In accordance with the program objectives, the Public Health Laboratory Service (now part of the Health Protection Agency) initiated enhanced surveillance of severe S. pyogenes disease from January 1, 2003, through December 31, 2004. Cases were defined according to the US definition (S. pyogenes isolated from a sterile site) (4). Also included were cases in which S. pyogenes was isolated from a nonsterile site in combination with probable streptococcal toxic shock syndrome (STSS) or another severe manifestation (pneumonia, necrotizing fasciitis, puerperal sepsis, meningitis, or septic arthritis). STSS was defined according to US specifications that differentiate between confirmed and probable cases on the basis of recovery of a sterile or nonsterile site isolate, respectively (4).

To maximize case ascertainment, cases were identified from 2 sources: isolate referrals to the national reference laboratory (Streptococcus and Diphtheria Reference Unit [SDRU]) and surveillance reports made to the Communicable Disease Surveillance Centre (CDSC) (5). SDRU provides reference microbiology services to the United Kingdom; CDSC surveillance covers England, Wales, Northern Ireland, the Channel Islands, and the Isle of Man. Cases identified from each source were reconciled with each other by using automated techniques to match records on the basis of personal identifiers (date of birth, sex, hospital number, National Health Service number, SOUNDEX-coded surname) and geographic location, followed by loose matching and manual checking to allow records to differ slightly on any given matching parameter. Referring laboratories were sent a study questionnaire to obtain further information on the demographic profile of the patient, disease manifestations, markers of clinical severity, outcome, and possible sources of infection (6). Ethnicity of patients was sought and classified according to census groupings for rate calculation. Where an isolate had not been received by SDRU, this isolate was also requested.

Group A streptococcal isolates referred to SDRU were characterized according to their M protein by using conventional serologic and emm gene typing (7,8). Antimicrobial drug–susceptibility testing was conducted by referral laboratories according to local standard operating procedures.

Responses to completed questionnaires were entered and stored in a custom-made Access (Microsoft, Redmond, WA, USA) database. All reports were checked to ensure they met the case definition. Repeat episodes were defined as those occurring in the same patient >30 days after the initial episode; reports received within 30 days were considered part of the same episode. Data were extracted for statistical analysis into STATA statistical software release 8.2 (Stata Corporation, College Station, TX, USA). Descriptive statistics were undertaken on confirmed cases with χ² and t tests used to test statistical significance of differences between subgroups. Incidence rates were calculated by using midyear resident population estimates for the respective years, age groups, sexes, and regional populations, with exact 95% confidence intervals (CIs) calculated according to the Poisson distribution. The 2001 census data were used as denominators for calculating rates according to ethnic group. All denominators were obtained from the Office for National Statistics. Stepwise unconditional logistic regression analysis was conducted to examine the independence of explanatory variables and development of STSS; the likelihood ratio test was used to evaluate significance of explanatory variables within each model.

All analyses were made on data from the United Kingdom, Channel Islands, and Isle of Man, except for estimated rates of infection, which were calculated for the areas with dual reporting (England, Wales, Northern Ireland, Channel Islands, and Isle of Man). The last 2 areas were omitted for age-, sex-, and ethnicity-specific rate calculations because of unavailability of these population denominators.

As part of a wider European initiative to improve our understanding of the epidemiology of severe S. pyogenes infections, the United Kingdom has amassed one of the largest collections of such cases recorded. The >3,700 cases diagnosed in 2003–2004 resulted in a rate of 3.33/100,000 population for England, Wales, Northern Ireland, the Channel Islands, and Isle of Man. This rate was similar to rates reported for other European countries and the United States in the early 2000s (1,9,10), although lower than some estimates from Canada (11–13). In the 2-year study period, the overall rate of severe S. pyogenes infections decreased from 3.53/100,000 to 3.12/100,000. This overall decrease was largely caused by a substantial decrease in 2 regions, Yorkshire and Humber, and London. Use of multiple sources for case ascertainment was an improvement over previously used methods. Previous methods, which relied solely on voluntary laboratory reporting, would have yielded a rate of 2.65/100,000 during this period. As with any study dependent on participation of local reporters, this study may have missed additional diagnosed cases.

Rates of severe diseases associated with S. pyogenes were markedly higher in male patients than in female patients, an observation not consistently found in other countries (12,14) but commonly found among patients with bacteremic infections in the United Kingdom (15,16). We did not observe any increased rate of severe S. pyogenes infections in black patients of African or Caribbean origin, as was found in a large study in the United States (17). Estimated rates were lower for black patients than for Asian or white patients. Because our study relied on clinician and microbiologist reporting of ethnicity, the proportion of patients reported as white may have been overestimated as a result of assumptions made without confirmatory information.

Marked seasonal patterns in severe S. pyogenes disease were evident during the study period, with an initial peak in December–January, followed by a strong peak in March–April. Preliminary comparisons among Strep-EURO participants suggest similar patterns in other European countries (18). Why these diseases should peak in late winter and early spring is not known. Seasonal patterns of viral respiratory infections with respiratory syncytial virus and influenza virus, which could make patients vulnerable to invasive S. pyogenes infections, may play a role in early- to mid-winter S. pyogenes peaks but would not explain the main spring peak seen in the United Kingdom (Health Protection Agency, unpub. data).

Clinical information provided for these case-patients highlights the severity of these infections; 19% died within 7 days of the initial culture-positive specimen being obtained. This finding is consistent with overall case-fatality estimates during enhanced surveillance in the United Kingdom in 1994–1997 (25%) and estimates reported in other countries (14,17,19,20). However, only 1 of these studies defined a time frame for estimates or included the role of infection in the patient’s death, as our study did. Case-fatality rates were particularly high for case-patients with necrotizing fasciitis, who accounted for only 5% of all cases but 10% of all deaths. Completion of the questionnaire could also have been biased in favor of more severe or interesting cases.

Among case-patients identified in this study, 12% had pneumonia, a value substantially higher than that noted by enhanced surveillance in the United Kingdom in 1994–1997 (5%) (20). However, our finding was consistent with those in studies in other countries (11,13,17,19,21,22). The case-fatality rate in our study (32% within 7 days of initial diagnosis) was substantially higher than that expected with community-acquired pneumonia (23).

STSS developed in 8% of the case-patients identified in our study. These patients had poor survival rates; 45% died within a week of initial diagnosis. STSS was most likely to develop in young adults, which is consistent with findings of a US study that reported a lower median age for STSS patients (24). Infection with an emm/M3 strain was associated with an increased risk for STSS. This finding is consistent with previous (unadjusted) findings from the United States (17) but different from findings from Canada, which only identified emm/M9, a strain uncommon in the United Kingdom, as associated with STSS (13). Alcoholism was associated with increased risk for STSS; this association was also found in other studies (13,22). However, risk for STSS was 3-fold greater for patients who were reported to have used nonsteroidal antiinflammatory drugs, despite adjustment for whether patients had necrotizing fasciitis, which is strongly associated with STSS. Because no data were collected about time, dose, indications for use, or which agent was taken, a causal link between use of nonsteroidal antiinflammatory drugs and STSS cannot be inferred from our findings. A confounding factor, such as delay in receiving appropriate treatment, which we did not adjust for in our analysis, could explain this finding. Patients who took nonsteroidal antiinflammatory drugs may have had early signs, such as extreme pain, which indicated a more severe infection. Nonetheless, this remains an interesting and potentially important observation in a controversial area where evidence supporting either immunologic impairment or augmentation caused by nonsteroidal antiinflammatory drugs remains unresolved (25). Other novel associations with STSS were made; e.g., STSS was less likely to develop in injection drug users or patients with malignancies than in other patients. These findings point to immunocompetence as a necessary mediator for development of STSS.

Questionnaires concerning possible predisposing factors have highlighted skin lesions as the most commonly identified potential source of infection, which is similar to findings in other countries (18,22). However, given the occult nature of S. pyogenes infections, as indicated by the high proportion (21%) of cases with no identified focus, several of these cases may have originated from respiratory tract colonization. This colonization could lead to transient bacteremia, which in turn seeded local tissue sites, possibly in the presence of local trauma.

In contrast to preliminary findings from other Strep-EURO participants (18), a substantial proportion of case-patients were injection drug users (20% overall and 40% among young adults). Regional breakdowns of risk factors were not undertaken in our analyses. However, regional differences in report rates are in part explainable by injection drug use–related cases, with highest overall rates in the Yorkshire and Humber region. This region had been identified as having a particular problem of severe S. pyogenes disease in injection drug users (26). Patterns of isolate referrals to the national reference laboratory over the past decade suggest an increase in severe S. pyogenes infection in injection drug users (27). The reasons behind this change remain unclear and warrant further investigation.

Most severe S. pyogenes cases in our study occurred sporadically in the community; only 9% were associated with healthcare interventions. One fourth of all case-patients and nearly 1 in 6 young adults had no particular risk factors identified. These findings highlight the likely economic effect of these infections and the challenges in developing any effective prevention measure.

Analysis of information collected in this study has yielded some unique insights into these infections and has begun to provide an evidence base for mounting public health initiatives in the United Kingdom (28). High and rapidly ensuing mortality rates among these patients emphasize the need for early recognition and rapid treatment. Maintaining a high index of suspicion, especially where there are signs of possible necrotizing fasciitis, could clearly be life-saving. Further analysis of emm/M type distribution, a key virulence factor inducing immunologic memory, will assist in assessment of the potential effect of vaccines currently under development. Changes in the epidemiology of severe S. pyogenes infections since the last period of enhanced surveillance in the mid-1990s underline the need for periodic monitoring to detect changes in disease manifestations, risk groups, and microbiologic characteristics to develop strategies for control and management of these infections.