INTRODUCTION
Streptococcus pneumoniae is one of the most significant childhood pathogens. A systematic review of disease burden in children aged <5 years estimated that in 2000 there were 14·5 million cases of severe pneumococcal disease, causing 821 000 deaths worldwide [Reference O'Brien1]. In Australia, in 2002, the incidence of invasive pneumococcal disease (IPD) was 11·5/100 000 overall (or 57·3/100 000 in children aged <5 years). Of 2271 cases notified, 44% presented with pneumonia, 5% with meningitis and 35% with unlocalized bacteraemia [Reference Gilbert2, Reference Roche and Krause3].
The 7-valent pneumococcal conjugate vaccine (Prevnar®, Wyeth; PCV7), which contains serotype antigens 4, 6B, 9V, 14, 18C, 19F and 23F, has been available in Australia since 2001, when it was introduced for routine use in Aboriginal children and other high-risk groups. There was limited uptake in the non-Aboriginal population until it was added to the routine immunization schedule in January 2005. Since then, it has been available free of charge for all infants as a three-dose schedule given at ages 2, 4 and 6 months [4]. In October 2011, a 13-valent pneumococcal conjugate vaccine [Prevenar 13® (PCV13), containing the seven PCV7 antigens plus serotype 1, 3, 5, 6A, 7F and 19A antigens] replaced PCV7. Since 1999, all isolates from IPD cases have been referred for serotyping to one of three Pneumococcal Reference Laboratories (PRL) in Australia, including one in New South Wales (NSW) where the population in June 2009 was 7·1 million, or nearly one third of the total Australian population. Based on the number of notified cases of IPD, we estimate that isolates are referred from at least 95% of cases in NSW each year.
Studies in many countries have documented the impact of routine use of PCV7 in infants on S. pneumoniae serotype distribution but few have examined corresponding changes in antibiotic resistance rates and results are varied [Reference Bettinger5–Reference Pirez8]. There is little recent information on pneumococcal resistance in Australia [Reference Gosbell and Neville9, Reference Roche10]. The aim of this study was to compare serotype distributions of Streptococcus pneumoniae isolates from patients aged < 5 and ⩾ 5 years with IPD in NSW, Australia before and after introduction of PCV7 and also to determine rates of antibiotic resistance, in relation to changes in serotype distribution, in IPD isolates from children aged < 5 years old in NSW before and after the introduction of PCV7, and before the introduction of PCV13, into the routine infant immunization schedule.
METHODS
Specimens
Between 2002 and 2009, 5177 S. pneumoniae isolates from patients with IPD were referred to the NSW Pneumococcal Reference Laboratory at the Centre for Infectious Diseases and Microbiology, Westmead Hospital. After exclusion of isolates that were duplicates (n = 102), not viable on receipt (n = 75), not S. pneumoniae (n = 26), non-typable (n = 45) or for which no corresponding patient date of birth was available (n = 43) there were 4886 unique isolates with known serotype and patient age, of these, 1170 (23·5%) were from children aged <5 years. Antibiotic susceptibility testing could not be performed on 166 (14%) of these 1170 isolates because the isolate could not be located from storage (n = 22), was not viable on subculture (n = 43) or, in the early years of surveillance, had been sent as a formalized specimen (n = 101), leaving 1004 isolates from children aged <5 years for susceptibility testing.
Serotypes were grouped as: (a) those included in PCV7; (b) those included in the PCV13, excluding PCV7 serotypes; and (c) non-PCV serotypes.
Serotyping
Pneumococcal isolates were serotyped by Neufeld's Quellung reaction using pool, type and factor-specific antisera (Statens Serum Institut, Copenhagen, Denmark).
Antibiotic susceptibility testing
Commercial broth microdilution minimum inhibitory concentration (MIC) panels (Sensititre, TREK Diagnostic Systems Ltd, UK) were used for quantitative susceptibility testing. This panel includes 18 antibiotics and three blank cells as growth controls. Antibiotics (range of concentrations in mg/l) used in this panel were: azithromycin (0·25–2), amoxicillin/clavulanate (2/1–16/8), cefuroxime (0·5–4), meropenem (0·25–2), erythromycin (0·25–2), chloramphenicol (2–16), cotrimoxazole (0·5/9·5–4/76), vancomycin (0·5–4), telithromycin (0·06–2), ceftriaxone (0·06–2), levofloxacin (0·5–16), clindamycin (0·06–2), cefipime (0·12–8), penicillin (0·03–80), tetracycline (0·5–8), linezolid (0·25–4), moxifloxacin (0·25–8), and gatifloxacin (0·5–8).
MICs corresponding to susceptible and resistant were classified as recommended by the Clinical and Laboratory Standards Institute (CLSI) for all antibiotics except penicillin. The CLSI MIC cut-offs for penicillin-susceptible, -intermediate and -resistant (S/I/R) isolates were changed, in 2008, depending on disease type. However, for consistency with previous data we retained the previous National Committee for Clinical Laboratory Standards (NCCLS) categories namely susceptible (S) ⩽0·06 mg/l, intermediate (I) 0·12–1 mg/l, and resistant ⩾2 mg/l (R).
Statistical analysis
Changes in the distribution of serotypes and antibiotic susceptibility within serotypes between periods were compared using χ2 or Fisher's exact test as appropriate. Because not all isolates with a known serotype could be tested for antibiotic susceptibility, to obtain susceptibility estimates for groups of serotypes we weighted the serotype-specific susceptibility patterns by the distribution of serotypes in the groups. For these adjusted proportions, variance was calculated using methods for a standardized proportion [Reference Kirkwood and Sterne11] and the z approximation to the binomial distribution was used to make comparisons. P values <0·05 were considered statistically significant for all comparisons.
RESULTS
Changes in serotype distribution between time periods
The mean annual numbers and cumulative percentages of the main serotypes in referred isolates, in both age groups and time periods, are shown in Table 1 and Figures 1 and 2. The average numbers of isolates referred per annum fell significantly between 2002–2004 and 2005–2009, due to decreases in PCV7 serotypes, which fell by 86% (from 210 to 29·2) in the <5 years age group and by 54% (from 358 to 164·8) in the ⩾5 years age group (Table 1).
Fig. 1. Mean annual number and cumulative percentages of serotype groups and selected serotypes identified in invasive pneumococcal disease isolates referred to the NSW Pneumococcal Reference Laboratory, from children aged <5 years before and after introduction of 7-valent pneumococcal conjugate vaccine into the routine childhood immunization schedule (in 2005).
Fig. 2. Mean annual number and cumulative percentages of serotype groups and selected serotypes identified in invasive pneumococcal disease isolates referred to the NSW Pneumococcal Reference Laboratory, from children and adults aged ⩾5 years before and after introduction of 7-valent pneumococcal conjugate vaccine into the routine childhood immunization schedule (in 2005).
Table 1. Changes in numbers and proportions of isolates referred in different serotype groups and age groups before and after introduction of the 7-valent pneumococcal conjugate vaccine
* Non-PCV7 group includes all serotypes except those serotypes in the PCV7 vaccine.
The following differences in proportions were statistically significant between time periods: † P < 0·0001, ‡ P = 0·01, § P = 0·001.
This was partly offset by significant increases in the average numbers of isolates referred each year in non-PCV7 serotypes which increased 121% (from 28 to 62) and 49% (from 173·6 to 259·4) in the <5 and ⩾5 years age groups, respectively. This was mainly due to serotype 19A, which rose from 4% to 33% of the total in the <5 years age group and from 4% to 12% in the ⩾5 years age group (Table 1). Although numbers are small, even when the two age groups were combined, there were significant proportional increases between periods in serotypes 1, 3, 6C, 7F, 15B, 22F and 33F (Table 1 and Figs. 1 and 2).
Antibiotic susceptibility
The distribution of 1170 isolates from children aged <5 years and the 1004 isolates available for testing, between serotype groups (PCV7, PCV13 and selected non-PCV serotypes) and time periods is shown in Table 2. Of the 1004 isolates available for testing, 415 (41%) were susceptible to all antibiotics tested. Table 3 shows the numbers and proportions (weighted to represent all referred isolates with known serotype) of PCV7, 19A and other serotypes, which were penicillin intermediate (pen-I; MIC 0·12–1 mg/l) or resistant (pen-R; MIC ⩾2 mg/l) by time period. As expected, there were major differences between serotypes. Pen-R isolates were more prevalent in PCV7, than in non-PCV7 serotypes in both time periods; with the highest resistance rates in 19F and 9V. Pen-R rates did not change significantly between time periods overall, or in either PCV7 or non-PCV7 serotype groups, but there was a significant increase in pen-R serotype 14 only. Pen-I was less common than pen-R in PCV7 isolates but more common in non-PCV7 isolates. The combined rates of pen-I and pen-R in PCV7 isolates did not change, significantly, but increased, in non-PCV7 isolates, from 24% to 34% between time periods, largely due to the increase in proportion of serotype 19A isolates, most of which were pen-I.
Table 2. Numbers and proportions of referred isolates from <5-year-olds tested for antibiotic susceptibility before (2002–2004) and after (2005–2009) the introduction of 7-valent pneumococcal conjugate vaccine, by serotype group
* Other non-PCV serotypes: 23A, 10A, 9N, 35F, 24F, 25A, 35B (<5 isolates each); 12F, 20, 8, 2, 5, 17F, 9A/L, 18A/B/F, 19B/C, 23B, 13, 16F, 21, 33A, 35A, 22A, 7C, 15A (⩽1 isolate each).
Table 3. Weighted numbers and proportions of referred isolates from <5-year-olds that were penicillin-intermediate (pen-I) and penicillin-resistant (pen-R) before (2002–2004) and after (2005–2009) the introduction of the 7-valent pneumococcal conjugate vaccine (PCV7), by serotype group
* Penicillin minimum inhibitory concentration (MIC): I = 0·12–1 mg/l.
† Penicillin MIC: R = ⩾2 mg/l.
‡ Significant differences between time periods were: a increased proportion of penicillin-resistant serotype 14 isolates (P < 0·005); b increased proportion of penicillin-intermediate isolates (P < 0·001).
There was a significant overall decrease in the rate of erythromycin resistance between time periods (Table 4). Of PCV7 isolates, the relatively high resistance rates, especially in serotype 14 isolates, did not change between time periods. In contrast, in non-PCV7 serotypes, an initially low rate of erythromycin resistance in 2002–2004 increased significantly in 2005–2009, due to increases in the overall proportion and erythromycin resistance rate of serotype 19A.
Table 4. Weighted numbers and proportions of referred isolates from <5-year-olds that were erythromycin-resistant before (2002–2004) and after (2005–2009) the introduction of the 7-valent pneumococcal conjugate vaccine (PCV7), by serotype group
* Ery-R = erythromycin resistant; minimum inhibitory concentration >0·05 mg/l.
† Increased erythromycin resistance in non-PCV7 isolates (P = 0·005).
‡ Increased erythromycin resistance in serotype 19A (P = 0·04).
§ Decreased erythromycin resistance rate in all isolates (P = 0·01).
All isolates were susceptible to vancomycin and linezolid, all but one to telithromycin and >99% to gatifloxacin, levofloxacin and moxifloxacin (data not shown). The weighted numbers and proportions of isolates resistant to other antibiotics are shown in Table 5. Two of 1004 isolates tested (0·2%, both serotype 19F) were resistant to all antibiotics tested except vancomycin, linezolid, telithromycin, gatifloxacin, levofloxacin and moxifloxacin. The highest rates of resistance in both periods were to macrolides (azithromycin, erythromycin), cotrimoxazole and penicillin, although macrolide resistance decreased significantly between periods. Ceftriaxone resistance was uncommon, but increased significantly (from 2% to 6%). All but one of the ceftriaxone-resistant isolates were also resistant to penicillin (data not shown). Clindamycin and tetracycline resistance rates also increased (Table 5).
Table 5. Weighted numbers and proportions of isolates from <5-year-olds that were resistant to individual antibiotics before (2002–2004) and after (2005–2009) the introduction of the 7-valent pneumococcal conjugate vaccine (PCV7)
† Penicillin-resistant and penicillin-intermediate (non-susceptible) isolates combined.
‡ Statistically significant differences in resistance rates between 2002–2004 and 2005–2009: aP = 0·002, bP = 0·005, cP = 0·0005, dP = 0·01, eP = 0·0004.
There was also a significant increase in multidrug resistance (MDR) (i.e. resistance to three or more antibiotic classes) between time periods. For example penicillin, erythromycin and cotrimoxazole (Pen/Ery/Cot) MDR increased from 4% to 7% (25/572 vs. 31/432, P = 0·04). In 2002–2004, nearly all Pen/Ery/Cot (21/25, 84%) MDR isolates belonged to serotype 19F compared to only 35% (11/31, P < 0·001) after 2005, when 19F was replaced as the predominant MDR serotype by 19A, which increased from 4% (1/25) to 39% (12/31) of all MDR isolates (P < 0·01). Other serotypes represented in Pen/Ery/Cot MDR isolates were 14 (11%, no change between time periods), 6A, 6B, 9V and 23F (1–2 isolates each).
DISCUSSION
This is the first study in Australia and one of the few anywhere, in which the changes in serotype distribution have been directly correlated with serotype-specific and overall antibiotic susceptibility in a comprehensive collection of IPD isolates, routinely referred for serotyping. In common with previous studies [Reference Dortet12–Reference Williams15], we confirmed a rapid and marked fall in the mean annual numbers and proportions of PCV7 serotypes referred after the introduction of routine childhood immunization, in Australia, particularly in the vaccine target group. This was also apparent to a lesser extent in older age groups due to herd effect. In contrast, there was a significant increase in absolute numbers and proportions of non-PCV7 serotypes in both age groups, to which the main contributor was serotype 19A with significant, but less marked, contributions from 1, 3, 6C, 7F, 15B, 22F and 33F.
Before introduction of PCV7, penicillin resistance was common in several of the most common ‘childhood’ serotypes [Reference Linares14], especially 19F and 9V. As expected, antibiotic resistance of IPD isolates fell in many countries, at least initially, following widespread vaccine use [Reference Dagan and Klugman16, Reference Stephens17], because of the marked fall in the numbers of cases due to PCV7 serotypes. However, this was offset by increases in antibiotic-resistant non-PCV7 serotypes, especially 19A, which spread throughout North America and many other countries [Reference Liao7, Reference Jacobs13, Reference Fenoll18–Reference Reinert21].
There has been little previous information about antibiotic resistance in invasive pneumococci in Australia. A national survey in 2005, of all age groups, showed that 16·5% of 351 invasive isolates were penicillin non-susceptible (MIC ⩾0·12 mg/l), including 5·4% that were pen-R [Reference Gottlieb22]. Our results (24% non-susceptible overall, and 12% pen-R in 2002–2004, in the <5 years age group) are somewhat higher, but the post-PCV7 changes (to 31% and 9%, respectively) are broadly consistent with those reported in international studies of antibiotic resistance following introduction of PCV7. Some studies have shown an initial reduction in penicillin non-susceptibility (pen-I plus pen-R) rates, later followed by increases [Reference Dortet12, Reference Linares14, Reference Fenoll18] and others, no significant change [Reference Bettinger5]. Although penicillin non-susceptibility rates have increased after introduction of PCV7 in many countries, the actual rates vary widely in different regions from 54% to 74% in parts of Africa and Asia to 20–30% in many European countries (MIC >0·06) [Reference Linares14] and 31% Australia (present study).
Most previous studies, like ours, have reported very low rates of resistance to vancomycin, telithromycin, linezolid and gatifloxacin [Reference Yoshioka23] and high rates of macrolide (16–32%) and cotrimoxazole (17–22%) resistance [Reference Borg24–Reference Shibl26]; changes in resistance rates in response to PCV7 use vary [Reference Calbo27, Reference Tyrrell28].
As reported elsewhere [Reference Mantese29], we found that resistance to third-generation cephalosporins, while generally still low, increased after introduction of PCV7. This is of concern, since this class of antibiotics is the mainstay of meningitis treatment and could lead to increased reliance on vancomycin, which is now recommended, in addition to a third-generation cephalosporin, for empirical therapy of suspected pneumococcal meningitis. Fortunately, pneumococci are still universally susceptible to vancomycin [Reference Ahmed30, Reference Tunkel31]. Most cases of IPD, other than meningitis, (including those due to pen-I and pen-R pneumococci) will respond to parenteral penicillin, which should be the initial treatment of choice in community-acquired pneumonia.
Recent increases in serotype 19A prevalence have been reported previously in Australia [Reference Williams15, Reference Hanna32, Reference Lehmann33] and elsewhere [Reference Liao7, Reference Williams15, Reference Hsu34–Reference Techasaensiri37], as have smaller increases in other non-PCV7 serotypes, including 1, 3, 7F, 15, 22 and 33 [Reference Imohl38]. In NSW, serotype 19A isolates had a high pen-I rate, which increased after the introduction of PCV7 and was the major contributor to the unchanged overall levels of penicillin non-susceptibility. Serotype 19A was generally uncommon before the introduction of PCV7. In NSW, it represented only 4% of isolates from <5-year-olds in 2002–2004 but increased to 33% in 2005–2009, with a fivefold increase in number of isolates referred. This is most likely due to a combination of antibiotic use and immune pressure [Reference Choi39].
While serotype 19A has increased in many countries, there have been differences in the dynamics of change. For example, in Canada a high proportion of the increased number of serotype 19A isolates were pen-R and MDR and belonged to clonal complex (CC)271/320. This was attributed to a capsule-switching recombination event (with the predominant pre-vaccine pen-R serotype 19F) and subsequent clonal expansion [Reference Pillai40]. In NSW, the predominant pre-vaccine pen-R serotype, 19F, also mostly belonged to CC271/320 [Reference Xu41]. However, post-vaccine serotype 19A isolates in NSW are mainly pen-I and the increase in their numbers is more likely to be due to expansion of a pre-existing pen-I 19A clone, than to capsule switching. This is consistent with the findings of Hanage et al. [Reference Hanage42], who showed that increases in pen-R non-PCV7 serotypes were due to expansion of pre-existing clones rather than acquisition of de novo resistance or serotype switching. Further work is underway to identify the predominant clonal complexes in 19A isolates in NSW.
In the USA [Reference Beall43], the increase in pen-R serotype 19A began before and continued after the introduction of PCV7 (in 2001), but stabilized in 2005. As in Canada, this was largely due to expansion of pen-R CC271/320, but in the USA it replaced the predominantly pen-I CC199 as the most prevalent 19A CC, suggesting that antibiotic use, rather than vaccine pressure, was the major driver of the increase in 19A. However, a concomitant increase in a putative ‘vaccine escape’, pen-S/I ST695 serotype 19A (a variant of CC199, which switched capsule and adjacent penicillin-binding protein genes with serotype 4) would be better explained by vaccine pressure. Vaccine pressure would also explain the finding, in an infant PCV7 immunization study, that acquisition of serotype 19A was significantly higher in fully immunized (with three doses) than unimmunized children [Reference van Gils44]. Thus the global emergence of serotype 19A appears to have been driven by both antibiotic use and immune pressure (and probably other factors), to varying degrees, and to result from recombination involving capsular and/or antibiotic resistance loci, clonal expansion or both [Reference Lee45].
Our study was limited by the fact that we were unable to test the susceptibility of IPD isolates in all age groups. However, given the major differences in serotype-specific resistance, we believe that serotype distribution is a reasonable predictor of antibiotic resistance. As expected, the changes in serotype distribution in the ⩾5 years age group was qualitatively similar to but less marked than that in the vaccine target age group, reflecting the well-described herd effect. The study demonstrates the complexity of changes in pneumococcal antibiotic resistance. The predicted major decrease in β-lactam antibiotic resistance, following the introduction of PCV7, has not been realized despite the marked fall in overall incidence of IPD. Although PCV13 includes serotype 19A and other serotypes that have increased since the introduction of PCV7, any attempt to predict its impact on future serotype or antibiotic resistance distributions of IPD isolates would be premature. Therefore, continued laboratory surveillance of IPD isolates (at least) is essential to inform future treatment guidelines and vaccine development.
ACKNOWLEDGEMENTS
The antibiotic susceptibility testing performed for this study was supported by a grant from Wyeth (Australia). Routine serotyping of isolates from patients with invasive pneumococcal disease is funded by the Australian Department of Health and Aging. The views expressed in this publication do not necessarily represent the position of the Australian Government. We thank laboratory staff throughout NSW for referring isolates and also Danny Ko and Damla Power for assistance with performing the susceptibility tests. H. Gidding is funded by an NHMRC CRE postdoctoral fellowship (APP1031963).
DECLARATION OF INTEREST
None.
