1043
Measles Incidence, Vaccine Efficacy, and Mortality in Two Urban African Areas with High Vaccination Coverage Peter Aaby, Kim Knudsen, Tht)ger Gorm Jensen, Jesper Thirup, Anja Poulsen, Morten Sodemann, Maria Clotilde da Silva, and Hilton Whittle
From the Institute of Anthropology and Statistical Research Unit, University of Copenhagen, and International Medical Cooperation Committee, Denmark; MCH Programme, Ministry of Public Health, Bissau, Guinea-Bissau; and MRC Laboratories, Fajara, The Gambia
Prevention of measles among infants is particularly important due to the very high case fatality rate among the youngest children [1]. The delayed impact of measles may be especially strong for children who have had measles before 1 year of age [2, 3]. It is commonly expected that as immunization coverage increases, the age of patients with measles will rise and, at the same time, the risk of exposure before age 9 months will diminish considerably [4]. These predictions have not always come true [5] ~ Recent studies from urban African areas have shown that a high proportion (13 %-45 %) of measles cases occur in children younger than 9 months [4-7], but there has been no prospective studies in these areas or age groups to assess the size ofthe problem. Retrospective surveys and hospital data have indicated incidences of 3 %-12 % before the age of vaccination [4-7], but these may have been biased by errors in age recall, omission of deaths, or identification of children at the health center. A high incidence of measles among inf~nts in urban African areas is related to the influx of susceptible individuals from rural areas, a high degree of crowding in urban areas, and the gathering of young infants in health institutions [5, 6, 8-10]. Several studies from Africa also indicate rather high vaccination failure rates [11], which could contribute to the continued risk of measles transmission. Received 8 January 1990; revised 8 May 1990. Financial support: Ministry of Health (Guinea-Bissau) and International Medical Cooperation Committee, Danchurchaid, Council for Development Research, Medical Research Council, and Social Science Research Council (Denmark). Reprints or correspondence: Mr. Peter Aaby, Institute of Anthropology, Frederiksholms Kanal4, 1220 Copenhagen K, Denmark. The Journal of Infectious Diseases 1990;162:1043-1048 © 1990 by The University of Chicago. All rights reserved. 0022-1899/90/6205-0007$Or .00
The present prospective study was undertaken in two neighboring districts in Bissau, the capital of Guinea-Bissau, to obtain a better understanding of the measles incidence in the first years of life and its relation to vaccination coverage. Though both areas had relatively high coverage for measles vaccination (>60%), it was clearly higher in the area where a randomized control trial of the protective efficacy of the Edmonston-Zagreb (EZ) and the Schwarz measles vaccines was undertaken [12]. As part of that trial, because it was thought that the improved vaccination coverage in the area might make detection of an effect of early vaccination difficult, surveillance was also instituted in the neighboring district, where measles vaccination was routinely given at age 9 months.
Subject and Methods Study populations. Bandim 1 and Bandim 2 are neighboring districts in Bissau, and their socioeconomic standards, house construction, and ethnic backgrounds are similar. In a sample survey undertaken during this study, the median number ofpersons per house was 15 (range, 2-36) in Bandim 1 and 14 (range, 1-38) in Bandim 2. The population of Bissau during the study was 1'\)150,000, with a population of I'\) 10,000 persons in each of Bandim 1 and Bandim 2. In both areas, children born between 1 August 1984 and 31 September 1985 who were registered in the district before age 4 months were followed until June 1987. In Bandim 1, only the 274 children randomized (before age 4 months) to receive Schwarz measles vaccine at 9 months were analyzed in the present study [12]. In Bandim 2, 448 children fulfilled the criteria for inclusion in the study. Registration and vaccination systems. Since 1978, child health and mortality have been studied in Bandim 1 [1, 9-13]. As part of that study, local assistants make monthly visits to all houses in the eight zones (neighborhoods) to register pregnancies, births, and
Downloaded from http://jid.oxfordjournals.org/ at New York University on June 30, 2015
Measles incidence, vaccine efficacy, and mortality were examined prospectively in two districts in Bissau where vaccine coverage for children aged 12-23 months was 81% (Bandim 1) and 61% (Bandim 2). There was little difference in cumulative measles incidence before 9 months of age (6.1% and 7.6%, respectively). Between 9 months and 2 years of age, however, 6.1% contracted measles in Bandim 1 and 13.7% in Bandim 2. Even adjusting for vaccination status, incidence was significantly higher in Bandim 2 (relative risk 1.6, P = .04). Even though 95% ofthe children had measles antibodies after vaccination, vaccine efficacy was not more than 68% (95% confidence interval [CI] 39%-84%) and was unrelated to age at vaccination. Unvaccinated children had a mortality hazard ratio of 3.0 compared with vaccinated children (P = .002), indicating a protective efficacy against death of 66% (CI 32%-83%) of measles vaccination. These data suggest that it will be necessary to vaccinate before age 9 months to control measles in hyperendemic urban African areas.
1044
Aaby et al.
the cumulative measles morbidity and mortality incidences. In this analysis, measles immunization was assumed to be effective from 3 weeks (21 days) after vaccination. Several children developed measles 4-5 weeks after vaccination. As this could be due to minor inaccuracies in the date of infection, analyses also were made with 35 days as the time limit for change in immunization status. Apart from controlling for age, adjustments for three factors (area, sex, and vaccination status) were made using a multivariate Cox regression analysis, which deals with survival data with censored observation [16]. Measles incidence and overall mortality were analyzed for all children from age 4 months. The effect of vaccination against measles was also analyzed in a model that included only children correctly vaccinated after age 8.5 months (255 days). In this analysis, children who moved, died, had measles, or were vaccinated before age 8.5 months were excluded. Separate analyses were made where children who contracted measles were censored from the moment they contracted measles. Vaccine efficacy. Children were considered vaccinated if they had a vaccination card with a measles vaccination indicated. Since vaccinations are monitored regularly in both areas, there are undoubtedly few vaccinated children classified as unvaccinated due to loss of the card. Vaccine efficacy (VE) was calculated from the age-specific incidence rates for unvaccinated (IRuv) and vaccinated children (lRv): VE = [(IRuv - IRv) x 100]/IRuv. In the multivariate Cox regression model, VE was estimated using the same equation, where IRs were adjusted for age, area, and sex. Adjustments were made for these factors because the incidence of measles varied by age, area, and sex.
Results Coverage and quality ofmeasles vaccination. In Bandim 1, 25 % of the cohort received the Schwarz measles vaccine by age 9.7 months, 50% by 11.1 months, and 75% by 14.1 months (figure 1). For cohort children aged 12-23 months, the coverage was 81 % with the Schwarz measles vaccine. The coverage rate has been adjusted for children who moved or died. Among children who received a Schwarz measles vaccine and did not report measles infection, 137 were tested after age 18 months; 95 % (130) had HAl antibodies to measles. In Bandim 2, 25 % of the children had received measles vaccine by age 9.7 months, 50% by 15.0 months, and 75% by 26.2 months (figure 1). For children aged 12-23 months, the coverage was 61 %. However, 5 % of the children in Bandim 2 had been vaccinated before age 8.5 months, giving a coverage rate of 56 % if only correctly vaccinated children are considered. Among children with no history of measles, 54 were tested for antibodies to measles when 2 years old. All 49 children with a documented vaccination had HAl antibodies (including four children vaccinated before 9 months), whereas none of the five unvaccinated children had antibodies. Measles cases and quality of diagnoses. Most cases of measles occurred during the last months of 1985 and the first
Downloaded from http://jid.oxfordjournals.org/ at New York University on June 30, 2015
deaths. Every third month, children under age 3 years are called for weighing and vaccination in their neighborhood. Once every 4 weeks, the unvaccinated children of each zone are called to the health center for vaccination. Blood samples are collected just before the Schwarz measles vaccination at 9 and 18 months. The vaccinations are supervised by Danish medical students. In January 1985, just before the initiation of the current study, measles immunization coverage was 84% (615/731) for children aged 12-35 months. Bandim 2 is likewise divided into eight zones. The registration of pregnancies, births, and deaths among children o
U
30
10
otp:;:=:;:::::..-...,....._...,....._ _-_-...,...-...,...-...,...6
8
10
12
14
16
18
20
22
24
26
--J 28
30
Age i'I Months - - Bandm1
6 months of 1986. The measles case fatality rate (CFR) is documented in table 1. There was a trend toward higher mortality among children from Bandim 2 compared with Bandim 1 (relative risk [RR] 3.66,95% confidence interval [CI] 0.59-22.80). The CFR for measles in Bandim 1 before the initiation ofthe vaccine study was similar to that subsequently found in Bandim 2. Because most cases in Bandim 1 were examined and treated at home, the lower CFR in Bandim 1 in 1985-1987 may be related to the treatment. In Bandim 1, 24 children contracted measles before being immunized with the Schwarz measles vaccine. Because 4 had measles in the week after measles vaccination, it was not possible to determine if seroconversion was due to infection or vaccination. The other 20 unvaccinated children had HAl measles antibody titers of ~OO mID when tested before measles vaccination (~9 months). Among children with no history of measles, only 11 % (231207) had a similar titer before vaccination (P < .001; X2 = 93.4). In Bandim 2, 31 of the children with a history of measles were tested for measles antibodies at age 2 years; 23 had also received measles vaccine. All were positive for HAl antibodies; 52 % (12/23) of the children had a titer of ~25,600 mID, whereas a similar titer was found in only 27% (13/49) of the vaccinated children who had no history of measles (P < .05; X2 = 4.46). Six of the eight unvaccinated children who reported measles tested positive for measles antibodies. Measles after vaccination. Several children in Bandims 1 and 2 had measles >3 weeks after measles vaccination. As shown in table 2, vaccinated children were somewhat more likely to be secondary cases than were unvaccinated children
-----_. Bandm 2
Table 1. Measles case fatality rate in Bissau children
Measles Incidence, Vaccine Efficacy, and Mortality in Two Urban African Areas with High Vaccination Coverage Peter Aaby, Kim Knudsen, Tht)ger Gorm Jensen, Jesper Thirup, Anja Poulsen, Morten Sodemann, Maria Clotilde da Silva, and Hilton Whittle
From the Institute of Anthropology and Statistical Research Unit, University of Copenhagen, and International Medical Cooperation Committee, Denmark; MCH Programme, Ministry of Public Health, Bissau, Guinea-Bissau; and MRC Laboratories, Fajara, The Gambia
Prevention of measles among infants is particularly important due to the very high case fatality rate among the youngest children [1]. The delayed impact of measles may be especially strong for children who have had measles before 1 year of age [2, 3]. It is commonly expected that as immunization coverage increases, the age of patients with measles will rise and, at the same time, the risk of exposure before age 9 months will diminish considerably [4]. These predictions have not always come true [5] ~ Recent studies from urban African areas have shown that a high proportion (13 %-45 %) of measles cases occur in children younger than 9 months [4-7], but there has been no prospective studies in these areas or age groups to assess the size ofthe problem. Retrospective surveys and hospital data have indicated incidences of 3 %-12 % before the age of vaccination [4-7], but these may have been biased by errors in age recall, omission of deaths, or identification of children at the health center. A high incidence of measles among inf~nts in urban African areas is related to the influx of susceptible individuals from rural areas, a high degree of crowding in urban areas, and the gathering of young infants in health institutions [5, 6, 8-10]. Several studies from Africa also indicate rather high vaccination failure rates [11], which could contribute to the continued risk of measles transmission. Received 8 January 1990; revised 8 May 1990. Financial support: Ministry of Health (Guinea-Bissau) and International Medical Cooperation Committee, Danchurchaid, Council for Development Research, Medical Research Council, and Social Science Research Council (Denmark). Reprints or correspondence: Mr. Peter Aaby, Institute of Anthropology, Frederiksholms Kanal4, 1220 Copenhagen K, Denmark. The Journal of Infectious Diseases 1990;162:1043-1048 © 1990 by The University of Chicago. All rights reserved. 0022-1899/90/6205-0007$Or .00
The present prospective study was undertaken in two neighboring districts in Bissau, the capital of Guinea-Bissau, to obtain a better understanding of the measles incidence in the first years of life and its relation to vaccination coverage. Though both areas had relatively high coverage for measles vaccination (>60%), it was clearly higher in the area where a randomized control trial of the protective efficacy of the Edmonston-Zagreb (EZ) and the Schwarz measles vaccines was undertaken [12]. As part of that trial, because it was thought that the improved vaccination coverage in the area might make detection of an effect of early vaccination difficult, surveillance was also instituted in the neighboring district, where measles vaccination was routinely given at age 9 months.
Subject and Methods Study populations. Bandim 1 and Bandim 2 are neighboring districts in Bissau, and their socioeconomic standards, house construction, and ethnic backgrounds are similar. In a sample survey undertaken during this study, the median number ofpersons per house was 15 (range, 2-36) in Bandim 1 and 14 (range, 1-38) in Bandim 2. The population of Bissau during the study was 1'\)150,000, with a population of I'\) 10,000 persons in each of Bandim 1 and Bandim 2. In both areas, children born between 1 August 1984 and 31 September 1985 who were registered in the district before age 4 months were followed until June 1987. In Bandim 1, only the 274 children randomized (before age 4 months) to receive Schwarz measles vaccine at 9 months were analyzed in the present study [12]. In Bandim 2, 448 children fulfilled the criteria for inclusion in the study. Registration and vaccination systems. Since 1978, child health and mortality have been studied in Bandim 1 [1, 9-13]. As part of that study, local assistants make monthly visits to all houses in the eight zones (neighborhoods) to register pregnancies, births, and
Downloaded from http://jid.oxfordjournals.org/ at New York University on June 30, 2015
Measles incidence, vaccine efficacy, and mortality were examined prospectively in two districts in Bissau where vaccine coverage for children aged 12-23 months was 81% (Bandim 1) and 61% (Bandim 2). There was little difference in cumulative measles incidence before 9 months of age (6.1% and 7.6%, respectively). Between 9 months and 2 years of age, however, 6.1% contracted measles in Bandim 1 and 13.7% in Bandim 2. Even adjusting for vaccination status, incidence was significantly higher in Bandim 2 (relative risk 1.6, P = .04). Even though 95% ofthe children had measles antibodies after vaccination, vaccine efficacy was not more than 68% (95% confidence interval [CI] 39%-84%) and was unrelated to age at vaccination. Unvaccinated children had a mortality hazard ratio of 3.0 compared with vaccinated children (P = .002), indicating a protective efficacy against death of 66% (CI 32%-83%) of measles vaccination. These data suggest that it will be necessary to vaccinate before age 9 months to control measles in hyperendemic urban African areas.
1044
Aaby et al.
the cumulative measles morbidity and mortality incidences. In this analysis, measles immunization was assumed to be effective from 3 weeks (21 days) after vaccination. Several children developed measles 4-5 weeks after vaccination. As this could be due to minor inaccuracies in the date of infection, analyses also were made with 35 days as the time limit for change in immunization status. Apart from controlling for age, adjustments for three factors (area, sex, and vaccination status) were made using a multivariate Cox regression analysis, which deals with survival data with censored observation [16]. Measles incidence and overall mortality were analyzed for all children from age 4 months. The effect of vaccination against measles was also analyzed in a model that included only children correctly vaccinated after age 8.5 months (255 days). In this analysis, children who moved, died, had measles, or were vaccinated before age 8.5 months were excluded. Separate analyses were made where children who contracted measles were censored from the moment they contracted measles. Vaccine efficacy. Children were considered vaccinated if they had a vaccination card with a measles vaccination indicated. Since vaccinations are monitored regularly in both areas, there are undoubtedly few vaccinated children classified as unvaccinated due to loss of the card. Vaccine efficacy (VE) was calculated from the age-specific incidence rates for unvaccinated (IRuv) and vaccinated children (lRv): VE = [(IRuv - IRv) x 100]/IRuv. In the multivariate Cox regression model, VE was estimated using the same equation, where IRs were adjusted for age, area, and sex. Adjustments were made for these factors because the incidence of measles varied by age, area, and sex.
Results Coverage and quality ofmeasles vaccination. In Bandim 1, 25 % of the cohort received the Schwarz measles vaccine by age 9.7 months, 50% by 11.1 months, and 75% by 14.1 months (figure 1). For cohort children aged 12-23 months, the coverage was 81 % with the Schwarz measles vaccine. The coverage rate has been adjusted for children who moved or died. Among children who received a Schwarz measles vaccine and did not report measles infection, 137 were tested after age 18 months; 95 % (130) had HAl antibodies to measles. In Bandim 2, 25 % of the children had received measles vaccine by age 9.7 months, 50% by 15.0 months, and 75% by 26.2 months (figure 1). For children aged 12-23 months, the coverage was 61 %. However, 5 % of the children in Bandim 2 had been vaccinated before age 8.5 months, giving a coverage rate of 56 % if only correctly vaccinated children are considered. Among children with no history of measles, 54 were tested for antibodies to measles when 2 years old. All 49 children with a documented vaccination had HAl antibodies (including four children vaccinated before 9 months), whereas none of the five unvaccinated children had antibodies. Measles cases and quality of diagnoses. Most cases of measles occurred during the last months of 1985 and the first
Downloaded from http://jid.oxfordjournals.org/ at New York University on June 30, 2015
deaths. Every third month, children under age 3 years are called for weighing and vaccination in their neighborhood. Once every 4 weeks, the unvaccinated children of each zone are called to the health center for vaccination. Blood samples are collected just before the Schwarz measles vaccination at 9 and 18 months. The vaccinations are supervised by Danish medical students. In January 1985, just before the initiation of the current study, measles immunization coverage was 84% (615/731) for children aged 12-35 months. Bandim 2 is likewise divided into eight zones. The registration of pregnancies, births, and deaths among children o
U
30
10
otp:;:=:;:::::..-...,....._...,....._ _-_-...,...-...,...-...,...6
8
10
12
14
16
18
20
22
24
26
--J 28
30
Age i'I Months - - Bandm1
6 months of 1986. The measles case fatality rate (CFR) is documented in table 1. There was a trend toward higher mortality among children from Bandim 2 compared with Bandim 1 (relative risk [RR] 3.66,95% confidence interval [CI] 0.59-22.80). The CFR for measles in Bandim 1 before the initiation ofthe vaccine study was similar to that subsequently found in Bandim 2. Because most cases in Bandim 1 were examined and treated at home, the lower CFR in Bandim 1 in 1985-1987 may be related to the treatment. In Bandim 1, 24 children contracted measles before being immunized with the Schwarz measles vaccine. Because 4 had measles in the week after measles vaccination, it was not possible to determine if seroconversion was due to infection or vaccination. The other 20 unvaccinated children had HAl measles antibody titers of ~OO mID when tested before measles vaccination (~9 months). Among children with no history of measles, only 11 % (231207) had a similar titer before vaccination (P < .001; X2 = 93.4). In Bandim 2, 31 of the children with a history of measles were tested for measles antibodies at age 2 years; 23 had also received measles vaccine. All were positive for HAl antibodies; 52 % (12/23) of the children had a titer of ~25,600 mID, whereas a similar titer was found in only 27% (13/49) of the vaccinated children who had no history of measles (P < .05; X2 = 4.46). Six of the eight unvaccinated children who reported measles tested positive for measles antibodies. Measles after vaccination. Several children in Bandims 1 and 2 had measles >3 weeks after measles vaccination. As shown in table 2, vaccinated children were somewhat more likely to be secondary cases than were unvaccinated children
-----_. Bandm 2
Table 1. Measles case fatality rate in Bissau children
