Systematic review and meta-analysis of pertussis epidemiology in Latin America and the Caribbean: 1980-2015.

OBJECTIVES: In Latin America and the Caribbean (LAC), pertussis disease incidence has reportedly increased since 2000 despite high vaccine coverage. A systematic review of pertussis literature and a meta-analysis was conducted to understand the burden of disease in LAC.
METHODS: A systematic literature review was completed, using relevant search terms. Original articles describing pertussis epidemiology and vaccine coverage in LAC published between 1980 and 2015 were identified. Applying a Bayesian meta-analysis random-effects model, we calculated pooled estimates and corresponding 95% credible intervals (95% CrIs) for pertussis incidence, case fatality ratio (CFR), pertussis prevalence among contacts, and coverage with three doses of diphtheria, tetanus, and pertussis (DTP) vaccine (DTP3).
RESULTS: A total of 59 studies meeting our selection criteria were identified, representing 15 countries. Of the 59, 15 of them provided incidence data, with 7 of the 15 reporting a pertussis case definition. The pertussis incidence estimate for the 1980-1999 period was 17.8 cases per 100 000 persons (95% CrI: 5.9-29.7); for the 2000-2015 period, it was 2.5 cases per 100 000 persons (95% CrI: 1.8-3.2). For the 1980-2015 period, the CFR, in 19 studies reviewed, was 3.9% (95% CrI: 2.9%-4.9%); for that same period, in 5 studies reviewed, pertussis prevalence among contacts was 24.9% (95% CrI: 13.7%-36.1%). Pooled DTP3 vaccine coverage estimates, in a total of 20 studies reviewed for the following three time periods, were: 1980-1990, 72.4% (95% CrI: 64.6%-80.2%); 1991-2000, 79.0% (95% CrI: 66.1%-91.9%); and 2001-2015, 90.0% (95% CrI: 87.7%-92.3%).
CONCLUSION: A decrease in pertussis incidence and an achievement of moderately high DTP3 vaccine coverage since the early 2000s was observed. The review highlights the need for increased publication of pertussis data at the country level and for LAC as a whole in order to better understand the true burden of the disease. Application of a standardized case definition and use of active case finding would aid in obtaining more accurate estimates of the disease burden in LAC.

Globally, pertussis is a poorly controlled disease that causes substantial morbidity and mortality. An estimated 16 million cases of the disease and approximately 195 000 deaths occur worldwide every year (1). After introduction of diphtheria, tetanus, and pertussis (DTP) vaccines during the early 1980s and reported vaccination coverage with three doses of DTP (DTP3) above 90% by the late 1990s, the incidence of pertussis had decreased substantially in Latin America and the Caribbean (LAC) (2–6). More recently, however, pertussis disease incidence has reportedly increased since 2000 (7–10). For example, in Argentina, Brazil, and Colombia, the number of reported cases during 2013 increased by 116%, 582%, and 2 967%, respectively, compared with cases reported during 2000 (11), and outbreaks have been reported in multiple countries during the last decade (6, 12–15). In most LAC countries, pertussis disease is nationally notifiable through a passive surveillance system; however, variation exists among surveillance systems, making comparability of data in LAC a challenge (7).

We conducted a systematic review of literature on pertussis in LAC that was published between 1980 and 2015, in order to more fully understand the burden of disease and vaccine coverage in LAC and to provide historical context as certain LAC countries consider a transition from a whole-cell (wP) to an acellular (aP) primary vaccine series.

METHODS

We conducted a systematic review and meta-analysis of original articles meeting our selection criteria. We systematically searched the following electronic databases: MEDLINE, EMBASE, Cumulative Index to Nursing and Allied Health Literature (CINAHL), Web of Science, Cochrane Library, Latin American and Caribbean Health Sciences Literature (LILACS), Pan American Health Organization (PAHO) Virtual Health Library, World Health Organization Library Database (WHOLIS), Scientific Electronic Library Online (SciELO), and CAB Direct and Global Health. For the search, we used the following disease search terms specifically focused on LAC: whooping cough OR pertus* OR whooping cough OR tos ferina OR tosferina OR coqueluche.

Eligibility criteria

We sought original articles describing pertussis epidemiology and pertussis vaccination coverage published in English, French, Spanish, or Portuguese from January 1980 through December 2015 for all populations in South America, Central America, Mexico, and the Caribbean. To be included, articles were required to describe studies or surveillance that focus on humans as subjects; contain greater than 30 index cases in studies reporting prevalence among contacts; focus on the general population; and be published in a peer-reviewed journal (an exception was made for government reports found through a systematic electronic database review). We excluded review articles, outbreak reports, case reports, position papers, reports concerning special populations (e.g., prisoners), and dissertations. A library science expert assisted with the search. Appendix 1 gives additional details on the search strategy.

Study selection

Three reviewers independently screened all titles and abstracts that the search yielded. Results were reconciled, and irrelevant and duplicate citations were excluded. The reviewers independently reviewed the full text of the remaining available articles to judge their eligibility for inclusion in the review. For all included articles, we abstracted information regarding report characteristics, sample selection methods, sample characteristics, sample participation rates, report findings (for the principal outcomes), and relevant P values or confidence intervals. When denominators were not given, we back-calculated for the denominator using the numerator and the outcome estimates provided. The data were entered into a database using Microsoft Excel software (Microsoft Corporation, Redmond, Washington, United States of America).

Outcomes measures

The principal outcomes included were: pertussis incidence in the general population of all age groups (defined as confirmed cases of pertussis/100 000 persons); pertussis prevalence among contacts (defined as laboratory-confirmed, symptomatic cases/100 evaluated contacts); pertussis case fatality ratio (CFR) (defined as number of deaths/100 confirmed pertussis cases); DPT3 vaccination coverage (defined as the number of subjects aged 6 months through 5 years vaccinated with three DPT vaccine doses/100 eligible subjects assessed); pertussis laboratory confirmation (defined as pertussis cases confirmed with a specific laboratory technique); and pertussis hospitalization (average duration of days hospitalized for a laboratory-confirmed case of pertussis). DPT3 coverage is a widely accepted indicator for assessing routine immunization coverage, and it is frequently used as a proxy for assessing the quality of immunization systems.

Analysis

Applying a Bayesian meta-analysis random-effects model, we calculated pooled estimates and corresponding 95% credible intervals (CrIs), with 20 articles describing pertussis vaccination coverage and with 15 articles describing pertussis incidence rates. A random-effects model assumes the studies are a sample from all possible studies. Thus, this approach takes into account variance both within each study and between studies (16). If an outcome had insufficient data points available for statistical pooling, a descriptive summary of the data was presented instead.

Pertussis vaccination coverage data were stratified into three time periods (1980–1990; 1991–2000; and 2001–2015), and pertussis incidence data were stratified into two time periods (1980–2000 and 2001–2015). In addition to plotting the individual study estimates and CrIs, we calculated pooled random-effects estimates and 95% CrIs for each category. Stratification of the data into the above time periods is based on articles describing data clustered around specific time frames. For vaccination coverage data, we also conducted Bayesian hierarchical logistic regression to determine trend over time. All analyses were implemented by using the SAS PROC MCMC (Markov chain Monte Carlo) procedure. SAS version 9.3 (SAS Institute Incorporated, Cary, North Carolina, United States) was used for the analysis.

RESULTS

Of the 3 980 references identified, 3 513 were duplicates or irrelevant publications, 24 were unavailable to the Centers for Disease Control and Prevention (CDC) library, and 384 were excluded based on inclusion criteria, leaving 59 articles for review (8, 17–74) (Figure 1), representing 15 countries (Table 1). The largest numbers of the included articles were from Brazil (32.2%), Argentina (15.2%), and Colombia (11.9%). Twenty articles (32.2%) reported findings concerning vaccination coverage (20, 21, 24, 26, 27, 30, 35–37, 39, 43, 44, 48, 50, 52, 53, 56–59) (Appendix 2), and 40 (67.7%) reported findings concerning pertussis epidemiology (8, 17–19, 22, 23, 25, 28–34, 38, 40–42, 45–47, 49, 51, 54, 55, 60–74) (Appendix 3). The largest numbers of the articles were published in either Spanish (45.7%) or English (32.2%) (Table 1).

FIGURE 1.

Literature search strategy and selection process

TABLE 1.

Characteristics of studies included in the analysis

Publication Characteristics	N	%	References
Language
Spanish	27	46	(21–25, 28, 30, 33, 34, 36, 38, 40–43, 45, 49, 51, 52, 54–56, 61, 62, 68, 72, 73)
Portuguese	12	20	(19, 27, 35, 39, 44, 47, 50, 57–60, 71)
English	19	32	(8, 17, 18, 20, 26, 29, 31, 32, 46, 48, 53, 63–67, 69, 70, 74)
French	1	2	(37)
Country	59	100
Argentina	9	15	(8, 26, 31, 32, 34, 36, 40, 55, 70)
Brazil	19	32	(17–19, 27, 29, 35, 39, 44, 47, 50, 57–60, 63, 66, 69, 71, 74)
Chile	3	5	(28, 30, 49)
Colombia	7	12	(24, 41–43, 54, 56, 73)
Costa Rica	2	3	(21, 33)
Ecuador	1	2	(20)
Guyana	1	2	(48)
Haiti	1	2	(53)
Honduras	2	3	(38, 52)
Martinique	1	2	(37)
Mexico	5	8	(22, 23, 46, 61, 68)
Panamá	1	2	(45)
Peru	3	5	(64, 65, 67)
Uruguay	3	5	(51, 62, 72)
Venezuela	1	2	(25)

Produced by authors from analysis results.

%

Pertussis epidemiology

Forty publications reported pertussis epidemiologic data from 11 countries: Argentina, Brazil, Chile, Colombia, Costa Rica, Honduras, Mexico, Panama, Peru, Uruguay, and Venezuela (Appendix 3). Fifteen publications reported pertussis incidence outcome data (8, 19, 22, 23, 29, 30, 38, 46, 54, 55, 60, 68, 70, 73, 74). Two publications (55, 68) reported data for two years each (respectively, for 1995 and 2000, and for 2010 and 2011). Another publication (74) reported data for four years, for 2010 through 2013. Although 8 of these 15 articles (53.3%) (8, 19, 29, 30, 46, 55, 72, 74) reported a case definition, only 6 (40.0%) used the WHO pertussis case definition (75). In addition, variation among the case definitions in the remaining articles was noted. For example, the number and type of presenting symptoms and the duration of cough required to meet the case definition differed in each study. Seven of the 8 articles (87.5%) included laboratory confirmation as part of their case definition. Two articles (30, 68) did not specify the type of diagnostic used for laboratory confirmation; in the remaining 5 articles, 1 included only culture for confirmation (29), 3 included both culture and polymerase chain reaction (PCR) (46, 55, 74), and 1 included culture, PCR, and serology (8). By applying a random-effects model that included the 15 publications reporting incidence data, the pooled estimates of annual pertussis incidence in the general population during 1980–1999 and 2000–2015 were, respectively, 17.8 cases per 100 000 persons (95% CrI: 5.9–29.7) and 2.5 cases per 100 000 persons (95% CrI: 1.8–3.2) (Figure 2).

FIGURE 2.

Random-effect estimate of pertussis incidence, 1980–2015

The prevalence of symptomatic pertussis among contacts of laboratory-confirmed pertussis cases was reported in five publications (17, 46, 51, 61, 63). All five focused on pertussis among infants or school children, but they reported symptomatic pertussis among all close contacts, using epidemiologic link or laboratory confirmation with PCR, culture, or serology. Among contacts of confirmed cases evaluated, an estimated 24.9% (95% CrI: 13.7%–36.1%) had laboratory-confirmed, symptomatic pertussis (Figure 3). None of the studies reported whether the contacts had received chemoprophylaxis before symptom onset, or the length of time from case symptom onset to contact specimen collection.

FIGURE 3.

Random-effect estimate of pertussis prevalence among contacts of confirmed cases, 1980–2015

Nineteen publications reported mortality data (8, 28, 29, 31–34, 38, 42, 45, 46, 55, 63, 66, 70-74). CFR was estimated to be 3.9% (95% CrI: 2.9%–4.9%) (Figure 4). Of the 19 publications, 7 of them (36.8%) were hospital-based studies focusing on infants and school-age children, whereas 12 (63.2%) used surveillance data covering infants, adolescents, and adults.

FIGURE 4.

Random-effect estimate of the fatality ratio in cases of pertussis in Latin America, 1980–2015

Twenty-four articles reported laboratory confirmation data (18, 22, 25, 28, 29, 31–34, 40, 41, 45, 47, 49, 51, 61–65, 67, 69, 71, 73). Laboratory confirmation using culture, PCR, and/or direct fluorescent antibody (DFA) testing alone was reported, respectively, in 13 (54.2%), 14 (58.3%), and 3 (12.5%) publications. Of the 24 articles with laboratory confirmation data, 7 studies (29.2%) used PCR and culture as complimentary diagnostic tests (40, 51, 62, 63, 65, 67, 69). Use of PCR as a laboratory confirmation method was reported as early as 1995 in 1 article (49), but it became more common over time, with the majority of publications published after 2006 reporting PCR for pertussis confirmation.

Eight publications, from four countries (Argentina, Brazil, Panama, and Uruguay), reported hospital duration data among patients with confirmed pertussis (31, 32, 34, 40, 45, 51, 71, 72). Six publications reported mean hospital duration, ranging from 5.2 to 14.2 days. Two publications reported a median hospital duration, one of 7.0 days and the other of 9.7 days.

Vaccination coverage data

Pertussis vaccination coverage data were reported in 20 publications (20, 21, 24, 26, 27, 30, 35–37, 39, 43, 44, 48, 50, 52, 53, 56–59), covering 10 countries. The 20 reports included 8 (40.0%) from Brazil, 3 (15.0%) from Colombia, 2 (10.0%) from Argentina, and 1 each from Chile, Costa Rica, Ecuador, Guyana, Haiti, Honduras, and Martinique (Appendix 2). All publications reported coverage with whole cell vaccines (DTP), because acellular vaccines were not used by the countries at the time of the coverage assessment. Sixteen publications (80.0%) estimated vaccination coverage data by using household surveys (20, 24, 26, 27, 35–37, 39, 43, 44, 50, 52, 53, 56–58), 3 publications (15.0%) (30, 48, 59) used the national immunization program data to estimate vaccination coverage, and 1 publication (5.0%) used a school-based survey (21). In 13 of the 20 publications (65.0%), survey samples were selected through systematic or random selection of participants, whereas in 7 studies (35.0%) the entire eligible population was surveyed. The age of the eligible population varied among the publications; however, 65.0% reported an eligible population less than age 5 years. Vaccination status was confirmed by vaccination cards or any written proof of vaccination in 10 of the publications (50.0%), whereas in 8 publications (40.0%), verbal report by guardians or parents was accepted if vaccination cards were unavailable.

DPT3 coverage data were assessed over three time periods, depending on the year the data were collected, as follows: 1980–1990 (20, 24, 35, 36, 39, 48, 52, 56, 58); 1991–2000 (27, 37, 39, 44, 50, 57); and 2001–2015 (21, 26, 30, 53, 59). Pooled estimates of DTP3 vaccination coverage over the time periods of 1980–1990, 1991–2000, and 2001–2015 were, respectively: 72.4% (95% CrI: 64.6%–80.2%), 79.0% (95% CrI: 66.1%–91.9%), and 90.0% (95% CrI: 87.7%–92.3%) (Figure 5). The overall pooled estimate for the entire 1980–2015 time period is 83.7% (95% CrI: 75.1%–91.4%). Furthermore, using a Bayesian hierarchical logistic regression model, we estimated an overall increase in vaccination coverage over time, as shown by the solid red line (with upper and lower CrIs indicated by the dashed red lines) in Figure 6.

FIGURE 5.

Random-effect estimate of DTP3 Vaccine coverage in Latin America, 1980–2015

FIGURE 6.

Trends in Vaccine Coverage, 1980–2015

DISCUSSION

This is the first systematic review of pertussis epidemiology and vaccine coverage in Latin America and the Caribbean, thus contributing to a better understanding of pertussis disease in LAC countries. From our analysis we observed a clear decrease in pertussis incidence during 1980-1999 and 2000–2014, as well as achievement of moderately high DTP3 vaccination coverage since the early 2000s.

The decline in pertussis incidence observed between 1980 and 2000 is consistent with WHO data and likely attributable to the widespread introduction of pertussis vaccines during the 1980s. Because of the limited data available, we were unable to stratify pertussis incidence data by finer time periods and, for example, assess whether any increase in pertussis incidence occurred within the last decade. However, the WHO Strategic Advisory Group of Experts (SAGE) pertussis working group recently reviewed pertussis surveillance data from 19 countries, including 4 from LAC (Brazil, Chile, Cuba, and Mexico), and determined there has been no true resurgence of pertussis in these 4 LAC countries. Instead, increases were attributable to naturally occurring cyclic patterns, surveillance systems changes, or decreases in vaccination coverage (76, 77). It is also important to note that despite recent increases in reported pertussis cases in LAC, the majority of countries are not reporting case counts as high as those reported during the 1980s (11).

Our review highlighted the complexities of comparing pertussis incidence among countries. For example, case definitions for pertussis varied among publications. Consequently, there might have been under-or overestimates of the burden of disease. Among those publications with a case definition, the definitions were inconsistent, and fewer than half of them used the WHO pertussis case definition (75). Creating a standardized pertussis case definition is complicated because of variations in clinical signs and symptoms of pertussis among different age groups (78, 79). However, a standardized, age-specific clinical case definition for LAC would make it easier to compare data between countries in LAC and to estimate the true burden of pertussis there (79).

Similarly, we note the challenges associated with using diagnostic methods with different levels of sensitivity and specificity. Although culture is considered the gold standard for pertussis diagnosis, it has a relatively low sensitivity. Therefore, CDC encourages using complementary diagnostic methods (combination of PCR, culture, and serology) to strengthen a surveillance system’s potential for accurate case confirmation. Of the 24 articles that we reviewed that had laboratory confirmation data, only 14 (58.3%) used culture as a diagnostic method, and only 7 (29.2%) used PCR and culture as complementary diagnostic methods. DFA is not recommended by WHO and CDC as a pertussis laboratory diagnostic because of low sensitivity and specificity (80–82); nevertheless, certain countries continue to use it as a diagnostic method. The majority of the studies we reviewed for laboratory confirmation methods were hospital based; therefore, their diagnostic methods might not be the same as those used in the national surveillance system. Efforts to standardize specimen collection, transportation, and diagnostic techniques and to establish continuous quality control and quality assurance programs will increase case confirmation and improve comparability of laboratory data within and across LAC countries.

Our review indicated an increase in DTP3 vaccination coverage over the three time periods. Our modeled analysis reported an increase in vaccination coverage over time although we observed an overlap in credible intervals among pooled estimates for the three time periods. The majority of the studies included in our review were household surveys and might provide a more accurate estimate of coverage in LAC than administrative data, which often overestimate vaccination coverage. Our 2001–2015 DTP3 coverage estimate met the WHO Global Immunization Vision and Strategy (GIVS) global target of ≥ 90% vaccination coverage (83), but did not meet the PAHO Technical Advisory Group on Vaccine-preventable Diseases target of ≥ 95% coverage among children aged < 1 year (9). To ensure high protection levels among children aged < 5 years and prevent increases in disease burden, LAC countries still need to make efforts to meet and maintain these targets.

The CFR of 3.9% estimated in our analysis was consistent with other estimates of pertussis CFR among lower-income countries (77) but much higher than the United States estimates of 0.15% during a similar time period. The high CFR in our analysis may be an overestimate due to reporting bias if the countries’ surveillance systems were not sensitive enough to identify less severe cases of pertussis. Although we were unable to calculate CFR by different age groups in our review, the majority of reported cases in LAC are infants, and thus this CFR likely represents infant mortality. Infants are at highest risk for severe pertussis disease and complications; therefore, rapid access to quality tertiary care is critical to prevent infant deaths. The LAC CFR might be a reflection of different levels of health care access and quality, especially in rural areas. A high level of suspicion among health care providers and early diagnosis and treatment are important for reducing pertussis mortality.

We estimated a pertussis prevalence rate of 24.9% among close contacts of confirmed pertussis cases. Secondary attack rates among susceptible household contacts of confirmed cases as high as 80% to 90% have been reported in previous studies (84, 85). Early contact investigation to ensure that contacts of confirmed cases receive prophylaxis is key in preventing secondary transmission, especially among households with persons at high risk (e.g., pregnant women, infants, and the immunocompromised) (86). In multiple studies, siblings are frequently the source of infant infection; therefore, vaccination of siblings to prevent primary infection and subsequent household transmission is important (17, 87, 88).

Our review can provide baseline data to evaluate the impact of vaccination changes in LAC on pertussis disease burden. Certain LAC countries are considering a transition from a whole-cell (wP) to an acellular (aP) primary vaccine series. Countries that have already introduced acellular vaccines for the infant doses include Costa Rica and Mexico (7), as well as Panama (personal communication with the pertussis surveillance coordinator of the Panama Ministry of Health). However, data from multiple countries outside Latin America that use an aP primary vaccine series (77, 89–95) indicate the duration of protection offered by aP vaccines wanes over time. In addition, data from modeling studies indicate the transition from wP to aP vaccines might result in a resurgence of disease (96–98). Therefore, a switch to an aP vaccine series at this time might potentiate further increases in pertussis incidence in Latin America. On the basis of available evidence, PAHO’s Technical Advisory Group on Vaccine-preventable Diseases recommends that countries currently using wP vaccines should not switch to aP vaccines (9).

The introduction of tetanus-diphtheria-acellular pertussis (Tdap) booster vaccines for pregnant women in certain Latin American countries (Argentina, Brazil, Chile, Colombia, Costa Rica, Mexico, and Panama) might affect pertussis incidence by decreasing the disease burden among infants too young to be vaccinated (99–102). Data from the United Kingdom estimate high maternal vaccination effectiveness (103, 104). There is currently no PAHO recommendation for routine vaccination of pregnant women. However, the WHO acknowledges that maternal vaccination is likely the most cost-effective strategy for reducing infant disease. WHO therefore suggests that countries experiencing high rates of infant morbidity or mortality from pertussis consider use of maternal vaccination in addition to routine infant vaccination (77).

Limitations

Our results may have been influenced by several limitations. To ensure our review included high-quality data that accurately represented the epidemiology of pertussis in LAC, we applied strict eligibility criteria to our search. However the number of articles meeting our eligibility criteria was limited, and only Argentina, Brazil, Chile, Colombia, Honduras, and Mexico were represented in our review of incidence data. Such paucity and geographic underrepresentation of data meeting our inclusion criteria limit the generalizability of our findings and may have led to over-or underestimates of pertussis incidence. In addition, the limited data prevented us from assessing changes in pertussis incidence within the last decade. Similarly, the heterogeneity of study methods and pertussis case definitions among the eligible articles limited data comparability and generalizability.

Conclusions

Our review highlights the challenge of estimating the overall burden of pertussis in an area with substantial variation in surveillance systems, health care access, and vaccination coverage, both within and across countries. Our findings underline the need for increased publication of pertussis epidemiology data at a country level and within LAC as a whole. Such publications could provide a better understanding of the true burden of pertussis disease in LAC, especially as more countries implement maternal Tdap vaccination and consider transition to an acellular primary series. Use of a standardized case definition, laboratory confirmation with complementary diagnostic methods, and active case finding among contacts of confirmed pertussis cases would also aid in obtaining a more accurate estimate of pertussis disease burden in LAC.

Acknowledgments.

We thank Nong Shang for his statistical advice.

Disclaimer.

Authors hold sole responsibility for the views expressed in the manuscript, which may not necessarily represent the official position or policy of the Centers for Disease Control and Prevention, the RPSP/PAJPH, or PAHO.