Eighteen-Year Longitudinal Study of Uncomplicated and Complex Acute Otitis Media During the Pneumococcal Conjugate Vaccine Era, 2006–2023

Abstract

Background

We analyzed the demographic and risk factors, middle ear fluid (MEF) pathogens, pneumococcus serotype distribution, and bacterial antibiotic nonsusceptibility among children with uncomplicated acute otitis media (uAOM) and complex acute otitis media (cAOM) over 3 timeframes: 2006–2009 (7-valent pneumococcal conjugate vaccine [PCV7] era), 2010–2014 (early 13-valent pneumococcal conjugate vaccine [PCV13] era), and 2015–2023 (late PCV13 era).

Methods

A total of 1537 children were enrolled over 18 years and prospectively followed from 6 to 36 months of age. Upon diagnosis of AOM, tympanocentesis was performed for MEF collection and culture. Electronic medical records were analyzed to identify uAOM and cAOM episodes.

Results

Analysis of demographic data showed that male sex, family history of AOM, and daycare attendance increased the odds of developing cAOM compared to uAOM. Streptococcus pneumoniae was less likely in cAOM, and Haemophilus influenzae more likely as compared to uAOM. AOM caused by S pneumoniae decreased significantly in the early and late PCV13 eras. This was driven by decreases in cAOM caused by PCV13 S pneumoniae strains, especially serotype 19A. Streptococcus pneumoniae penicillin nonsusceptibility was associated with cAOM and declined in the early PCV13 era.

Conclusions

The risk factors for developing cAOM compared to uAOM are similar. PCV13 significantly reduced cAOM and penicillin nonsusceptibility associated with S pneumoniae, driven by reduction in cases caused by serotype 19A. Haemophilus influenzae continued to be a dominant cause of cAOM. Although non-PCV13 S pneumoniae serotypes emerged in the late PCV13 era, the lower level of cAOM caused by S pneumoniae was sustained.

The risk factors for developing complex acute otitis media (AOM) compared to uncomplicated AOM are similar. PCV13 reduced complex AOM and penicillin nonsusceptibility associated with Streptococcus pneumoniae. Haemophilus influenzae continued to be a dominant cause of complex AOM.

Acute otitis media (AOM) predominantly affects children between 6 and 24 months of age. AOM is usually treated with antibiotics in the United States (US), although elsewhere, mild to moderate cases are treated supportively with a high rate of spontaneous resolution [1]. Most infections resolve without complications (uncomplicated AOM [uAOM]) [2]. Some children have more serious disease, defined in this study as complex AOM (cAOM). Children with cAOM have become a focus of concern by public health and regulatory authorities because they experience greater morbidity from infections and their treatment accounts for more than half of medical care costs associated with AOM [3]. After introduction of pneumococcal conjugate vaccines (PCVs), changes in bacterial pathogen distribution and their antibiotic susceptibility were observed among children with AOM [4–8]. Knowledge gaps persist as no single long-term, prospective comparative study of both types of infection in the same population has been published. Most studies do not differentiate uAOM and cAOM, and many rely on nasopharyngeal cultures rather than tympanocentesis. Nasopharyngeal culture is a poor predictor of middle ear bacteriology [9].

Here, we report an 18-year longitudinal tympanocentesis study from a single US site that involved 6- to 36-month-old children with uAOM, cAOM, or both. We sought to understand demographics and risk factors, otopathogen composition, pneumococcal serotype distribution, and bacterial antibiotic nonsusceptibility over 3 different timeframes related to the recommendation of PCVs in the US: 2006–2009 (7-valent PCV [PCV7] era), 2010–2014 (early 13-valent PCV [PCV13] era), and 2015–2023 (late PCV13 era).

METHODS

Study Design

Young children participated in a prospective, longitudinal study of AOM between January 2006 and August 2023 in Rochester, New York. Children were generally enrolled at age 6 months and followed until 36 months. The details of the study design have been previously described [4]. All children were required to receive the full primary series of PCV7 or PCV13 immunizations according to US Centers for Disease Control and Prevention recommendations (doses at 2, 4, and 6, months; booster dose between 12 and 15 months). Written informed consent was obtained from parents, and the institutional review boards of the University of Rochester or Rochester Regional Health approved the study.

Sample Collection

Children were enrolled from community-based pediatric practices, including members of all social classes living in urban, suburban, and rural communities representative of the diverse populations in the US. When a child was diagnosed with AOM in the clinic according to the diagnostic criteria of the American Academy of Pediatrics [10], otoscopy findings were confirmed by tympanocentesis culture in most cases to identify middle ear fluid (MEF) otopathogens. For patients with a perforated tympanic membrane (TM), a deep aspiration of the MEF or swab sample to obtain MEF was collected within 24 hours.

Pathogen Culture and Antibiotic Susceptibility

Standard microbiology processing and identification techniques were used for the isolation of Streptococcus pneumoniae, Haemophilus influenzae, Moraxella catarrhalis, and Streptococcus pyogenes. Pneumococcal serotypes were determined using pure cultures by Quellung reaction using Latex pools and serotype-specific antisera (Serum Staten Institut, Denmark). Oxacillin nonsusceptibility for S pneumoniae and β-lactamase production for H influenzae and M catarrhalis isolates were determined using Sensi-Disc (Becton Dickinson). There were no cases of AOM caused by S pyogenes. Antibiotic susceptibility of S pneumoniae isolates to 16 antibiotics was determined with the VITEK 2 Gram-Positive Susceptibility Card-AST-GP68, AST-GP74 and AST-ST02 (bioMérieux, Inc). The antibiotic susceptibility interpretations were classified as susceptible, intermediate, or resistant based on current Clinical and Laboratory Standards Institute breakpoints for nonmeningeal isolates [11].

Data Collection and Definition of uAOM and cAOM

Demographic and AOM risk factor data were obtained by parent interview and/or from the electronic medical record. Each episode of AOM was retrospectively reclassified as uAOM or cAOM. uAOM was defined as infection of the middle ear with signs and symptoms (fever, irritability, ear tugging, headache) for ≤72 hours. cAOM was defined as presentation of 1 of the following 5 conditions: (i) recurrent AOM (defined as children with 3 AOM episodes within 6 months or ≥4 AOM episodes within 12 months); (ii) AOM with TM rupture; (iii) relapsed AOM (second AOM episode occurring >2 weeks from the initial AOM visit but <1 month from the initial otitis media [OM] episode) [12]; (iv) treatment failure (second AOM episode occurring within 14 days from the initial AOM visit) [12]; or (v) AOM with local or systemic complications such as mastoiditis, intracranial abscess, and facial nerve palsy. Classification of cAOM was made on an episode basis. If the child met the definition of recurrent AOM, all AOM episodes with MEF collection were included in cAOM. Infrequently, when parents brought the enrolled child to an urgent care facility, AOM diagnosis was confirmed the following day by a study clinician and MEF was collected. There were AOM episodes where reexamination did not occur; these were counted as AOM episodes and were taken into consideration for the uAOM/cAOM definition. For analysis of AOM risk factors, children who participated in the study and had never experienced an AOM were included in certain analyses.

Statistical Analysis

Categorical variables were summarized as proportions and frequencies. Fisher exact test or the χ² test was used to understand differences in these variables.

All MEF sample results for a specific episode of AOM were combined. The presence or absence of otopathogens isolated per AOM episode was used as the response in logistic regression with the era used as an exposure variable and the type of episode (uAOM or cAOM), demographic/risk factors, clinical site, and age as covariates. Model selection was performed using stepwise elimination to minimize the Akaike information criteria. The relationship of the variables with cAOM was studied using logistic regression with cAOM as the response. Children without AOM were also evaluated to compare demographics and risk factors against children with AOM.

The presence or absence of antimicrobial sensitivity was modeled in a similar fashion with sensitivity as the response.

Results were presented as adjusted odds ratios (ORs) and 95% confidence intervals (CIs). Unadjusted ORs were calculated using the oddsratio.wald function in the R package epitools (version 0.5-10.1). All models were fit using the glm function in R software (version 4.4.1).

RESULTS

Children in the Cohort

A total of 1537 children between 6 and 36 months of age were enrolled on a rolling basis across the 18-year study time frame (Supplementary Table 1A); 1121 children were enrolled at about 6 months of age and constitute the primary analysis cohort (Table 1). The remainder were actively enrolled by parent request at a later age. Some parents who were offered enrollment when their child was 6 months old declined but later would request tympanocentesis for symptomatic relief of pain or because of antibiotic treatment failure, TM rupture, or recurrent AOM. These children were enrolled and followed prospectively thereafter. Due to their higher number and/or severity of AOM episodes, these children were enrolled as cAOM cases. There was no effort to recruit a particular type of child with AOM. One hundred ninety-two children vaccinated with PCV7 were included in the PCV7 era. Children who received PCV13 immunizations were divided into 2 eras: 404 children in the early PCV13 era, and 525 children in the late PCV13 era. Five hundred ninety-one never had an AOM episode (No OM group, 52.7%) and 530 children had at least 1 AOM episode. These 530 children were further divided into uAOM (n = 281 [53.0%]), cAOM (n = 181 [34.2%]), and cAOM and uAOM (n = 68 [12.8%]).

Table 1.

Demographic and Clinical Characteristics of Children Enrolled at Approximately 6 Months of Age

Characteristic	Overall Total	PCV7 Era Jan 2006–Dec 2009				Early PCV13 Era Jan 2010–Dec 2014				Late PCV13 Era Jan 2015–Aug 2023
		n = 192				n = 404				n = 525
		No AOMa	uAOMb	cAOMc	uAOM & cAOMd	No AOMa	uAOMb	cAOMc	uAOM & cAOMd	No AOMa	uAOMb	cAOMc	uAOM & cAOMd
	N = 1121	n = 96	n = 45	n = 40	n = 11	n = 262	n = 90	n = 37	n = 15	n = 233	n = 146	n = 104	n = 42
Sex
Male	593	49	25	26	5	124	51	24	9	119	70	66	25
	52.9%	51%	55.6%	65%	45.5%	47.3%	56.7%	64.9%	60%	51.1%	47.9%	63.5%	59.5%
Female	528	47	20	14	6	138	39	13	6	114	76	38	17
	47.1%	49%	44.4%	35%	54.5%	52.7%	43.3%	35.1%	40%	48.9%	52.1%	36.5%	40.5%
Ethnicity
Non-Latino	1018	93	43	39	11	225	84	37	15	205	132	94	40
	90.8%	96.9%	95.6%	97.5%	100%	85.9%	93.3%	100%	100%	88%	90.4%	90.4%	95.2%
Latino	103	3	2	1	0	37	6	0	0	28	14	10	2
	9.2%	3.1%	4.4%	2.5%	0%	14.1%	6.7%	0%	0%	12%	9.6%	9.6%	4.8%
Race
White	855	81	39	37	10	159	74	32	14	170	119	85	35
	76.3%	84.4%	86.7%	92.5%	90.9%	60.7%	82.2%	86.5%	93.3%	73%	81.6%	81.7%	83.3%
Black	117	9	3	0	1	59	6	3	0	23	6	5	2
	10.4%	9.4%	6.7%	0%	9.1%	22.5%	6.7%	8.1%	0%	9.9%	4.1%	4.8%	4.8%
Multiracial	76	2	2	2	0	18	3	1	1	23	12	9	3
	6.8%	2.1%	4.4%	5%	0%	6.9%	3.3%	2.7%	6.7%	9.9%	8.2%	8.7%	7.1%
Other	14	1	0	0	0	2	2	0	0	4	3	1	1
	1.2%	1%	0%	0%	0%	0.8%	2.2%	0%	0%	1.7%	2.1%	1%	2.4%
Not documented	59	3	1	1	0	24	5	1	0	13	6	4	1
	5.3%	3.1%	2.2%	2.5%	0%	9.2%	5.6%	2.7%	0%	5.6%	4.1%	3.8%	2.4%
Breastfeedinge
Exclusively	423	28	12	4	2	52	29	8	4	119	78	63	24
	37.7%	29.2%	26.7%	10%	18.2%	19.8%	32.2%	21.6%	26.7%	51.1%	53.4%	60.6%	57.1%
Partially	251	23	7	16	1	48	13	8	3	67	34	21	10
	22.4%	24%	15.6%	40%	9.1%	18.3%	14.4%	21.6%	20%	28.8%	23.3%	20.2%	23.8%
None	423	45	26	9	8	159	41	19	8	46	34	20	8
	37.7%	46.9%	57.8%	22.5%	72.7%	60.7%	45.6%	51.4%	53.3%	19.7%	23.3%	19.2%	19%
Not documented	24	0	0	11	0	3	7	2	0	1	0	0	0
	2.1%	0%	0%	27.5%	0%	1.1%	7.8%	5.4%	0%	0.4%	0%	0%	0%
Smoker
Yes	183	6	5	1	1	52	8	1	1	56	32	15	5
	16.3%	6.2%	11.1%	2.5%	9.1%	19.8%	8.9%	2.7%	6.7%	24%	21.9%	14.4%	11.9%
No	893	90	40	36	10	208	81	36	14	166	100	79	33
	79.7%	93.8%	88.9%	90%	90.9%	79.4%	90%	97.3%	93.3%	71.2%	68.5%	76%	78.6%
Not documented	45	0	0	3	0	2	1	0	0	11	14	10	4
	4%	0%	0%	7.5%	0%	0.8%	1.1%	0%	0%	4.7%	9.6%	9.6%	9.5%
Family history of AOMf
Yes	561	43	27	24	5	85	40	23	10	127	84	70	23
	50%	44.8%	60%	60%	45.5%	32.4%	44.4%	62.2%	66.7%	54.5%	57.5%	67.3%	54.8%
No	543	53	18	15	6	171	46	14	4	103	60	34	19
	48.4%	55.2%	40%	37.5%	54.5%	65.3%	51.1%	37.8%	26.7%	44.2%	41.1%	32.7%	45.2%
Not documented	17	0	0	1	0	6	4	0	1	3	2	0	0
	1.5%	0%	0%	2.5%	0%	2.3%	4.4%	0%	6.7%	1.3%	1.4%	0%	0%
Daycare attendanceg
Yes	539	31	17	28	8	108	48	29	12	75	79	82	22
	48.1%	32.3%	37.8%	70%	72.7%	41.2%	53.3%	78.4%	80%	32.2%	54.1%	78.8%	52.4%
No	568	65	28	11	3	153	42	8	3	154	63	21	17
	50.7%	67.7%	62.2%	27.5%	27.3%	58.4%	46.7%	21.6%	20%	66.1%	43.2%	20.2%	40.5%
Not documented	14	0	0	1	0	1	0	0	0	4	4	1	3
	1.2%	0%	0%	2.5%	0%	0.4%	0%	0%	0%	1.7%	2.7%	1%	7.1%

Data are presented as No. (%).

Abbreviations: AOM, acute otitis media; cAOM, complex acute otitis media; PCV7, 7-valent pneumococcal conjugate vaccine; PCV13, 13-valent pneumococcal conjugate vaccine; uAOM, uncomplicated acute otitis media.

^aNo AOM: children who never had AOM.

^buAOM: children who had uAOM.

^ccAOM: children who had at least 1 episode of cAOM (recurrent AOM, AOM with tympanic membrane rupture, AOM with antibiotic failure, relapse AOM, AOM with local or systemic complications such as mastoiditis).

^duAOM & cAOM: children who had at least 1 episode of both uAOM and cAOM. Individual episodes of AOM are classified as uAOM or cAOM. A child can therefore have both types of AOM and are included in this group.

^eBreastfeeding exclusively: breastfeeding only to 12 months of age; breastfeeding partially: breast and formula feeding.

^fFamily history of AOM is defined as immediate family members having a history of AOM infections.

^gDaycare attendance was defined as daycare attendance at any time point during the study.

Surgical interventions were inconsistently recorded over the eras. On the order of 7.6% of children had tympanostomy tube placement, 2% had adenoidectomy, and 1.5% had tonsillectomy. The dropout rate was 15.8%.

Demographic Features and Risk Factors for uAOM and cAOM

Male sex, family history of AOM, and daycare attendance increased the odds of cAOM compared to uAOM (Figure 1). Data on demographics and clinical characteristics were also compared between children with and without AOM episodes (No OM group). Daycare attendance was the highest risk factor for having any AOM (Supplementary Figure 1A and 1B).

Figure 1.

Demographic and risk factors for complex acute otitis media (cAOM) as compared to uncomplicated acute otitis media among children enrolled at approximately 6 months of age. Log-adjusted odds ratios (ORs) and 95% confidence intervals (CIs) are shown. Male sex (OR, 1.85 [95% CI, 1.21–2.83]), daycare attendance (OR, 3.77 [95% CI, 2.41–6.03]), and a family history of ear infections (OR, 1.79 [95% CI, 1.17–2.77]) were associated with an increased risk of cAOM.

uAOM and cAOM Episodes in the 3 Study Eras

Among 530 children with ≥1 AOM, 380 (71.7%) gave consent at the AOM visit and underwent tympanocentesis, yielding MEF samples from 802 AOM episodes (193 children [51%] had multiple episodes with MEF samples). The data in Table 1 suggest that, over the 3 time periods, 50.0%, 64.9%, and 44.4% of the children had no AOM, respectively, that is, consistent with a vaccine impact in the early years that faded, perhaps with pneumococcal replacement. In a similar vein, the percentage of children with cAOM was 20.8%, 9.2%, and 19.8%, again suggesting a similar pattern.

Due to the random staffing and logistic challenges or parent refusal at specific episodes of AOM, 150 children with AOM did not undergo tympanocentesis. Although there were demographic differences between the children who did and did not have tympanocentesis (Supplementary Table 2A and 2B), these are unlikely to affect the differences seen between uAOM and cAOM.

Supplementary Table 3 A shows the number of uAOM and cAOM episodes with MEF collection across time. Of 802 AOM episodes, 38.9% (312) were classified as uAOM. Among 490 cAOM episodes, 57.8% (283) were recurrent AOM, 18.6% (91) were relapsed AOM, 4.7% (23) were AOM with rupture, and 19.0% (93) were AOM treatment failures.

Distribution of Otopathogens Before and After the Introduction of PCV13

The percentage of children with pneumococci isolated during AOM episodes decreased over time (Table 2, Supplementary Table 4A). Compared to the PCV7 era, the odds of isolating pneumococci in the early PCV13 era trended lower and were 43% lower in the late PCV13 era. Including opportunistically enrolled patients, the odds of isolating pneumococci in the early PCV13 era were 40% lower (OR, 0.60 [95% CI, .41–.87]). No difference was seen comparing the early and late PCV13 eras. There was no significant relationship between the isolation of H influenzae or M catarrhalis and era.

Table 2.

Changes in Otopathogen and Serotype Distribution Over the Eras in Children Enrolled at Approximately 6 Months of Age

Pathogen and Serotype Distribution	Early PCV13 Era Jan 2010–Dec 2014	Late PCV13 Era Jan 2015–Aug 2023
Streptococcus pneumoniae	0.65 (.41–1.02)	0.57 (.36–.89)
S pneumoniae uAOM	ns	ns
S pneumoniae cAOM	0.42 (.22–.78)	0.44 (.25–.80)
Haemophilus influenzae	ns	ns
H influenzae uAOM	ns	ns
H influenzae cAOM	ns	ns
Moraxella catarrhalis	ns	ns
M catarrhalis uAOM	ns	0.32 (.13–.81)
M catarrhalis cAOM	ns	ns
6 serotypes in PCV13 but not PCV7	0.28 (.12–.63)	0.077 (.012–.30)
PCV13 serotypes	0.25 (.10–.56)	0.11 (.037–.30)
Serotype 19A	0.23 (.08–.60)	0.029 (.003–.14)
7 serotypes in PCV20 but not PCV13	18.7 (3.71–342.1)	12.32 (2.36–226.97)
Non-PCV serotypes	5.89 (2.25–18.53)	18.51 (6.91–59.69)

Data are presented as odds ratios (95% confidence intervals) for otopathogen isolation and serotype expression over the eras. The PCV7 era serves as reference.

Abbreviations: cAOM, complex acute otitis media; ns, not significant; PCV7, 7-valent pneumococcal conjugate vaccine; PCV13, 13-valent pneumococcal conjugate vaccine; PCV20, 20-valent pneumococcal conjugate vaccine; uAOM, uncomplicated acute otitis media.

The difference in otopathogen distribution over time was driven primarily by changes in the frequency of pneumococci in cAOM cases (Figure 2). In children with cAOM, compared to the PCV7 era, pneumococcal isolation was 58% lower in the early PCV13 era, and 56% lower in the late PCV13 era (Table 2). No difference was seen in children with uAOM. There was no relationship between era and H influenzae positivity in children with cAOM or uAOM. There were lower odds of M catarrhalis isolation in children with uAOM in the late PCV13 era.

Figure 2.

Interaction between acute otitis media (AOM) type and otopathogen positivity over time among children enrolled at approximately 6 months of age. Proportion (number of AOM episodes associated with each otopathogen/total uncomplicated AOM or complex AOM episodes) of otopathogens detected. Error bars are 95% confidence intervals (unadjusted). *Significant difference compared to the 7-valent pneumococcal conjugate vaccine era after the adjustment for demographic and risk factors. Abbreviations: cAOM, complex acute otitis media; HF, Haemophilus influenzae; MC, Moraxella catarhalis; PCV7, 7-valent pneumococcal conjugate vaccine; PCV13, 13-valent pneumococcal conjugate vaccine; PN, Streptococcus pneumoniae; uAOM, uncomplicated acute otitis media.

Relationship Between Otopathogens and cAOM

Compared to uAOM, cAOM showed 35% reduced pneumococcal isolation (OR, 0.65 [95% CI, .46–.92]) and 44% increased H influenzae isolation (OR, 1.44 [95% CI, 1.05–2.00]). No relationship was found with M catarrhalis.

Isolation of a single otopathogen in an AOM episode was linked to a higher likelihood of uAOM than cAOM, although it was not statistically significant. Neither culture-negative results nor detection of multiple otopathogens were associated with cAOM (Supplementary Table 5A).

Since the cAOM dataset contains cases of treatment failure after antibiotic exposure and repeated antibiotic treatments, we reanalyzed the data excluding treatment failures (Supplementary Table 5A and 5B) or culture-negative episodes. The findings were consistent with the analysis that included these episodes.

Distribution of Pneumococcal Serotypes Before and After the Introduction of PCV13

The proportion of pneumococci expressing serotypes included in PCV13 significantly decreased over time (Table 2, Figures 3 and 4, and Supplementary Table 6). Compared to the PCV7 era, the odds of isolating pneumococci expressing the 6 serotypes in PCV13 but not in PCV7 were 72% lower for the early PCV13 era and 92.3% lower for the late PCV13 era. The odds of isolating pneumococci expressing the 13 serotypes included in PCV13 were 75% lower in the early PCV13 era and 89% lower in the late PCV13 era. These changes were mainly driven by pneumococci expressing serotype 19A, the odds of which were 77% lower in the early PCV13 era and 97.1% lower in the late PCV13 era. The isolation of pneumococci expressing serotype 3 did not change significantly over time.

Figure 3.

Interaction between acute otitis media (AOM) type and serotype over time among children enrolled at approximately 6 months of age. Proportion (number of AOM episodes associated with each serotype class/total uncomplicated or complex AOM episodes) of serotype class detected. Error bars are 95% confidence intervals (unadjusted). Abbreviations: cAOM, complex acute otitis media; PCV, pneumococcal conjugate vaccine; PCV7, 7-valent pneumococcal conjugate vaccine; PCV13, 13-valent pneumococcal conjugate vaccine; PCV20, 20-valent pneumococcal conjugate vaccine; uAOM, uncomplicated acute otitis media.

Figure 4.

Serotype distributions over time between uncomplicated acute otitis media (uAOM) and complex acute otitis media (cAOM) among all participants. Proportion (number of AOM episodes associated with each serotype/total uAOM or cAOM episodes) of each serotype. Data from all participants are shown due to the small number of individual serotypes. Only determined serotypes are shown. *P < .05, Fisher exact test. Abbreviations: cAOM, complex acute otitis media; NT, nontypeable; PCV7, 7-valent pneumococcal conjugate vaccine; PCV13, 13-valent pneumococcal conjugate vaccine; PCV15, 15-valent pneumococcal conjugate vaccine; PCV20, 20-valent pneumococcal conjugate vaccine; uAOM, uncomplicated acute otitis media.

PCV13 serotypes were replaced by non-PCV13 serotypes over time. Compared to the PCV7 era, the OR for MEF isolation of pneumococci expressing the 7 serotypes included in 20-valent PCV (PCV20) but not in PCV13 was 18.7 for the early PCV13 era and 12.32 for the late PCV13 era. Non-PCV serotypes were seen more frequently in the early PCV13 era (OR, 5.89) and late PCV13 era (OR, 18.51). For all children, serotype 35B isolation drastically increased in the late PCV13 era compared to the early PCV13 era (OR, 15.75 [95% CI, 2.87–223.33]). For children enrolled at approximately 6 months, serotype 35B only appeared in the late PCV13 era.

Relationship Between Pneumococcal Serotype and cAOM

Due to the rarity of pneumococcal PCV13 serotypes in the late PCV13 period, this era was excluded from this analysis. Pneumococci with the 6 serotypes in PCV13 but not in PCV7, the PCV13 serotypes, and serotype 19A were more likely seen in cAOM (OR, 3.33 [95% CI, 1.22–9.9521], 3.43 [95% CI, 1.27–10.15], and 6.78 [95% CI, 2.07–26.88], respectively) versus uAOM episodes.

Antibiotic Nonsusceptibility Over Time and Relationship to cAOM

The odds of antibiotic nonsusceptibility of pneumococci to penicillin were higher in children with cAOM versus uAOM (OR, 2.65 [95% CI, 1.23–5.98]), as was multiclass nonsusceptibility (OR, 2.84 [95% CI, .95–9.24]), but this was not statistically significant (Table 3).

Table 3.

Streptococcus pneumoniae Antibiotic Nonsusceptibility and Associated Serotype Distribution—All Participants

Antibiotic Tested (No. of Readouts)	PCV7 Era				Early PCV13 Era				Late PCV13 Era
	Jan 2006–Dec 2009				Jan 2010–Dec 2014				Jan 2015–Aug 2023
	No. of Nonsusceptible Isolates/No. of Readouts (%)a		Associated Serotypes (No.)		No. of Nonsusceptible Isolates/No. of Readouts (%)		Associated Serotypes (No.)		No. of Nonsusceptible Isolates/No. of Readouts (%)		Associated Serotypes (No.)
	uAOM	cAOM	uAOM	cAOM	uAOM	cAOM	uAOM	cAOM	uAOM	cAOM	uAOM	cAOM
Penicillin (150)	3/9 (33%)	21/34 (62%)	19A: 2 ND: 1	6A: 3 6C: 1 19A: 14 ND: 3	8/33 (24%)	11/31 (35%)	3: 1 6A: 1 15A: 1 19A: 2 23A: 1 23B: 1 ND: 1	11A: 1 19A: 9 35B: 1	3/12 (25%)	14/31 (45%)	23B: 1 23F: 1 35B: 1	15A: 1 15B: 2 15C: 1 35B: 8 35D: 2
Amoxicillin (138)	0/8 (0%)	0/29 (0%)	…	…	0/31 (0%)	0/26 (0%)	…	…	1/11 (9%)	7/33 (21%)	23B: 1	15B: 1 35B: 4 35D: 1 NT: 1
Cefotaxime (153)	1/9 (11%)	10/34 (29%)	19A: 1	6A: 2 19A: 7 ND: 1	3/33 (9%)	4/31 (13%)	3: 1 19A: 2	19A: 4	1/12 (8%)	4/34 (12%)	35B: 1	15B: 1 15C: 1 35B: 2
Ceftriaxone (157)	1/9 (11%)	2/34 (6%)	19A: 1	19A: 2	1/33 (3%)	0/31 (0%)	3: 1	…	0/13 (0%)	1/37 (3%)	…	35B: 1
Ertapenem (151)	0/9 (11%)	0/33 (21%)	…	…	0/33 (9%)	0/31 (19%)	…	…	0/11 (0%)	2/34 (6%)	…	35B: 2
Meropenem (152)	1/9 (11%)	7/34 (21%)	19A: 1	6A: 1 19A: 5 ND: 1	3/33 (9%)	6/31 (19%)	3: 1 19A: 2	19A: 5 35B: 1	1/12 (9%)	16/33* (47%)	35B: 1	15B: 2 15C: 1 35B: 11 35D: 1 NT: 1
Ofloxacin (151)	0/9 (0%)	0/34 (0%)	…	…	0/33 (0%)	0/31 (0%)	…	…	1/11 (9%)	0/33 (0%)	11C: 1	…
Erythromycin (91)	4/4 (100%)	17/17 (100%)	19A: 2 ND: 2	6A: 3 6C: 1 19A: 10 ND: 3	8/12 (67%)	8/11 (73%)	11C: 1 15A: 1 15B: 1 15C: 1 19A: 2 23A: 1 33F: 1	11A: 1 15B: 2 19A: 5	4/13 (31%)	17/34 (50%)	11A: 1 15B: 1 21: 1 35B: 1	10: 1 15A: 1 15B: 2 15C: 1 23B: 2 35B: 8 35D: 2
Tetracycline (153)	2/9 (22%)	9/34 (26%)	19A: 1 ND: 1	6A: 1 19A: 8	4/33 (12%)	5/31 (16%)	3: 1 15A: 1 19A: 2	19A: 5	0/12 (0%)	3/34 (9%)	…	15A: 1 15B: 1 35B: 1
Chloramphenicol (150)	0/9 (0%)	1/34 (3%)	…	6A: 1	0/32 (0%)	0/31 (0%)	…	…	0/11 (0%)	0/33 (0%)	…	…
TMP-SMX (154)	2/9 (22%)	15/34 (44%)	19A: 2	6A: 2 6C: 2 19A: 9 ND: 2	8/33 (24%)	9/31 (29%)	3: 1 6A: 1 11C: 1 19A: 2 23B: 1 33F: 1 ND: 1	11A: 2 19A: 7	0/12 (0%)	11/35 (31%)	…	15B: 2 15C: 1 23B: 3 23F: 1 33A: 1 35B: 2 NT: 1
Multiclass antibiotic nonsusceptibilityb (157)	3/9 (33%)	15/34 (44%)	19A: 2 ND: 1	6A: 3 6C: 1 19A: 9 ND: 2	4/33 (12%)	6/31 (19%)	3: 1 15A: 1 19A: 2	11A: 1 19A: 5	1/13 (8%)	13/37 (35%)	35B: 1	15A: 1 15B: 2 15C: 1 35B: 7 35D: 1 NT: 1

Of 277 Streptococcus pneumoniae isolates, 157 (57%) were randomly selected and tested for antibiotic nonsusceptibility. All participants are shown due to the small number of individual serotypes.

Amoxicillin, cefotaxime, ceftriaxone, chloramphenicol, ertapenem, linezolid, moxifloxacin, penicillin, tetracycline, vancomycin, levofloxacin, meropenem, ofloxacin, telithromycin, and TMP-SMX were tested. Only the 9 antibiotics with detected nonsusceptibility are shown. Intermediate and resistant sensitivities were categorized as nonsusceptible.

Only determined serotypes were included. Bolded serotypes are present in PCV13. If nontypeable is the only result for an episode, it is listed as NT.

Abbreviations: cAOM, complex acute otitis media; ND, indeterminate (determined only up to subgroup); NT, nontypeable; PCV7, 7-valent pneumococcal conjugate vaccine; PCV13, 13-valent pneumococcal conjugate vaccine; TMP-SMX, trimethoprim-sulfamethoxazole; uAOM, uncomplicated acute otitis media.

^aPercentage of isolates expressing the particular serotype that are nonsusceptible.

^bMulticlass antibiotic nonsusceptibility is defined as in Bajema et al [13].

^*Fisher exact test was used to compare uAOM and cAOM (P < .05).

Isolation of penicillin-nonsusceptible pneumococci was 67% lower in the early PCV13 era versus the PCV7 era (OR, 0.33 [95% CI, .12–.92]), and multiclass nonsusceptibility was lower over the eras, although the only statistically significant decrease was in the early PCV13 era (OR, 0.23 [95% CI, .065–.76] vs the PCV7 era). This decrease in nonsusceptibility during the early PCV13 era coincided with a decrease in serotype 19A isolates, which have a significant association with penicillin nonsusceptibility (OR, 14.57 [95% CI, 4.51–53.51]) and multiclass nonsusceptibility (OR, 13.55 [95% CI, 3.48–64.33]).

There was no relationship between S pneumoniae oxacillin resistance and era after adjustment for covariates. There was a relationship between cAOM and oxacillin resistance (OR, 2.60 [95% CI, 1.31–5.42]) (Supplementary Table 7A and 7B). There were greater odds of β-lactamase production by H influenzae in the early and late PCV13 eras (OR, 3.07 [95% CI, 1.30–7.81] and 2.64 [95% CI, 1.10–6.81], respectively). There was no relationship between β-lactamase production by H influenzae and cAOM. All M catarrhalis isolates tested produced β-lactamase.

DISCUSSION

We conducted an 18-year longitudinal study in Rochester, New York, to investigate the epidemiology of cAOM and uAOM in 6- to 36-month-old children using tympanocentesis and microbiological culture. Analysis of demographic data showed that male sex, family history of AOM, and daycare attendance increased the odds of developing cAOM compared to uAOM. For children enrolled at about 6 months, isolation of pneumococcus was less likely, and isolation of H influenzae more likely, in cAOM as compared to uAOM. AOM caused by pneumococci trended lower in the early PCV13 era and was significantly lower in the late PCV13 era. This was driven by decreases in cAOM caused by PCV13 pneumococcal strains, especially strains expressing serotype 19A. Pneumococcal penicillin nonsusceptibility was associated with cAOM and declined in the early PCV13 era. A key strength of our study is the comparison of uAOM and cAOM within the same study population simultaneously. These findings are likely generalizable to primary care practices, where households are less dense and where childcare commonly supports workforce participation outside of the home.

Demographic Risk Factors

Male sex was associated with an 85% increased risk, family history of AOM with a 79% increased risk, and daycare attendance with a 277% increased risk of cAOM over uAOM, consistent with other epidemiological studies of recurrent AOM [5, 14]. Although our cAOM population includes other complex episodes such as TM rupture, antibiotic treatment failure, and relapsed AOM, most episodes in this study were recurrent AOM infections. The risk factors we find for cAOM are clinically plausible. Males are generally more susceptible to the infections [15], family history of AOM may be a surrogate for genetic contribution to AOM susceptibility [16], and higher infectious disease risk is associated with attendance of childcare [17].

S pneumoniae Over Time

After the introduction of PCV13, the odds of isolating pneumococci during an episode of AOM declined by 43% in the late PCV13 era. A decline of 35% was seen in the early PCV13 era, although this was not significant in the primary cohort analysis group. Using the entire cohort of children, the odds of isolating pneumococci declined by 40%, which was significant. Over the 13 years following the introduction of PCV13, there was a 72% reduction in overall AOM caused by pneumococci expressing the 6 additional serotypes included in PCV13 but not PCV7 in the early PCV13 era, and a 92% reduction in the late PCV13 era. Similarly, there was a 75% reduction in episodes caused by the PCV13 serotypes in the early PCV13 era and an 89% reduction in the late PCV13 era. This decline was driven by a decrease in cAOM caused by pneumococci. PCV13 covered the serotypes associated with cAOM and led to a remarkable decrease in cAOM. Much of this change was driven by a decrease in pneumococci expressing serotype 19A. Isolation of pneumococci expressing serotype 3 from MEF was less common, contrary to its prevalence in invasive pneumococcal disease [18]. A similar strong reduction among different populations experiencing cAOM caused by PCV13 serotypes was reported in multiple studies [8, 19, 20]. Previously we have shown that children with episodes of recurrent AOM are immunologically impaired [21]. The strong decrease in cAOM caused by pneumococci suggests that PCV13 was able to prevent cAOM in those high-risk children through direct or herd protection.

Additional serotypes included in PCV20 and non-PCV serotypes were significantly increased in both PCV13 eras, suggesting that serotype replacement occurred after introduction of PCV13, but not enough to return pneumococcal isolation to pre-PCV13 levels. Previous studies that included cAOM showed no significant increase in non-PCV serotypes [8, 22]. However, those studies only included the early PCV13 era. Although we did not find a significant association of pneumococci expressing serotype 35B and cAOM, serotype 35B strains have become a predominant otopathogen in our study population [23].

H influenzae and M catarrhalis Over Time

We did not observe a relationship between era and isolation of H influenzae or M catarrhalis in cAOM. Haemophilus influenzae was the dominant otopathogen in the early and late PCV13 eras. Dagan et al reported an unexpected significant reduction in nonpneumococcal OM in their study [24]. They also reported higher vaccine impact in the crowded Bedouin population compared to the Jewish population, possibly due to differences in lifestyle and resulting otopathogen exposure.

Microbiological Risk Factor for cAOM

In our study, initiated after the introduction of PCV7, isolation of pneumococci was 35% less likely in children with cAOM versus uAOM. Prior studies conducted in the pre-PCV era identified pneumococcus as the predominant otopathogen, giving evidence of the benefit of PCV7 [25]. Isolation of H influenzae was 44% more likely in children with cAOM versus uAOM. This is consistent with the association between H influenzae and cAOM, particularly recurrent AOM, as previously reported [26].

Our analysis detected multiple otopathogens in 7%–11% of episodes, even among those with cAOM; however, isolation of multiple otopathogens and failure to isolate any otopathogen were not associated with cAOM. Dagan et al reported that 35% of complex OM episodes (including recurrent, nonresponsive, treatment failure, or chronic) involved multiple otopathogens, especially in Bedouins (64.2%), possibly due to their living conditions [27]. Multiple otopathogen detection (21%) was reported among Aboriginal indigenous children with acute perforation [28]. The difference among studies suggests that multiple pathogen isolation may be dependent on the lifestyle and population. Culture-negative episodes accounted for approximately 36% of uAOM and 41% of cAOM cases in our population. We speculate that children with cAOM may have more exposure to surreptitious use of leftover antibiotics, resulting in a greater proportion of negative culture results. Excluding treatment failure and culture-negative cases did not change the result of our analysis comparing otopathogen isolation and risk of cAOM.

Penicillin nonsusceptibility of pneumococci was 165% more likely to be associated with cAOM. Nonsusceptibility to penicillin provides a survival advantage for pneumococci after exposure to penicillin, which is typically the case for cAOM episodes.

Antibiotic Nonsusceptibility Over Time

We found a 67% decrease in penicillin nonsusceptibility and a 77% decrease in multiclass antibiotic nonsusceptibility for pneumococci in the early PCV13 era versus the PCV7 era, predominantly due to reduction of serotype 19A. Ben-Shimol et al reported a 70% decrease in penicillin nonsusceptibility and a 55% decrease in multiclass antibiotic nonsusceptibility in 2014–2016 compared to 2004–2008 [29], consistent with our findings. Our group reported that antibiotic nonsusceptibility of pneumococci significantly increased from 2014 onward, largely due to increased prevalence of serotype 35B [30].

The main limitation of this study is our inability to estimate the true incidence of uAOM and cAOM, as our study design did not capture appropriate denominator data.

CONCLUSIONS

Risk factors for developing cAOM compared to uAOM are similar. PCV13 significantly reduced cAOM and penicillin nonsusceptibility associated with pneumococci driven by serotype 19A. Haemophilus influenzae continued to be a dominant cause of cAOM. Although non-PCV13 pneumococcal serotypes emerged in the late PCV13 era, the lower level of cAOM caused by pneumococci was sustained.

Supplementary Data

Supplementary materials are available at The Journal of Infectious Diseases online (http://jid.oxfordjournals.org/). Supplementary materials consist of data provided by the author that are published to benefit the reader. The posted materials are not copyedited. The contents of all supplementary data are the sole responsibility of the authors. Questions or messages regarding errors should be addressed to the author.

Notes

Acknowledgments. We thank all physicians, nurses and clinical staff, the parents who consented, and children who participated in this study. We thank all of the laboratory technicians at M. P.'s laboratory who performed bacterial identifications and antibiotic susceptibility testing on children's samples.

Author contributions. N. F.: Writing–original draft, writing–review and editing, data curation, formal analysis, investigation, methodology, validation, visualization. F. N. S.: Writing–original draft, writing–review and editing, formal analysis, investigation, methodology, validation, visualization. R. K.: Writing–review and editing, investigation. P. B.: Writing–review and editing, conceptualization, data curation, formal analysis, investigation, methodology, supervision. E. G.: Writing–review and editing, data curation, investigation. L. W.: Writing–review and editing, conceptualization, formal analysis. M. A.: Writing–review and editing, project administration. A.M.: Writing, review and editing, project administration. L. R. G.: Writing–review and editing, conceptualization, funding acquisition, methodology, resources. A. A.: Writing–review and editing, conceptualization, funding acquisition, investigation, methodology, project administration, supervision. M. E. P.: Writing–original draft, writing–review and editing, conceptualization, funding acquisition, investigation, methodology, project administration, resources, supervision, visualization. All authors had full access to all the data in the study and accept responsibility to submit for publication.

Data sharing. The original data contributions presented in the study are included in the article or the Supplementary Material. Further inquiries can be directed to the corresponding author.

Financial support. Rochester General Hospital Research Institute is the study sponsor/co-funder and Pfizer provided additional funding for this study analysis. Funding was provided by the National Institute on Deafness and Other Communication Disorders of the National Institutes of Health (grant number R0108671) and the Centers for Disease Control and Prevention (contract numbers 75D30119C06842 and 75D30121C12195) to M. P. for sample collection.

Potential conflicts of interest. The institutions employing N. F., F. N. S., R. K., E. G., P. B., and M. P. received funds from Pfizer to conduct the study. F. N. S. is compensated for contributions to Practical Reviews in Otolaryngology. L. W., M. A., A. M., L. R. G., and A. A. are employees of and hold stock and/or stock options in Pfizer, Inc.

All authors have submitted the ICMJE Form for Disclosure of Potential Conflicts of Interest. Conflicts that the editors consider relevant to the content of the manuscript have been disclosed.