INTRODUCTION
Over the past 15–20 years, vaccination coverage in Cambodian children has increased markedly, leading to poliovirus (PV) elimination and marked reductions in other vaccine-preventable diseases (VPDs) [1]. During 2000–2010, the percentage of children receiving all the vaccines recommended in the national immunization schedule by age 12 months increased from 31% to 74% [2]. However, significant immunity gaps likely exist in Cambodian adults for many VPDs because of low vaccination coverage in the country before 1993. Recent global experience has shown that large pockets of adult susceptibility can hinder achieving and maintaining VPD elimination and control goals [Reference Luo3, 4]. This paper focuses on polio, rubella, and measles, VPDs that have regional or global eradication, elimination, or control goals.
Cambodia has been identified as one of the countries in the Western Pacific Region (WPR) at increased risk for widespread PV transmission after wild-type poliovirus (WPV) introduction, or emergence of a circulating vaccine-derived poliovirus (cVDPV) [Reference Lowther5] as happened in Cambodia during 2005–2006. WPV outbreaks in Albania, Namibia, Cape Verde, Tajikistan, Republic of Congo, and China [Reference Luo3, 4, Reference Prevots6–Reference Khetsuriani10] have exposed underlying immunity gaps in adults that necessitated wide age-range outbreak response vaccination campaigns and put global polio eradication at risk. Understanding adult PV susceptibility in Cambodia is required for accurate polio risk assessment and preparedness planning.
Like many countries approaching measles elimination, Cambodia has seen a shift in the age distribution of reported cases towards older children, adolescents, and adults. The proportion of laboratory-confirmed measles cases in persons aged ⩾15 years increased from <6% in 2010 to >16% in 2011 after a nationwide campaign targeted at children age <5 years (National Immunization Program, Ministry of Health, Kingdom of Cambodia, unpublished data). Assessment of the impact of this post-campaign shift on overall measles incidence in Cambodia is limited by variable case ascertainment and classification over time and multiple geographically staggered supplementary immunization activities (SIA). How much the age-distribution shift could potentially hinder progress towards measles elimination depends on the extent of underlying measles susceptibility in the adult population, which is currently unknown in Cambodia.
Rubella is currently endemic in Cambodia. In 2013, the Global Alliance for Vaccines and Immunization (GAVI) supported the introduction of rubella-containing vaccine (RCV) into Cambodia via a nationwide campaign of children aged from 9 months to 14 years and in the routine childhood immunization schedule – a major step towards rubella control in the country. However, >25% of reported rubella cases in Cambodia in 2012 were in women of child-bearing age (WCBA). Knowledge of age-specific rubella immunity can generate national estimates of congenital rubella syndrome (CRS) cases and help design a comprehensive vaccination strategy, which includes routine childhood vaccination and protection for WCBA from rubella virus infection [11].
We conducted a nationwide serological survey in 2012 to determine the population immunity to polio, rubella, and measles in Cambodian women aged 15–39 years. The objectives of this survey were to determine age-group-specific seroprevalence of antibodies to WPV types 1–3 and identify potential gaps in PV immunity in adolescent and adult women, generate age-group-specific measles immunity profiles, assess susceptibility to rubella virus infection in WCBA and estimate CRS incidence in Cambodia.
METHODS
During November–December 2012, we conducted a nationwide cross-sectional serological survey of women aged 15–39 years in Cambodia. Cambodia's 24 provinces were divided into five regions: the capital, Phnom Penh, and four geographical regions with approximately equal population sizes (Fig. 1).
Fig. 1. Map of Cambodia illustrating the five regions for the 2012 vaccine-preventable disease serological survey and the number of urban and rural enumeration areas (EA) included the survey for each region. Cambodian provinces by region: West (Battambang, Kampong Chhang, Kampong Speu, Koh Kong, Pailin, Preah Sihanouk, Pursat provinces); North (Banteay Mean Chey, Kampong Thom, Kratie, Mondolkiri, Otdar Mean Chey, Preah Vihear, Ratanakiri, Siem Reap, Steung Treng provinces), Southeast (Kampong Cham, Prey Veng, Svay Rieng provinces); and Southwest (Kampot, Kandal, Kep, Takeo provinces).
Sample size and sample selection
A target sample size of 400 women within each age group (15–19, 20–24, 25–29, 30–34, 35–39 years) was calculated based on an estimated rubella seroprevalence of 80% [Reference Miyakawa12–Reference Malakmadze14], a desired precision of ±5% with 95% probability of achieving that precision, a design effect of 1·5, and a 10% non-response rate based on results from previous surveys.
The 2010 Demographic and Health Survey (DHS) selected 611 enumeration areas (EAs) from the 28 764 EAs enumerated in the 2008 Cambodia General Population Census by probability proportional to size (PPS); our sampling frame consisted of these EAs. In each region 20 EAs were selected by PPS. The DHS survey oversampled urban EAs; we corrected this by allocating the number of selected urban and rural EAs within each region based on the proportion of the population living in urban and rural areas.
During the DHS listing, teams drew maps of the selected EAs delineating boundaries and identifying all households (HHs). For our survey, teams used these maps and updated the HH listings in the field to reflect any changes since the 2010 DHS. In each EA, 22 HHs were selected by systematic sampling using previously described survey methods [Reference Luman15]. Although we aimed to enrol 20 women per cluster, 22 HHs were visited to adjust for the estimated number of age-eligible women per HH and non-participation rate. All eligible women in selected HHs were invited to participate, and teams revisited HHs up to three times.
Data and specimen collection and testing
Field teams visited selected HHs and enrolled eligible women after providing written information about the survey and obtaining consent from participants. A brief questionnaire, including demographic information and tetanus vaccine history, was completed and 5 ml of blood was obtained by venepuncture. Serum was stored at 4–8 °C and transported to the National Institute of Public Health (NIPH) laboratory in Phnom Penh within 96 h of collection. In Phnom Penh, samples were centrifuged and aliquoted into two cryovials, one for measles and rubella antibody testing at NIPH and one for polio antibody testing at CDC (Atlanta, GA, USA). Cryovials were stored at −80 °C until tested at NIPH or shipped by air on dry ice to CDC. Measles and rubella testing was performed by anti-measles and anti-rubella virus IgG enzyme-linked immunosorbent assay (Enzygost, Siemens, Germany) according to the manufacturer's recommendations. Samples with corrected optical density (OD) values >0·2 were considered positive, samples with values <0·1 negative, and samples with values of 0·1–0·2 were considered equivocal and retested.
Samples were tested using a standard microneutralization assay for antibodies to PV types 1–3 according to established protocols at the Global Polio Specialized Laboratory, CDC [16, 17]. Briefly, 80–100 CCID50 of each PV serotype and twofold serial serum dilutions were combined and pre-incubated at 35 °C for 3 h before adding HEp-2(C) cells. After incubation for 5 days at 35 °C and 5% CO2, plates were stained with Crystal Violet and cell viability measured by OD via spectrophotometry. Each specimen was run in triplicate, with parallel specimens from each participant tested in the same assay run. Neutralization titres were estimated by the Spearman–Kärber method [Reference Kärber18] and reported as the reciprocal of the calculated 50% endpoint. Each run contained multiple replicates of a reference antiserum pool starting at 1:32 dilution to monitor performance variation. A serum sample was considered positive if antibodies were present at ⩾1:8 dilution.
Data management and analysis
Statistical analysis was conducted using SAS v. 9·3 (SAS Institute Inc., USA). Sampling weights were calculated to take into account each stage of selection, including the sampling probability in the original DHS EA selection. A non-response adjustment within each of the 12 original design strata was included using the weighting class approach. Post-stratification methods were used to adjust weights to conform to the population distribution of the 2008 Cambodian Population Census. Seroprevalence estimates and 95% confidence intervals (CIs) were calculated taking into account the stratified cluster sampling design and sampling weights. Point estimates <0·20 or >0·80 were reported with Wilson CIs, otherwise Wald CIs were reported. Second-order Rao–Scott χ 2 tests were used to assess differences in seroprevalence across age groups, regions, and rural/urban residence.
For estimates of CRS incidence and annual number of CRS cases in Cambodia, age-structured catalytic models were fitted as described previously [Reference Cutts and Vynnycky19] to the observed age-stratified serological data overall and for the different regions using maximum-likelihood to estimate the average force of infection (rate at which susceptibles are infected) for rubella for those aged <13 and ⩾13 years, denoted by λ y and λ o , respectively. The equation for the proportion of people that were susceptible by age a (s(a)) assumed that the force of infection differed between those aged <13 years and those aged ⩾13 years. For the age groups represented in the data, which referred to people aged ⩾13 years, the equation was:
$$s\left( a \right) = \exp\left( {-13{\lambda _y}} \right)\exp\left( { - {\lambda _o}\left( {a-13} \right)} \right).$$
The CRS incidence per 100 000 live births in 5-year maternal age groups up to age 44 years (A i ) in each region was then calculated using the expression
$$0.65 \times s\left({{A_i}} \right) \times (1-{\exp}\left({-16{\lambda_o}/52} \right) \times 100000,$$
where s(A i ) is the estimated proportion of women in age group A i that are susceptible and λ o is force of infection estimated for the given region. The CRS incidence per 100 000 live births in women aged 15–44 years was then calculated as the average of the corresponding value in age group A i , weighted by the number of live births occurring in that maternal age group in 2010. As in previous analyses [Reference Miyakawa12, Reference Cutts and Vynnycky19], the risk of a child being born with CRS was assumed to be 65% if the mother was infected during the first 16 weeks of pregnancy and zero thereafter. The CRS incidence was then multiplied by estimates of the number of live births in women in the corresponding age group to obtain the number of CRS cases. The number of live births was calculated by multiplying age-specific fertility rates by the number of women in each age group in 2010, both extracted from the United Nations population database, 2008 revision. CIs on the force of infection and CRS incidence for each region and catalytic model were obtained by bootstrapping, using 1000 bootstrap datasets generated using the approach employed by Shkedy et al. [Reference Shkedy20].
Human subjects' protection
We obtained written informed consent from participants and from parents or guardians of women aged <18 years. The protocol was reviewed at CDC and determined to be a programme evaluation and therefore exempt from Institutional Review Board review. The survey protocol was approved by the Cambodian National Ethics Committee for Health Research.
RESULTS
Field teams visited 2200 HHs; 1829 (83%) had at least one eligible woman, 271 (12%) had no eligible women, and 100 HHs (4·5%) were not able to be contacted to determine if any eligible women lived in the HH despite up to three visits being made. These 100 HHs were treated in the weighting adjustment and the non-response calculation as if they had one non-responding eligible woman. Of the 2167 women enrolled, 2154 had samples available for serological testing and are included in the analysis. Overall response rate was 92%, response rate by region ranged from 85% to 98%. Table 1 shows the distribution of participants by age group and region. Median age of participants was 26 years, and 63% of participants were pregnant or had been previously. Twenty-one of 24 provinces had clusters that were selected for inclusion in the survey; the three provinces without clusters included in the survey (Kep, Mondolkiri, Pailin) constitute 1·2% of Cambodia's population according to the 2008 Census.
Table 1. Distribution of serosurvey participants by age group and region, Cambodia, 2012
PV antibody serological results
Overall, 29·8% (95% CI 27·4–32·3) of women were seronegative for antibody to ⩾1 PV serotypes. PV antibody seropositivity ranged from 83·3% for PV3 to 93·4% for PV2. Antibody seropositivity did not differ by location for any of the PVs, but did differ significantly by age for both PV1 and PV3 (P < 0·001, Table 2). PV1 antibody seropositivity increased with increasing age; from 67·9% in those aged 15–19 years to 97·2% in those aged 35–39 years. PV3 antibody seropositivity also increased with age; from 72·9% in those aged 15–19 years to 92·6% in those aged 35–39 years. In the 15–19 years age group, 48·2% (95% CI 42·4–54·1) were susceptible to WPV types 1 or 3; 35·4% (95% CI 30·5–40) of those aged 20–24 years were susceptible to ⩾1 type. Antibody seropositivity for all PVs did not differ between rural and urban areas (data not shown).
Table 2. Antibody seropositivity to polioviruses by age group and region in women aged 15–39 years, Cambodia, 2012
CI, Confidence interval.
* Estimates adjusted to account for sampling weights and survey design.
† Two participants without sample available for polio testing.
‡ Design effect (DE). P1: DE = 1·3; P2: DE = 1·1; P3: DE = 1·4. Estimated intra-class correlation (ICC) = (DE − 1)/(b – 1), where b is the average number of responses per cluster. P1: ICC = 0·016; P2: ICC = 0·006; P3: ICC = 0·019.
Measles antibody serological results
Of 2154 women tested for measles antibodies, 21(1%) had samples with repeated equivocal results and were considered to be seropositive. Analysing these results as seronegative or excluding equivocal results did not change the seroprevalence estimates significantly for any subgroup (data not shown). Overall measles seroprevalence was 95·9% (95% CI 92·8–96·4), and the estimated seropositivity was >95% for every subgroup except those aged 15–19 years. Seroprevalence was significantly lower for this age group than for all other age groups (89·6% vs. 97·4%, P < 0·001). Seroprevalence did not differ by region or rural/urban location (Table 3).
Table 3. Antibody seropositivity to measles virus by age group, region, and urban/rural residence in women aged 15–39 years, Cambodia, 2012
CI, Confidence interval.
* Estimates adjusted to account for sampling weights and survey design.
† Measles (overall) design effect = 2·4 and intraclass correlation = 0·066.
Rubella antibody serological results
Rubella antibody seronegativity was 26·7% overall. Women aged 15–19 years were significantly less likely to be seropositive for antibody to rubella virus (59·7%) than women in all other age groups (71·4–79·0%, P < 0·001). A significant decreased seropositivity was apparent in women aged 20–24 (71·4%) compared to those aged 25–34 (78·9–79·0%, P < 0·05). Women from rural areas were less likely to be seropositive for antibody to rubella virus (71·7%) than women from urban areas (78·7%, P = 0·015). Women in the southwest and southeast provinces had the lowest estimated antibody seropositivity, but this difference was not significant (Table 4).
Table 4. Antibody seropositivity to rubella virus by age group, region, and urban/rural residence in women aged 15–39 years, Cambodia, 2012
CI, Confidence interval.
* Estimates adjusted to account for sampling weights and survey design.
† Rubella (overall) design effect = 2·1 and intra-class correlation = 0·054.
Model-based estimates of CRS incidence
For Cambodia overall, the best-fitting value for the force of rubella virus infection was 76/1000 persons per year (95% CI 65–87) for children aged <13 years and 33/1000 persons per year (95% CI 21–44) for those aged ⩾13 years. These estimates varied between regions, with the highest force of infection in those aged ⩾13 years estimated for Phnomh Penh (61/1000 persons per year (95% CI 31–97). The CRS incidence for Cambodia overall was estimated therefore to be 167/100 000 live births in women aged 15–44 years (95% CI 107–220). Predicted CRS incidence ranged from 76/100 000 live births in West Cambodia to 270/100 000 live births in Southeast Cambodia (Fig. 1), reflecting differences in the force of infection between these areas. Incidence decreased with increasing age group from 222/100 000 births in women aged 15–19 years to 115/100 000 in women aged 35–39 years (Table 5). For 2010, 645 babies (95% CI 416–859) were estimated to be born with CRS in Cambodia.
Table 5. Estimates of the age-specific force of infection and the CRS incidence per 100 000 live births (for 5-year age groups) in different regions of Cambodia
Values in parentheses are 95% confidence intervals.
DISCUSSION
This nationally representative survey demonstrated significant gaps in the immunity of Cambodian women to PV1 and PV3, especially in the late teen years, where around half the cohort was susceptible to ⩾1 serotype. The increasing seroprevalence to PV1 and PV3 with age in our study, as has been seen elsewhere [Reference Alleman21, Reference Green22], likely reflects increased remote exposure to circulating WPV types 1 and 3 in older persons in our population and a greater opportunity for exposure to shed polio vaccine virus through contact with children. Although we did not measure susceptibility in men, previous studies have demonstrated lower PV antibody seropositivity in men than women [Reference Gidding23–Reference Stark26], suggesting the overall level of immunity in the adult population might be even lower than in our survey. Current immunity to polio in women in our survey is less than the population immunity documented prior to recent outbreaks in Tajikistan and China [Reference Khetsuriani10, Reference Wang27].
The recognition of a substantial immunity gap to PV in adolescents and young adults underlines the importance of maintaining good acute flaccid paralysis surveillance, including in persons aged >15 years, and the need to maintain high levels of polio vaccine coverage in young children. Although financially and logistically challenging, proactive polio vaccination campaigns are less expensive than outbreak response campaigns and have been advocated for areas with identified immunity gaps [Reference Thompson28]. The WHO has recently broadened the recommended age for polio outbreak response campaigns to at least 15 years, given the increasing role of older unprotected persons in many outbreaks and the need to rapidly boost overall population immunity [29, 30]. Based on the epidemiology of the outbreak, recent response campaigns in China and the Republic of Congo were targeted at an even wider age range [Reference Luo3, 4].
The >95% overall measles antibody seroprevalence in Cambodian women is higher than both the modelled estimates of susceptibility targets necessary for elimination [Reference Gay31] and the level of population immunity in adults in the USA during the absence of endemic measles [Reference Hutchins32]. Previous studies in multiple settings have demonstrated no difference in measles seroprevalence between adult men and women [Reference de Melker33–Reference Liu35]. This would suggest that in the face of high population immunity in children, there is not a significant enough reservoir of susceptible adults to perpetuate measles virus transmission in Cambodia. Maintaining high vaccine coverage in children aged <15 years would likely effectively eliminate endemic measles, although limited transmission could still occur in groups with lower (<90%) immunity. The recent addition of a second measles dose to the national vaccination schedule and the 2013 measles-rubella vaccine campaign are therefore crucial steps towards achieving the permanent absence of endemic measles in Cambodia.
More than 25% of WCBA are susceptible to rubella virus – a level considered to indicate a high risk for CRS [36, Reference Cutts37]. The overall 26·7% rubella susceptibility is similar to the 29·8% susceptibility in neighbouring Vietnam [Reference Miyakawa12] prior to a large rubella outbreak during 2011–2012 which resulted in 281 reported CRS cases [1]. While there are, to our knowledge, no previous estimates of rubella seroprevalence or CRS cases from Cambodia, the estimated incidence rate of 176 cases/100 000 live births in our study is very similar to the overall estimate of 173/100 000 in the WHO's WPR in 1996 [Reference Cutts and Vynnycky19]. No serological data from Cambodia was available for this previous estimate and it was assumed that the incidence rate was similar to the overall regional rate [Reference Cutts and Vynnycky19]; if this is true, there has been little change in this rate between 1996 and 2012. Without the introduction of RCV, on the basis of the rubella force of infection and the age-specific fertility rate, at least 400 children are born every year in Cambodia with CRS. The medical, economic and personal burden of continued rubella virus transmission in Cambodia is therefore substantial, as we estimate that ~1 in 600 infants born in Cambodia will have CRS. The introduction of rubella vaccine into Cambodia covering children up to age 15 years should decrease the overall force of infection significantly but will not decrease the number of WCBA susceptible to rubella virus for many years [11]. Fortunately, sentinel CRS surveillance was established in Cambodia in 2011. This surveillance identified 32 CRS cases in 2012 [1], which while only 5% of our estimated CRS burden in the country was >10% of the total reported cases worldwide [38]; this highlights the absence of CRS surveillance in many areas of the world with continued rubella transmission. Maintaining high routine rubella vaccine coverage, monitoring for changes in rubella epidemiology, and expanding CRS surveillance will be essential after the campaign. Given the significant susceptibility of WCBA, rubella vaccination strategies targeting WCBA, either through routine services for WCBA or a vaccination campaign including men and women should be considered, as recommended by WHO and the Strategic Advisory Group of Experts on Immunization (SAGE) [11, 39], to accelerate progress towards rubella and CRS elimination.
Our survey has certain limitations. First, we surveyed women only, thus inference to the entire adult population should be done cautiously. If anything this would likely underestimate the true immunity gaps to PVs. Second, although we used rigorous methods to randomly select the sample, only 2154 women throughout Cambodia were included. For this reason, we do not provide estimates for individual provinces or for age groups within the regions. Finally, although we selected survey EAs randomly and used PPS methods, the initial sampling frame came from EAs included in the 2010 DHS survey, which was based on the 2008 census. This sampling strategy gave us access to detailed maps and household listings of EAs, helping to ensure that a true probability sample was selected. This improved the representativeness of our survey and permitted appropriate weighting to allow for unbiased population estimates. However, smaller or non-existent settlements in 2008 would be less likely to be included and therefore underrepresented in our survey.
Overall, our serosurvey shows a level of immunity for measles in adult women that is encouraging for sustained elimination of measles in Cambodia if high childhood vaccine coverage is maintained. However, the gap in immunity to PV identified in our survey in young women is of sufficient size that a significant outbreak in this age group is possible if polio were reintroduced into Cambodia. In addition, the high susceptibility to rubella suggests that until this age cohort passes through prime fertility years, a substantial number of CRS cases can be expected in Cambodia unless further action is undertaken to protect these women. This is probably true even with the reduced force of rubella infection expected to follow the MR vaccination campaign. Sensitive rubella and CRS surveillance after this campaign will be essential to monitor progress towards accelerated rubella and CRS control and changes in rubella epidemiology. Periodic repeat seroprevalence surveys, ideally integrated in health or demographic surveys, could provide objective evidence of changes in population susceptibility profiles over time and monitor the impact of changes in immunization programme practices towards achieving and maintaining VPD control goals.
ACKNOWLEDGEMENTS
We thank Rownak Khan, Penelope Campbell, Paul Rota, Richard Duncan, and Youngmee Jee for their assistance and support in the design and conduct of the study and Heather Scobie for thorough review and insightful comments on the manuscript and assistance with preparation of Figure 1. We also thank Deborah Moore, Yiting Zhang, Sharla McDonald, Larin McDuffie, William Hendley, Tamanna Ahmed, and Patricia Mitchell for polio neutralizing antibody serum analysis.
This work was supported by the US Centers for Disease Control and Prevention through Cooperative Agreement with the National Institute for Public Health, Cambodia (Cooperative Agreement number: 5U2GPS000939–05).
The findings and conclusions in this report are those of the author(s) and do not necessarily represent the official position of Centers for Disease Control and Prevention (CDC).
DECLARATION OF INTEREST
None.
