one of the greatest plagues of the first half of this century. As a freshman medical student in 1941, I learned that acute rheumatic fever killed more school children in the United States than all other diseases combined. Moreover, survivors afflicted with rheumatic heart disease had frequent recurrent rheumatic attacks. Great enthusiasm greeted the introduction of sulfonamide prophylaxis for secondary rheumatic attacks in the late 1930s [2]. The introduction of penicillin in the 1940s (to which group A streptococci were sensitive) offered the opportunity not only to prevent rheumatic recurrences but also to prevent initial attacks of the disease [3].

In the 1950s, Wyeth, Inc. introduced benzathine penicillin G, a repository penicillin compound of such limited solubility that a single intramuscular injection could create a depot that released minute amounts of penicillin detectable in blood for several weeks [4]. Benzathine penicillin G offered virtually perfect control of group A streptococcal infection in rheumatic children and adults. Such control established the fact that group A streptococci alone were the cause of recurrent rheumatic attacks. This hypothesis had been difficult to sell to many leading authorities who observed recurrence of acute rheumatic fever when the antecedent streptococcal infection was asymptomatic, as was not infrequently the case.

So dramatic were the uncontrolled pilot studies of monthly injections of benzathine penicillin G to children recovering from recent rheumatic fever that the results were readily accepted for publication [5]. The expected frequency of rheumatic recurrences in this cohort was approximately 17% per year, yet none were observed. Moreover, the drug was well tolerated. This preliminary trial enabled us to design and launch a properly controlled study [6] that yielded information about the outcome of acute rheumatic fever in the absence of subsequent intercurrent streptococcal infection and about the variables that determine the risk for rheumatic recurrences after such infections [7].

Interestingly, although factors influencing the rheumatic attack rate, such as age, duration from the previous rheumatic attack, presence and severity of rheumatic heart disease, as well as the number of previous rheumatic attacks, were all important variables, the nature of the antecedent infection was characterized only quantitatively-by the magnitude of the immune response (for example, the increase in antistreptolysin O titer). What was not yet considered was what kind of group A streptococcus initiated the attacks. Were there simply quantitative differences in virulence or also qualitative ones that affected the attack rate of rheumatic fever following streptococcal pharyngeal infections [8]?

That qualitative differences existed in group A streptococci that were related to the kind of disease they produced was hardly a novel notion in the 1960s [9]. It had been shown that certain pharyngeal strains, notably M-protein types 2 and 4, did not reactivate rheumatic fever [10]. Strains causing streptococcal pyoderma could cause acute glomerulonephritis but not rheumatic fever [11], even when they also infected the pharynx [12]. The virulent M-protein-rich, encapsulated strains frequently found in military epidemics of acute rheumatic fever belonged to a limited number of M-protein serotypes (for example, 1, 3, 5, 6, 18, 24, and others [13]). During convalescence from streptococcal pharyngitis, these strains rapidly attenuated, losing their virulence properties of rich M-protein content and large hyaluronate capsules (responsible for the formation of "mucoid" colonies on blood-agar plates). Moreover, such strains in vitro were difficult to keep in this high phase of virulence, and the strains required frequent passage through fresh human blood or through mice to remain virulent. Strains such as these could spread rapidly through closed populations (military recruits, schools, institutions), but they also became relatively noninfectious when attenuated [14]. Such attenuated strains, however, could still stubbornly colonize the throat, because the pharyngeal adherence factor of group A streptococci is a surface lipoteichoic acid and not M protein [15].

Therefore, a report in 1969 [16] of the absence of rheumatic fever following an extensive food-borne military epidemic of streptococcal sore throat prompted an editorial [17] suggesting that some strains of group A streptococci might be neither nephritogenic nor rheumatogenic and yet these strains could still produce streptococcal pharyngitis. Because the elusive factor(s) of "rheumatogenicity" of group A streptococci has remained undefined, American Heart Association committee recommendations for the prevention of either primary or secondary attacks of acute rheumatic fever have remained firm. Particularly problematic is the duration for which prophylaxis against recurrences needed to be sustained. Expert committee members argued between "forever" and some limit; the compromise was the more acceptable suggestion of "indefinitely".

The spectacular rate of decline in the 1970s in the prevalence of both primary and secondary acute rheumatic fever in the United States and other developed countries left many authorities puzzled by the continued high prevalence of group A streptococcal sore throat in school children [18]. Some attributed this decline primarily to prevention of acute rheumatic fever by the diligent and effective penicillin therapy for streptococcal sore throat [19]. This explanation failed to take into account the well-known fact that primary penicillin prophylaxis for streptococcal sore throat could not be effective for at least 30% of patients because at least this many of those who contract acute rheumatic fever do not recall ever having had an antecedent sore throat and, therefore, could not have been treated [3]. Vigorous use of effective penicillin therapy for reported sore throat might, however, eventually reduce the spread of highly virulent rheumatogenic strains in a close-living population and hasten the replacement by less dangerous streptococcal strains. Indeed, coincident with the disappearance of rheumatic fever in the United States was the observed disappearance of the characteristic rheumatogenic strains that were found among the M-protein serotypes known to be associated with epidemic rheumatic fever [20].

The remarkable decline in acute rheumatic fever in the 1970s encouraged my colleagues and me to consider discontinuing streptococcal prophylaxis in rheumatic patients who were classified as being at lowest risk for recurrences. In the absence of rheumatic fever in the environment, such management policies appeared to us not only ethical but consistent with what was already becoming common practice in the United States and elsewhere. Our decision seemed justified when acute rheumatic fever did not recur in this cohort despite immunologically documented intercurrent streptococcal infections [21]. Because these infections were diagnosed retroactively by serum antibody titers, we were unable to identify and study all of the antecedent infectious strains of streptococci. Those few strains that we did identify were group A streptococci of low virulence, and those that were M-typable did not belong to the serotypes that have been associated with acute rheumatic fever.

The resurgence of acute rheumatic fever in the United States in the 1980s provided the opportunity to seek out the identity and characteristics of rheumatogenic strains [22]. Unfortunately, it has been difficult to do this because routine clinical microbiology laboratories restrict their identification to "group A" strains, and few investigators now have sufficient interest or experience to go beyond this limited taxonomic approach. Nonetheless, in the recent focal acute rheumatic fever epidemics, especially those at military bases [23], the epidemiologists of the Centers for Disease Control and Prevention as well as others have identified some of the old rheumatogenic mucoid M-rich strains within foci of some of the U.S. outbreaks, most notably a mucoid M-18 strain notorious former military epidemics of acute rheumatic fever in the Rocky Mountain region. These strains seem now to be routed by intensive penicillin therapy, and acute rheumatic fever has been extinguished without eliminating all strains of group A streptococci from the affected population [24].

But when should we remove the mantle of penicillin protection from our most susceptible hosts? The Chilean study [1] brings this difficult decision into the 1990s. A special feature of the Chilean study is that it began in a local cohort riddled with acute rheumatic fever and from which rheumatogenic streptococci appeared to have been eradicated. The authors [1] noted that little evidence existed to suggest that their patients, in whom prophylaxis was discontinued, were exposed to highly rheumatogenic strains. The percentage of strains that were M typable was remarkably low—only 7% compared with 52% of throat isolates in their previous studies in this cohort. Yet pockets of rheumatic fever still persist in this area, producing a situation similar to that in other places in the world where focal populations are heavily affected by acute rheumatic fever, whereas nearby populations are almost completely spared (reviewed in reference 25). Indeed, the explosive epidemics at Great Lakes Naval Base, Illinois [26], were caused by M-identifiable, hugely encapsulated mucoid strains (for example, M-24) that were inoculated by newcomers from all parts of the country and that quickly swept through the close-living ranks of naval recruits. But a few hundred feet away, at the other training units of the same base, only sporadic cases of rheumatic fever were observed despite a high endemic rate of streptococcal pharyngitis in these nearby cohorts. I carried off that M-24 strain and freeze-dried it; it was locked up like a genie in a bottle. I insisted that my laboratory staff not release it unless they were receiving careful penicillin prophylaxis!

The late Dr. Edwin Beachey subsequently [27] used M-24 and other similar strains to determine the primary structure of the M-protein molecule and to identify its elongated form of recurring peptide units, which were used to isolate the type-specific epitopes that could potentially produce type-specific vaccines. Fischetti and colleagues [28] found this extended molecular M-protein structure in all strains preserved at the Rockefeller University that had been associated with acute rheumatic fever outbreaks. Are these strains the culprits in the severe rheumatic fever still occurring in focal populations such as in Soweto or among the native Tahitians or Maoris from New Zealand? Do they haunt some inner city areas in Miami and Cairo, some Japanese islands, and Utah communities (cited in reference 29)?

What, then, might be the agenda for the control of rheumatic fever in the current decade? The first 50 years of this century of research in acute rheumatic fever established with great difficulty that acute rheumatic fever did not recur without group A streptococcal infection. It should not take us another 50 years to discover which group A streptococcal strains cause this disease and how they do it. If, indeed, rheumatogenic strains occur most often within the limited number of M-serotypes that are found in populations affected by acute rheumatic fever, a multivalent vaccine containing the most culpable M types might be feasible [28, 30]. It could provide one strategy for the eventual eradication of acute rheumatic fever. Meanwhile, wherever acute rheumatic fever rears its ugly head, we are obliged to depend on effective penicillin therapy to rout the offending strains: When we have done so, it will be entirely safe to discontinue prophylaxis, even in our most vulnerable rheumatic hosts.