As part of a European study of severe S. pyogenes infections, the United Kingdom undertook enhanced surveillance during 2003–2004 (2,5,6). Case-patients were defined as persons with S. pyogenes isolated from a sterile site or from a nonsterile site if the patient had pneumonia, necrotizing fasciitis, puerperal sepsis, meningitis, septic arthritis, or if streptococcal toxic shock syndrome (STSS) developed (7). Methods are reported elsewhere (2,8).

To identify patient outcome, data from England and Wales were linked to death registrations obtained from the Office for National Statistics by using probabilistic methods (9). Deaths were identified that occurred up to 30 days after diagnosis of infection; time between diagnosis and death was measured from the date the culture-positive specimen was taken.

Data were analyzed by using STATA statistical software version 8.2 (Stata Corporation, College Station, TX, USA). A nonproportional test for equality of survivor function (Peto-Peto-Prentice) was used to assess differences between subgroups, with Cox proportional hazards regression used for multivariable analysis.

Of the 3,566 case-patients with severe S. pyogenes infection, 3,422 (96%) had sufficient identifiers to be linked to death registrations and were used for all further analyses. Overall, 698 (20%) case-patients died within 30 days after collection of culture-positive specimens. Risk for death was highest within 1 day after specimen collection (11% [375] extending to 16% [559]) for the first 7 days, beyond which risk for death dropped substantially.

Analysis of the certified cause of death identified an infectious underlying cause in 280 (50%) deaths occurring from any International Classification of Diseases, 10th Revision, condition classification within 7 days. S. pyogenes infection was specified as the cause of death in 5 of these (Table).

Figure 1

Figure 1. Kaplan-Meier analysis of time to death after diagnosis of severe Streptococcus pyogenes infection, by age, England and Wales, 2003–2004.

Age strongly influenced survival rate; oldest patients had the poorest survival rate-(p<0.001; Figure 1). Most deaths in patients <45 years of age (73/104 [70%]) occurred within the first 2 days, whereas deaths in older age groups were more dispersed over time. Other patient factors were independently associated with risk for death ( Appendix Table). Of patients without any identified concurrent illnesses, 87/518 (17%) died within 7 days.

Appendix Figure

Appendix Figure. Monthly number of cases and 7-day case-fatality rate for patients dying within 7 days after diagnosis of severe Streptococcus pyogenes infection, England and Wales, 2003-2004.

Case-fatality rates paralleled seasonal incidence (Appendix Figure), and were highest from December to April (17%–21%), gradually falling through the summer to their lowest point in October (6%). After adjustment for other significant factors, patients whose infection was diagnosed in October were 82% less likely to die than were those whose infection was diagnosed in January.

Figure 2

Figure 2. Kaplan-Meier analysis of time to death after diagnosis of severe Streptococcus pyogenes infection, by development of streptococcal toxic shock syndrome (STSS), England and Wales, 2003–2004.

Patients identified with necrotizing fasciitis had the highest risk for death within 7 days (34%), >2 times higher than patients with other clinical manifestations, after adjustment for other significant factors. Patients who reported gastrointestinal symptoms were 2 times as likely as those who did not to die within the first 7 days (p = 0.02). Only 9/76 (12%) patients who had gastrointestinal symptoms met the case definition for STSS. Cellulitis, the most common clinical symptom, was associated with more deaths (130/438 [30%] patients) than was any other condition. Survival probability in the 30 days after a culture-positive specimen was significantly reduced among patients in whom STSS developed (p<0.001; Figure 2); 47/178 (26%) of these patients died within a day of specimen collection.

Risk for death varied according to the emm/M-type responsible for the infection. The highest risk was associated with emm/M3 (33%) or emm/M1 (28%), with emm/M3 being borderline significant after adjustment for other significant factors.

This analysis highlighted the scale and rapidity of deaths in patients with severe S. pyogenes infection, reemphasizing the importance of early recognition of invasive disease and prompt initiation of antimicrobial drug and supportive therapy. A number of factors conferred a heightened risk for death, which other studies also have found: increasing age, diagnosis of necrotizing fasciitis or pneumonia, and underlying malignancy (1,10,11). Although necrotizing fasciitis carried the highest risk for all-cause mortality, it is a relatively rare condition, accounting for only 10% of all deaths, compared with the more common and typically less severe cellulitis (30% of deaths). Some evidence indicated that emm/M3 was more commonly associated with death than were other emm/M-types, also found elsewhere (1,11). Development of STSS was a strong predictor of poor outcome, although whether this syndrome is independently related to death is unclear, given that several of its constituent markers effectively denote the progressive failure of organ systems. For this reason, STSS was not included in the multivariable analysis. Notably, patients with gastrointestinal symptoms had a poorer outcome than others; presumably this is a sign of overwhelming sepsis possibly linked to toxin production. Although relatively uncommon (3% of case-patients), gastrointestinal symptoms clearly are of diagnostic importance and should be included in the severity assessment of patients with other signs and symptoms of S. pyogenes infection.

An intriguing and novel finding from this analysis is that risk for death mirrors seasonal changes in incidence (highest in the winter/spring months and falling substantially to a nadir in October). An unadjusted-for confounder could explain this pattern, such as a preceding viral infection, but no candidates with a suitable seasonal pattern are apparent. Although emm/M-type was included in the model, differences could exist in the circulation of specific subtypes with particular virulence profiles throughout the year, again not accounted for in this analysis. Seasonal differences in case-fatality rates may yield some important clues about the drivers behind the seasonal incidence pattern of S. pyogenes infections. An explanation centered wholly around transmission dynamics seems less favored because it would not explain the changing risk for death. An immunologically focused explanation would fit better because it could explain both the changing incidence rate and the changing case-fatality rate. The length of daylight in a given day can affect the production of vitamin D and melatonin. Both vitamin D and melatonin are known to impact immune function (12–14). Although neither appear to be consistent with an autumnal boost to host immunity, if the beneficial effect of vitamin D production during the summer has a cumulative maximal impact in the autumn (15), then this theory could offer an explanation.

Because S. pyogenes is a largely sporadic infection occurring diffusely throughout the population, opportunities for control of severe infections remain limited. The 26 serotypes included in a multivalent vaccine currently under phase II clinical trials represent 67% of all isolates collected in this study, and 80% of all deaths occurring within 7 days after diagnosis (16). The impact of any vaccine will clearly depend on many factors, with the possibility of serotype replacement undermining any such efforts. The challenge to provide a lasting control measure for these devastating infections will continue long into the future.