A total of 18,646 IPD patients were reported to the GERMS-SA surveillance program during 2012–2018 (Figure 1). Of these, 6,865 (36.8%) patients were enrolled in the GERMS-SA enhanced surveillance program and 6,247 (91.0%) of them had available outcome information. Of patients with outcome information, 6,228 (99.97%) had age information. Samples from 5,103 (81.9%) with known age were serotyped; of those, 4,906 (96.1%) yielded a valid serotype by Quellung or were positive for one of the 38 serotypes included in the multiplex PCR. Of those with valid serotypes, 4,656 (94.9%) had distinguishable serotype data. Of the distinguishable serotype cases, 4,272 (91.8%) cases had serotypes included in the PCV-13 vaccine or serotypes that comprised >1% of all cases. The cases used in this study made up 22.9% of overall IPD cases (4,272/18,646). The most represented age group was 25–44 years (40.7%) (Table 1), which was similar to that of all cases, for which the most represented age group was also 25–44 years (Appendix 2 Table 1). Of those with known sex, 51.2% were female, was similar to the distribution of sex of all cases (51.2% female of those with known sex) (Appendix 2 Table 1). Of those with known HIV status, 67.0% were HIV positive (Table 2; Appendix 1 Figure 2).

When we considered all IPD cases and all ages, the most common serotype identified was serotype 8 (11.9%), followed by 19A (8.9%), 12F (8.8%), and 3 (6.8%) (Table 2; Figure 2, panel A). This serotype distribution was similar to that of all 18,646 IPD cases from all GERMS-SA sites, in which the most common serotype identified was also serotype 8 (7.2%), followed by 12F (6.1%), 19A (6.0%), and 3 (4.2%) (Appendix 1 Figure 3; Appendix 2 Table 1). When considering all IPD cases and all ages, we found serotype 1, 6A, 19A, and 12F to decrease in total number of isolates over time period of study (Figure 2, panel B). In patients <15 years of age, the most common serotype was serotype 8 (14.0%), followed by serotype 1 (7.3%), 19A (6.7%), and 19F (6.3%) (Table 2; Figure 2, panel A). Serotypes 12F, 3, 1, and 6A consistently decreased over time of study in this age group (Appendix 1 Figure 4). In patients >15 years of age, the most common serotype was also serotype 8 (11.1%), followed by 12F (9.7%), 19A (9.6%), and 3 (7.9%) (Table 2; Figure 2, panel A). In this age group, we found serotype 1 to consistently decrease over the time period of the study (Appendix 1 Figure 5). The overall CFR was 32.1% when considering all IPD cases. In patients <15 years of age, CFR was 22.1%, and in those >15 years of age, CFR was 35.5% (Tables 1, 2). Serotypes with the highest CFR in all IPD cases were 6A (44.6%), 11A (43.8%), and 22F (40.5%). In IPD patients <15 years, the serotypes with the highest CFR were serotype 11A (64.3%), 22F (50.0%), and 13 (40.0%) (Figure 3). In IPD patients >15 years of age, serotypes with the highest CFR were 5 (50.0%), 6A (48.3%), 15B/C (47.5%), and 19F (43.1%) (Figure 4). In nmIPD patients of all ages, the most common serotype was 19A, followed by 8 and 3 (Appendix 1 Figure 6). The serotype with the highest CFR in nmIPD in all ages was serotype 11A (44.4%), followed by 10A (36.9%) and 6A and 19F (both 35.4%) (Appendix 2 Table 2).

When we analyzed only meningitis cases but considered all ages, we found that serotype 6A (61.9%) had the highest CFR, followed by 16F (59.0%) and 19A (58.0%). Serotype 6A also had the highest CFR when restricted to meningitis patients <15 years of age (71.4%) (Figure 5). In meningitis patients >15 years of age, serotype 1 had the highest CFR (62.6%) (Figure 6).

Nine years after PCV7 introduction and 5 years after PCV13 introduction in South Africa, the non-PCV13 serotype 8 is the most common serotype found in IPD patients followed by 19A, 12F, and 3. Serotype 1, a PCV13 serotype decreasing in total number over the study period, was associated with higher CFR than serotype 8 in analysis of meningitis cases including all ages. Serotype 6A was associated with higher CFR in the analysis of IPD for all ages, in particular in IPD cases of patients <15 years of age. In patients >15 years of age, we found that serotype 6A decreased in 2012 and 2013 but then stabilized until study end. Serogroup 15 non-PCV13 serotypes 15A and 15B/C were also associated with higher CFR than serotype 8. Other factors that we found were independently associated with death in >2 analyses were HIV status, age, meningitis infection, year, and province.

Serotypes associated with death in the prevaccine era (2003–2008) in South Africa in patients >15 years of age included serotypes 1 and 19F in IPD patients but no particular serotype in meningitis patients (19). Serotype 1, which is a PCV13 but not a PCV7 serotype, was associated with death in our analysis of meningitis cases of all ages in the vaccine era. Serotype 1 is associated with outbreaks of meningitis (20,21). In studies from mainly prevaccine years (1990–2009) (22,23), and particularly in a meta-analysis including studies from 1928–2010 (24), the risk for death from serotype 1 was found to be lower than that for other serotypes. In the prevaccine era, serotype 1 has been associated with a high invasive disease potential (25,26) (studies from France, Alaska, and Iceland) and noted to be one of the most common IPD serotypes globally in children <5 years of age (27). These studies cover different time periods and geographic locations than ours; distinct serotype 1 lineages circulating in South Africa, as described in du Plessis et al. (28), could explain the differences we found in association with death. Similar to a previous study of serotype 1 epidemiology in South Africa in 2003–2013 (29), we notice a continued decline of overall numbers of serotype 1 IPD cases in patients >15 years of age. The characteristics of serotype 1 pneumococci lineages and our results suggest that even though numbers are decreasing overall, this serotype should be monitored locally and globally in the future and the association of specific lineages with death and with meningitis should be analyzed.

Vaccines continue to play a crucial role in protection against pneumococcal disease. On the basis of our results in South Africa, we recommend protecting against the PCV13 serotypes 3, 6A, 19F, and 19A, as well as serotype 1, and monitoring the non-PCV13 serotype 12F for its effects on pneumococcal disease. Serotype 6A in particular was associated with death in analysis of IPD of all ages and of IPD in patients <15 years of age. Serotype 19F was associated with death in IPD patients in the vaccine era as well as in the prevaccine era (19); a component of both the PCV13 and PCV7 vaccines, serotype 19F seems to be associated with more severe disease both in our study death and in previous studies (14,24,30). Serotype 19A was the second most common serotype in IPD in both the prevaccine and the vaccine era; it is included in the PCV13 but not in the PCV7 vaccine and was noted to be one of the most important replacement serotypes in the United States in children <5 years of age (31). Serotype 3 (a PCV13 but not PCV7 serotype) and serotype 12F (a non-PCV serotype) were both among the 4 most common serotypes overall in our IPD study population; the total number of deaths was ≈100 for each. Although we did not find either serotype 3 or 12F independently associated with death in our analysis, the total numbers lead us to suggest monitoring these serotypes in the future.

Although serotype 1 was the most common serotype in the prevaccine era in South Africa (19), we found serotype 8 to be the most common in the current vaccine era. This finding in South Africa is similar to a recent report from Europe, North America, and Australia (32) and also a study summarizing serotypes globally (33). Including this serotype in new vaccine formulas could therefore reduce overall IPD cases. Serogroup 15 should be considered in new vaccine formulations, because we found 15B/C (IPD patients >15 years) and 15A (IPD patients <15 years) to be associated with death. To our knowledge, recent data of postvaccine dynamics of serogroup 15, as well as its association with CFR, are scarce. Savinola et al. (34) described the emergence of a 15A multidrug-resistant lineage as a result from a 19A capsule switch and van der Linden et al. (35) showed an increase in 15A after PCV-13 introduction in Germany. An increase of serogroup 15 in children with pneumococcal disease was noted in Hong Kong; Liyanapathirana et al. suggested consideration in future vaccine strategies (36). Because of these recent studies and the association we found with death, we suggest close monitoring of serogroup 15 in both adults and children.

Our study has limitations that make a direct comparison to other studies difficult. We used data from enhanced surveillance sites for available patient data; doing so excluded cases from smaller and rural hospitals and clinics. We also excluded from analysis cases for which serotype and outcome information were not available. Both limitations may have caused selection bias. Another limitation is that our 2 groups are not mutually exclusive; our analysis of IPD cases includes the meningitis cases, which we also analyzed separately as a subset. The inclusion of meningitis cases in the IPD analysis may be causing some of the serotypes to be associated with CFR. Temporal variation of serotypes after introduction of vaccines was not analyzed; although we considered the increase or decrease of serotypes in the conclusions, the consequences of these variations are hard to anticipate and may have introduced bias. In addition, we did not control for treatment or antimicrobial resistance, which may need to be studied in a future analysis. An advantage of this study is that, in an effort to detect differences in age groups, we analyzed the data in all ages as well as in 2 different age groups of <15 years and >15 years. Another advantage is that we conducted the analysis to detect possible differences of serotype associated with death in IPD cases as well as in a subset of meningitis cases.

In conclusion, our data suggest that PCV13 serotypes 1, 3, 6A, 19A, and 19F should continue to be monitored in surveillance studies. Although we cannot exclude that other serotypes may also be important, we recommend inclusion of serotypes 8 and 12F and serogroup 15 (serotypes 15A and 15B/C) in new vaccines, which may contribute to overall reduction of disease caused by S. pneumoniae.