Corresponding author Sophie E. Moore, MRC International Nutrition Group, Nutrition and Public Health Intervention Research Unit, Department of Epidemiology and Population Health, London School of Hygiene and Tropical Medicine, Keppel Street, London WC1E 7HT, UK. Tel.: +44 207 958 8138; Fax: +44 207 958 8111; E-mail: email@example.com
Objective To explore the relationship between calendar month of administration and antibody (Ab) response to vaccination in subjects from The Gambia and Pakistan, two countries with distinct patterns of seasonality.
Methods Three cohorts were investigated: Responses to rabies and pneumococcal vaccine were assessed in 472 children (mean age 8 years, males 53%) from rural Gambia. Responses to tetanus, diphtheria and hepatitis B (HBsAg) were investigated in 138 infants also from The Gambia (birth to 52 weeks of age, males 54%). Responses to rabies and Vi typhoid vaccines were assessed in 257 adults from Lahore, Pakistan (mean age 29.4 years, males 57%).
Results In Gambian children, significant associations were observed between month of vaccination and Ab response for the pneumococcal and rabies vaccines. As no consistent pattern by month was observed between the responses, it is assumed that different immunomodulatory stimuli or mechanisms were involved. In Pakistani adults, a significant pattern by month of vaccination was observed with both rabies and typhoid vaccine. No monthly influences were observed in the infant study to the tetanus, diphtheria or the HbsAg vaccines.
Conclusions Antibody responses to certain specific vaccines are influenced by month of administration. Further research is required to elucidate the precise mechanisms explaining these observations, but a co-stimulatory effect of seasonally variable environmental antigens is a likely cause. Future studies of Ab response to vaccination in countries with a seasonally dependent environment should consider month of vaccination when interpreting study findings.
Objectif Etudier le lien entre le mois d'administration et la réponse anticorps à un vaccin chez des sujets vivants en Gambie et au Pakistan; deux pays ayant des modèles distincts de saisonnalité.
Méthodes Trois cohortes étaient étudiées: les réponses aux vaccins contre la rage et le pneumocoque étaient étudiées chez 472 enfants (âge moyen 8 ans, 53% de sexe masculin) en zone rurale en Gambie. Les réponses aux vaccins contre le tétanos, la diphtérie et l'hépatite B (Ag-HBs) étaient étudiées chez 138 nourrissons provenant également de Gambie (de la naissance à 52 semaines d’âge, 54% de sexe masculin). Les réponses aux vaccins contre la rage et la typhoïde (Vi) étaient étudiées chez 257 adultes à Lahore, Pakistan (âge moyen 29,4 ans, 57% de sexe masculin).
Résultats Chez les enfants Gambiens, des associations significatives étaient observées entre le mois de vaccination et la réponse anticorps pour les vaccins contre le pneumocoque et la rage. Comme aucune relation logique était observée entre le mois d'administration et les réponses vaccinales, il est possible que différents mécanismes ou stimuli immuno-modulateurs soient impliqués. Chez les adultes Pakistanais, un lien significatif avec le mois de vaccination était observé avec les 2 vaccins contre la rage et la typhoïde. Aucune influence du mois était observé dans l’étude de la vaccination contre le tétanos, la diphtérie et l'hépatite B chez les nourrissons.
Conclusions La réponse anticorps à certains vaccins spécifiques est influencée par le mois d'administration. D'autres études sont nécessaires pour élucider le mécanismes précis pouvant expliquer ces observations, mais un effet co-stimulateur d'antigènes environnementaux variants selon les saisons est probablement en cause.
Objetivo Explorar la relación entre el calendario mensual de administración de vacunas y la respuesta de anticuerpos en sujetos de Gambia y Pakistán; dos países con diferentes patrones de estacionalidad.
Métodos Se investigaron tres cohortes: se evaluaron las respuestas a la vacuna de la rabia y a la vacuna neumocócica en 472 niños (edad promedio 8 años, varones 53%) de áreas rurales de Gambia. Se investigaron las respuestas al tétanos, difteria y hepatitis B (HBsAg) en 138 infantes también de Gambia (desde el nacimiento a las 52 semanas de edad, varones el 54%). Se establecieron las respuestas a las vacunas de la rabia y de la Vi tifoidea en 257 adultos de Lahore, Pakistán (promedio de edad 29.4 años, y varones un 57%).
Resultados En los niños de Gambia, se observaron significantes asociaciones entre el mes de vacunación y la respuesta de anticuerpos para las vacunas neumocócicas y de la rabia. Dado que no se observó un consistente patrón por mes entre las respuestas, se asume que diferentes mecanismos o estímulos inmunomoduladores han estado involucrados. En adultos paquistaníes, un significativo patrón en el mes de vacunación fue observado tanto en la vacuna tifoidea como en la de la rabia. No se observaron influencias en el mes en el estudio de los infantes en las vacunas del tétanos, difteria o HbsAg.
Conclusiones Las respuestas de anticuerpos a ciertas vacunas específicas están influenciadas por el mes de administración. Es necesario más investigación para dilucidar los mecanismos precisos que expliquen estas observaciones, pero un efecto coestimulador de antígenos de variables ambientales estacionales es una causa probable. Los futuros estudios de respuesta de anticuerpos a las vacunas en países con un medio ambiente dependiente de las estaciones deberán considerar el mes de vacunación al momento de interpretar los hallazgos del estudio.
Vaccination, together with sanitation, has made the most important contribution to public health in the past century, and has dramatically reduced the number of deaths attributed to infectious disease across the world. Many such vaccines are now easily available to all, and programmes such as the expanded programme of immunization (EPI) reach people even in the world's poorest countries. While most commercially produced vaccines elicit a protective response in most recipients, evidence suggests that a number of factors can influence the strength of the response between individuals. Considering these factors in studies of antibody (Ab) response to vaccination is therefore important.
Age at vaccination is the most commonly recognized to influence Ab responses, with variations specifically seen in the young and the elderly. In early life, numerous factors contribute to the limitation of Ab responses to both protein and polysaccharide vaccines. One factor is immunological immaturity, which may affect B cells and/or their interactions with antigen-presenting cells, T cells, follicular dendritic cells or other components of the lymphoid microenvironment (Siegrist 2001). Pre-existing maternal Ab levels are another factor: they may limit vaccine response through various mechanisms (Siegrist 2003), or stimulate vaccine response through breastfeeding, including specific enhancement of the immunoglobulin G2 response during Haemophilus influenzae type b (Hib) disease in childhood (Silfverdal et al. 2002). In the elderly, age-related changes in various aspects of immune function, such as an increase in the number of immune cells showing signs of replicative senescence, result in a reduced Ab response to vaccination (Rubins et al. 1998; Fisman et al. 2002) and underscore the need for specific research aimed at designing vaccines to meet the unique requirements of this population.
Gender is also thought to influence Ab response to vaccination, and it is suggested that the immune system is clearly dimorphic, with female animals from different species having higher levels of circulating immunoglobulins and presenting stronger Ab responses to immunization and infection (Da-Silva 1999). Hypothesized mechanisms for such differences include the differential modulation of immune factors by sex hormones (Da-Silva 1999).
Other suggested modifying factors include nutritional status and certain infections. In many countries, the distinct seasons define patterns in disease prevalence and nutritional status. Thus, season of administration may influence the Ab levels generated. We explored the effect that the month of vaccination had on Ab generation in three age groups of subjects from two countries exhibiting a distinct seasonal pattern: an infant cohort from The Gambia, West Africa; a child cohort also from The Gambia and an adult cohort from Lahore, Pakistan.
Materials and methods
The data come from a series of studies conducted at the Medical Research Council's (MRC) rural research station in Keneba, The Gambia, and from an ongoing study of adults living in urban Lahore, Pakistan. The study population and methodology from each site are detailed below.
Study 1. The vaccine response data were collected as part of a larger study investigating early life programming of human immune function (Moore et al. 2001). The study recruited 472 children (mean age 8.0 years, range 6.6–9.5) during an 11-month period from April 1998 to February 1999. Each child participated in the study for a period of 2 months.
At baseline, weight, height and mid-upper arm circumference were measured by the same observer (SEM) using regularly validated, standard equipment. In addition, a 6-ml fasted sample of venous blood was taken, centrifuged and plasma/serum aliquots were immediately frozen for the later analysis of pre-vaccination Ab titres and plasma levels of α1-antichymotripsin (ACT). A sample of whole blood was reserved for the assessment of haemoglobin (Hb) levels and a thick film was prepared for the examination of malaria parasites. Each child was then given a single injection of the 23-valent pneumococcal capsular polysaccharide vaccine, Pneumovax® (Merck, Sharpe and Dohme) and a preliminary dose of the human diploid-cell rabies Vaccine (HDCV; Pasteur-Merieux). Fourteen days after the first vaccination, a finger prick blood sample was collected for assessment of Ab response to the first rabies vaccination. Twenty-eight days after the first rabies vaccination, a second dose was given. At the same visit, a second finger prick blood sample was obtained for assessment of Ab response to the pneumococcal vaccine. Two months following the initial vaccination, a final finger prick blood sample was collected for the measurement of rabies antibodies.
All vaccines were from the same batch and were handled identically during the study period. Samples were kept refrigerated in the laboratory in Keneba and kept cool until the point of administration. Vaccines were administered by a team of three research nurses from the MRC unit.
Laboratory analysis. Pneumococcal capsular polysaccharide-specific immunoglobulin G (IgG) was measured at the Immunobiology Unit, Institute of Child Health, London. Standard enzyme-linked immunosorbent assay methods were used to quantify anticapsular IgG antibodies to four pneumococcal serotypes (1, 5, 14 and 23F) with the use of a standard reference serum, 89SF, as described previously (Quataert et al. 1995). Antibodies to types 1 and 5 were chosen, as they represent relevant antigens in West Africa; type 14, as the assay is highly specific and can be used as an internal control for the protocol and type 23, as it has low antigenicity. Anti-rabies Ab titres were determined at the Department of Virology, Central Veterinary Laboratories, Surrey, UK, using the rapid focus fluorescence inhibition test of the WHO (Smith et al. 1973). Plasma ACT levels were measured at MRC Human Nutrition Research in Cambridge, UK, using a nephelometric assay on a Cobas Bio centrifugal analyser (Roche).
Study 2. The study population was a cohort of 138 infants recruited into an in-depth study of infant immune development (Collinson 2002). These infants were born in five villages across the West Kiang region of The Gambia between April 1998 and June 1999. All infants in the study were immunized in accordance with the Gambia EPI programme, as outlined in Table 1. Subjects received vaccines of the same approved type and source as other Gambian infants. To guarantee a cold chain, all vaccines were collected directly by the principal investigator (ACC) and transported to the Keneba laboratory refrigerator. All vaccines were administered by the principal investigator using a uniform technique (Collinson 2002). Ab levels to diphtheria, tetanus and hepatitis B (HBsAg) were measured, with concurrent measurement of plasma micronutrient status.
Table 1. Gambian expanded programme of immunization schedule 1998–2000
BCG, Bacillus Calmette–Guerin; HBV, hepatitis B subunit vaccine; OPV, trivalent oral polio vaccine; Hib, Haemophilus influenzae type b conjugate vaccine; DTP, combined diphtheria, tetanus and whole cell pertussis vaccine.
0 (within 72 h of birth)
BCG, HBV, OPV
DTP, Hib, HBV, OPV
DTP, Hib, OPV
DTP, Hib, HBV, OPV
Measles, yellow fever, OPV
Laboratory analyses. Specific diphtheria and tetanus Ab concentrations were measured in plasma samples collected from each infant at birth (cord blood) and at 8, 16 and 52 weeks of age. Samples were measured at the WHO Collaborating Centre for Vaccinology and Neonatal Immunology, University of Geneva by ELISA on antigen-coated plates, as described in detail elsewhere (Ota et al. 2002). Anti-HBsAg antibodies were measured at MRC Laboratories, Fajara, by radioimmunoassay (Sorin Biochemica, Saluggia, Italy) on plasma samples collected at 52 weeks of age. Plasma micronutrient and ACT levels were measured at MRC Human Nutrition Research in Cambridge, UK. Plasma vitamin C levels were measured on a Cobas-Bio analyser (F. Hoffman-la-Roche Ltd, Basel, Switzerland). Prior to freezing, plasma samples for vitamin C analysis were mixed with an equal volume of 10% metaphosphoric acid in order to deproteinize the plasma. Plasma zinc concentrations were determined colorimetrically using a commercial kit (Wako Chemicals, Nauss, Germany). Plasma retinol levels were measured by high performance liquid chromatography using an assay procedure derived from that of Thurnham et al. (1988). Plasma ACT levels were measured using a nephelometric assay on a Cobas Bio centrifugal analyser.
Study population. The study population was drawn from a cohort of 2468 subjects born during 14 consecutive years, from 1964 to 1978, in an urban slum (Gowalmandi) in Lahore, Pakistan (Jalil et al. 1989). These infants and their mothers were followed from birth, partly longitudinally, by personal surveillance by one of the authors (FJ). The current study was part of a larger study looking at the effect of poverty, early life malnutrition and infections on adult health and mortality (Moore et al. 2004). From the 2468 infants initially registered, 76% (1885) were retraced during 2001. Of these, 733 had infant data available and were available for follow-up within the framework of the main study. From these subjects, a smaller subcohort was selected to participate in the more in-depth study of vaccine response. These subjects were selected based on the availability of detailed records from the time of their birth, living not too far from Lahore and willingness to participate in the more in-depth study. Subjects reporting previous receipt of a rabies vaccine were excluded from the study.
Fieldwork was conducted during the months of April to September 2002. A total of 257 (males 146, females 111) adults were successfully recruited. During the study, each adult was seen on days 0, 7 and 14. On the first visit (day 0), subjects were brought to the hospital where height was measured to the nearest 0.1 cm using a wall-mounted stadiometer and weight was determined to the nearest 100 g using electronic scales (Tanita UK, West Drayton, London, UK). A blood sample was then collected by venesection for the analysis of pre-vaccination serum Ab levels. Finally, a single dose of purified Vi surface polysaccharide extracted from Salmonella typhi (Aventis Pasteur) and a single dose of rabies vaccine (Rabies vaccine BP, Verorab; Aventis Pasteur) was given to each subject. Seven days later, a further blood sample was collected to assess the Ab response to the rabies vaccine. After sampling, a second dose of the rabies vaccine was given to each subject. On the final day of the study (day 14), a further blood sample was collected for the analysis of Ab response to the typhoid vaccine and to the second dose of the rabies vaccine. Throughout the study, vaccines were stored in refrigerators at the study site and only removed immediately prior to vaccination, hence maintaining the cold chain. The same batch of vaccines was used for the entire study.
Laboratory analyses. Anti-rabies Ab titres were calculated at the Central Veterinary Laboratories, Surrey, UK using the fluorescent antibody virus neutralization method, as described by Cliquet et al. (1998). Anti-Vi IgG analysis was conducted at the Laboratory of Developmental and Molecular Immunity, National Institutes of Child Health and Human Development, Bethesda, USA as described by Kossaczka et al. (1999). Full details of all laboratory procedures can be found elsewhere (Moore et al. 2004).
For the Gambian studies, scientific and ethics approval was granted by the MRC Scientific Coordinating Committee, and the joint Gambian Government and Medical Research Council Ethics Committee. Informed consent was obtained from the parents/guardians of all the participating children. Approval for the Pakistan study was granted by the Medical Ethics Committee for Research, King Edward Medical College, Lahore, Pakistan and by the Ethics Committee of Göteborg University, Sweden. The study was conducted with informed consent from all subjects.
Before analysis, all vaccine data were log transformed and results of absolute titres are presented as geometric mean values and 95% confidence intervals (CIs). The response variable used throughout the analysis was the difference between the logarithms of the post- and pre-vaccination Ab titres. For Ab response to rabies vaccine in Gambia study 1, subjects reporting previous exposure to rabies (animal bite with post-exposure vaccination) were excluded from the study, and hence pre-vaccination Ab titres were not measured. Data are therefore presented as the difference between the Ab response to the first and second dose of the vaccine. In Pakistan, while subjects reporting previous exposure to rabies were also excluded from the study, pre-vaccination titres were measured and data are presented as the difference between pre-vaccination levels and levels after the first dose of the vaccine and also the difference in levels after the first and second dose of the vaccine. Infant responses to diphtheria toxoid and tetanus toxoid (Gambia study 2) are presented as the difference between Ab levels measured on serum samples collected immediately before vaccination at week 8 and week 16. Responses were measured after two rather than three doses of the vaccine based on the pre hoc assumption that this would maximize variance in strength of response between individuals, with the three-dose schedule aiming to generate a maximal response among all recipients. To avoid making too strong an assumption about the relationship between the pre- and post-vaccination levels, we controlled each analysis for the pre-vaccination level centred on its mean (i.e. we fitted the pre-vaccination level minus the mean pre-vaccination level as a covariate). The exception to this was hepatitis B for which HBsAg Ab levels were measured in samples collected from the infants at 1 year of age. Samples were analysed at this time point to give a measure of Ab persistence, and the unadjusted post-vaccination responses were therefore analysed.
In the Gambian studies (both of which covered a period of approximately 1 year), the monthly variation (seasonality) was fitted using the first two pairs of Fourier terms: sin(θ) and cos(θ) and sin(2θ) and cos(2θ), where θ is the angle in radian representing the date in terms of its position in the annual cycle (on 1 January θ = 2π/365; on 31 December θ = 2π). These four terms were fitted in pairs referred to below as F1 = sin(θ) and cos(θ) and F2 = sin(2θ) and cos(2θ). Further model nomenclature used is explained by the following examples: X = covariate terms (other than seasonality and pre-vaccination Ab titre); F1|X = effect ofF1 controlled for X; F1 + F2 = joint effect of F1 and F2. As the data from Lahore only covered a 6-month interval from April to September 2002, the monthly variation was fitted using the first four orthogonal polynomials derived from the date of observation.
In all analyses, seasonality was fitted both without and in the presence of potentially confounding covariates. They were rarely significant and their results are not presented in detail. Different covariates were available for each study, as follows: Gambia study 1 – sex, body mass index (BMI) z-score, malaria parasitaemia (binary indicating presence or absence of parasites in a blood smear), plasma ACT and age; Gambia study 2 – sex, weight at birth, gestational age at birth; Lahore; sex, BMI, weight. All were measured at the time of vaccination unless otherwise stated.
The monthly pattern of each vaccine response is illustrated graphically by plotting the fitted ‘seasonality’ (either the F1 + F2 terms or all four orthogonal polynomials as appropriate) vs. month. As the seasonality was fitted to the logarithm of the post-vaccination Ab titre, the antilog of the seasonal pattern is plotted, thus yielding the factor by which the response for each month was multiplied relative to the average. The 98% CIs for the seasonal patterns were calculated from the coefficient variance–covariance matrix estimated from the information matrix.
In order to investigate whether a common seasonality affects all the Ab responses in each study, seemingly unrelated regression was used to fit the all patterns simultaneously (without covariates but including pre-vaccination titre levels) and to compare the fits when the seasonal term coefficients were and were not constrained to be identical. All statistical analyses were performed with stata 9 (Stata Corporation, College Station, TX, USA).
The average responses to vaccination, expressed as the proportional increase between pre- and post-vaccination levels are shown in Table 2. The only response which was not significant was that for tetanus in Gambia study 2.
Table 2. Response to vaccination
Antibody titre data
Covariate and antibody data
Data expressed as the proportional increase between pre- and post-vaccination levels.
†Factor by which the pre-vaccination level is multiplied. ‡It was not possible to estimate the response to Hepatitis B, as we did not have the pre-vaccination antibody levels.
Gambia study 1
Gambia study 2
Rabies dose 1
Rabies dose 2
Study 1, The Gambia
A significant trend, by month of study, was observed for age (P ≤ 0.0001), BMI-SD score (P = 0.0022), Hb (P ≤ 0.0001), malaria-parasite-positive blood films (P ≤ 0.0001), and plasma levels of ACT (P ≤ 0.0001) (Table 3). There was no significant difference in the gender distribution by month of study (P = 0.23).
Table 3. Subject characteristics by month and year of study, study 1, The Gambia
Figure 1 shows Ab responses to all four serotypes of the pneumococcal vaccine and response to the rabies vaccine by month of study. With the exception of serotype 23, month of study influenced response to vaccination: type 1 shows a simple sinusoidal pattern peaking in November (F1: P = 0.024), an effect that was enhanced by controlling for the covariates (F1|X: P = 0.010; F1|F2 + X: P = 0.019), although the basic pattern of response is unaltered. The response to type 5 shows bimodal seasonality: it was only the F2 terms that were significant (F2|F1: P = 0.030). This effect remained essentially unaltered after controlling for covariates (F2|F1 + X: P = 0.020). Type 14 has a more complex response (F1 + F2: P = 0.0012), which was lost on controlling for covariates. Although there was a loss of 22 points due missing covariates, these were apparently not entirely responsible for the seasonality seen before adding covariates to the model as it remained when these missing points were excluded (F1 + F2 among those with no missing covariate data: P = 0.0037). Of the covariates, age was not the only one to have a strong independent effect (age |F1 + F2: P = 0.004) but also reduced the effect because of seasonality more than any other individual covariate. Type 23 had a weak but significant simple sinusoidal effect (F1: P = 0.02) but this disappeared on controlling for F2 and/or covariates.
The response to the rabies vaccine was strong and complex. Although the simple sinusoidal fit was significant (F1: P = 0.037), the best fit was achieved when both pairs of Fourier terms were fitted (F1 + F2: P = 0.013). This complex effect remained after controlling for the covariates (F1 + F2|X: P = 0.024), although the simple sinusoidal seasonality did not (F1|X: ns).
The data do not support the hypothesis that there is a common seasonal pattern that applies to all vaccine responses: separate F1 + F2 models for each response gave a significantly better fit than a common F1 + F2 model (P = 0.0034) in seemingly unrelated regression. Type 1 is the only response that did not peak in July/August and this serotype was always involved whenever pair wise comparison of responses proved significant.
Study 2, The Gambia
Table 4 details the characteristics of the infants recruited into the study, according to month of birth. Mean birthweight was 2855 g, and no significant association was observed with month of birth (P = 0.7245). None of the other anthropometric measurements made at birth were related to month of birth. Mean gestational age was 38.6 weeks, and a significant association was observed with month of birth (P ≤ 0.0001), with a trend towards shorter gestational ages in the wet season births. Of the cord blood micronutrients measured, month of birth was significantly associated with plasma levels of vitamin C only (Table 4). No significant associations were observed with plasma zinc [mean (SD); 13.8 (3.34) μmol/l, P = 0.5048] or plasma retinol levels [0.619 (0.21) μmol/l, P = 0.1339]. No significant association was observed between month of birth and cord blood levels of ACT (Table 4).
Table 4. Subject characteristics by month and year of study, study 2 The Gambia
Gestational age (weeks)
Vitamin C (μmol/l)
Pre-immunization Ab titres to diphtheria toxoid were universally very low or undetectable, so titres measured at 16 weeks (4 weeks after the second vaccine dose) were compared after adjustment of the week 8 levels. Ab titres were higher in the female infants than in the male infants (1605 vs. 1070 mIU/ml, P = 0.0366), but there were no significant relationships with any of the other infant characteristics measured. No significant seasonal pattern was observed by the month of study, either before or after controlling for covariates (Figure 2).
As a consequence of high levels of passively transferred maternal Ab, tetanus toxoid Ab levels were relatively high in cord blood [geometric mean 3195 mIU/ml (95% CI 2632–3880)] and in plasma samples collected at 8 weeks, immediately before the infants received the first tetanus toxoid vaccine [580 mIU/ml (95% CI 484–695)]. At 16 weeks, the geometric mean level was 597 mIU/ml (95% CI 520–686); not significantly higher than the pre-immunization levels (P = 0.83). Furthermore, in 77 (58%) of 132 paired samples at 8 and 16 weeks (4 weeks after the second vaccine dose), the tetanus toxoid Ab level was lower than pre-immunization level. Using Ab levels at 16 weeks, no significant trend was observed according to the month of vaccination (Figure 2), either before or after controlling for covariates.
Data were available for anti-HBsAg Ab levels measured at 52 weeks of age for 121 (88%) of the infants. The geometric Ab titre was 77.1 mIU/ml (95% CI 56.3–105.6), and no significant associations were observed between Ab levels and other infant characteristics. No significant association was observed between response to vaccination and month of administration (Figure 2) before or after controlling for covariates.
The mean (SD) age of the Pakistani adults included in the current study was 29.4 (4.75) years, and 56.8% were males. BMI ranged from 14.1 to 46.0 kg/m2, with a mean (SD) value of 24.2 (5.27) kg/m2. There was no significant difference in the distribution by month of study for gender (P = 0.0954), age (P = 0.6329) or BMI (P = 0.1708).
As the data from the study in Lahore only covered a 6-month period, we fitted the monthly variation using orthogonal polynomials. Although polynomials up to degree four (z1, z2, z3 and z4) were fitted, only the first two were ever significant. Typhoid responses increased slightly between May and October (z1|z2: P = 0.032) with the Ab level increasing on average by 23% per month (95% CI: 4.5–46). This effect was undiminished when controlled for covariates (z1|z2 + X: P = 0.028), with Ab level increasing on average by 22% per month (95% CI: 2.9–44%). The monthly variation in response to the first rabies vaccination was not significant, either before or after controlling for covariates. However, the much stronger boost because of the second dose did significantly vary with month, peaking in August (z1 + z2: P = 0.0016). This effect was unaltered after controlling for covariates (z1 + z2|X: P = 0.004). Fitting the z1 + z2 + z3 model with seemingly unrelated regression revealed some evidence that the typhoid and rabies (dose 2) effects differed significantly (P = 0.021). Antibody responses to both typhoid and rabies vaccination are shown in (Figure 3)
A number of factors are thought to influence an individual's response to certain commonly used vaccines. Such factors include age, gender, nutritional status and infection at the time of vaccination. The current study has added to this evidence, with data to suggest that month of administration in two countries with a pronounced seasonality influences the serum Ab levels generated in different population groups and using a number of different vaccines.
A seasonal variation in markers of immune function has been reported in studies from West Africa. In The Gambia, we observed a seasonal variation in blood leucocyte and lymphocyte subpopulations measured during infancy, with higher lymphocyte and leucocyte counts in the rainy season (July–December), when compared with the dry season (January–June) (Collinson 2002). In a study of infants and children from Guinea-Bissau, however, Lisse et al. (1997) actually reported the opposite with lower absolute and percentage lymphocyte counts observed in samples collected during the annual rainy season. In addition, from Guinea-Bissau, Shaheen et al. (1996) found a seasonal variation in cell-mediated immunity, with a rise in anergy (a negative skin test response observed to all test antigens) observed during the rainy season. In The Gambia, we also saw a variation by season of measurement in the size of the thymus, as assessed by ultrasonography. At each age point (1, 8, 16 and 52 weeks of age) the mean thymic index was significantly smaller in infants when measured in the wet season (Collinson et al. 2003). This observation correlated with a seasonal variation in breast milk IL-7 levels, suggesting a possible role of breast milk immune factors in promoting thymic development (N'Gom et al. 2004). However, the studies reported in this paper are the first example we can find in the literature to report that month of administration of some commercially produced vaccines influences Ab response.
There are a variety of seasonally related factors that could account for this observed difference in Ab response. First, it is possible that it could be a consequence of methodological problems directly related to month of vaccination, such as differences in ambient temperature affecting the cold chain of the vaccines. However, this is considered highly unlikely in the current studies, as the vaccines were stored and administered under research protocol standards. In addition, batch differences in the vaccines cannot account for these observations, as in both the child cohort in The Gambia and the adult cohort in Pakistan where seasonal associations were observed, the same batch of vaccines was used for all subjects.
It is possible that the observed association is a consequence of the seasonal pattern in infectious or background antigen exposure in both these population groups. In rural Gambia, there is a profound seasonality in malaria transmission, with highest rates of hospital admissions observed in the rainy season months of October and November (Brewster & Greenwood 1993). Seasonal differences in either parasitaemia or malaria prophylaxis/treatment could therefore influence the Ab response to vaccination. Indeed, previous studies have shown an association between acute malaria and a decreased response to certain vaccines, including diphtheria and tetanus toxoid, and to meningococcal, Salmonella and Hib vaccinations (reviewed in Rosen & Breman 2004). While this could be a proposed mechanism for the associations observed, it seems unlikely, as children in The Gambia having blood films positive for parasites did not show a decreased response to either of the vaccines administered. In addition, plasma levels of ACT, indicative of an acute inflammation, did not account for the seasonal variation in Ab responses. Furthermore, and while a limitation of the current analysis is the truncated nature of data collection in the Pakistan study, with data collected between the months of April and September only, the relevance of a month by month variation in Ab response in a population where there is little malaria (Zaman et al. 1993), makes it unlikely that malaria infection is the primary causal factor in the observed association with month of vaccination. Of interest, and as malaria is known to affect the maternal Ab transfer (Duffy 2003), it may appear surprising that no association was observed between month of vaccination and pre-immunization levels to tetanus toxoid. However, malaria prophylaxis is prescribed routinely to pregnant women during the malaria season in this region of The Gambia, and very few placentas had evidence of active infection at the time of delivery and only a minority had evidence of malaria infection in earlier pregnancy (Collinson 2002). The large variance in cord and pre-immunization tetanus Ab levels coupled with small sizes of malaria-infected and -uninfected subgroups would, therefore give a low likelihood of detecting a seasonal effect of gestational malaria in this population cohort.
The associations with month of vaccination could also be due to other seasonally dependent infectious agents or environmental antigens. Studies from The Gambia have reported seasonal variations in the incidence of clinical presentation with both respiratory infections (Forgie et al. 1991; Brewster & Greenwood 1993) and diarrhoeal disease (Rowland et al. 1985; Brewster & Greenwood 1993), both showing peaks during and shortly after the annual rainy season. In Pakistan, a seasonal variation has also been reported in both upper and lower respiratory tract infections, with the reported incidence being highest during the cooler months of November to February (Zaman et al. 1993). In addition, a seasonal variation has been observed in the presence of diarrhoeal disease, with the maximum incidence observed during the months of April to June (Mahmud et al. 1993). While less is known about the seasonality of specific helminthic infections at either of the study sites, it is also possible that the current observations are the consequence of immune priming by infestations with worms.
Of specific relevance to the current study, a seasonal pattern in pneumococcal carriage has been observed in both infants and their mothers in The Gambia (Momodou Darboe, personal communication), with an apparent reduction in carriage during the months of May to August. Although the data are not directly comparable (collected in different years and from different subjects), this reduction in rates of carriage does seem to coincide with a reduced Ab response to the pneumococcal vaccine, and hence may suggest a priming of vaccine response by carriage. Priming by vaccine-specific Ab carriage cannot, however, explain the observed seasonal response for vaccines where no pre-vaccination carriage occurs (e.g. rabies). Further modelling of specific disease patterns in relation to Ab response may help elucidate whether the observed associations have an infectious aetiology.
There is some evidence in the literature to suggest that nutritional status may influence immune function, although such findings are not universal. It is also possible, therefore, that the observed difference is due to seasonal changes in the nutritional status of individuals living in these two areas. In rural Gambia, there is a pronounced seasonality in many markers of nutritional status: in energy balance (Prentice et al. 1981) and the status of certain micronutrients including vitamin A and its precursors (Bates et al. 1994), riboflavin (Bates et al. 1994), vitamin C (Bates et al. 1982,1994), iron (Bates & Prentice 1999) and folate (Bates et al. 1994). Although there are few published data describing the seasonality of nutritional status in urban Pakistan, it is unlikely that the variability of seasonally available foods will influence nutritional status in the population, and particularly in the urban poor. Studies looking specifically at vaccine response indicate that, while Ab production may be impaired in severe malnutrition (Powell 1982), in subjects with less severe nutritional deficiencies, there is no impact of nutritional status on Ab response to vaccination (Greenwood et al. 1986; Lakshami et al. 2000). Indeed, we have published elsewhere a detailed study to show that the Ab response to vaccination in rural Gambia is not related to nutritional status (Moore et al. 2003). It is therefore unlikely that the seasonal swings observed in nutritional status are responsible for the observed differences in Ab production to vaccination.
It is interesting that there were no associations between month of administration of the vaccine and Ab levels in the infant cohort from The Gambia. One possible reason is that the sample size in each group was too small, and that a larger population sample would be required to observe an association. Optionally, it may be that the strength of the Ab response to the specific vaccines used in the infant study (standard EPI vaccines) is not affected by the factors influencing the trends observed by month of vaccination in the other two studies. It is also possible, however, that an association with season includes the dimension of age, and that in infants, whose immune system is still relatively naïve, the level of the antigen exposure induced on vaccination is potent enough to induce a strong response in all, irrespective of other seasonally dependent factors. Alternatively, the diphtheria and tetanus Ab levels were measured in children at 16 weeks; an age at which most infants in this community are still predominantly breast fed and quite well protected from the morbidity and antigenic exposure associated with poor weaning practices. However, the fact that no association was observed with Hepatitis B Ab levels at 52 weeks of age suggests that the protective effect of breast milk is not the only factor influencing the lack of an association with month in this cohort. Finally, failure to detect a true effect in this age group might be a consequence of our choice of sample time points in relation to vaccine doses.
Understanding these associations could be aided by understanding the specific mechanisms of Ab generation to vaccination, as it is possible that the observed difference in response is a consequence of a selective response by certain types of vaccines. We used vaccines that can be immunologically classified as T-lymphocyte dependent (rabies, tetanus, diphtheria, Hepatitis B) or T-lymphocyte independent (typhoid and pneumoccocal). Indeed, from study 1 in The Gambia, the patterns of Ab response to the rabies and pneumococcal vaccines do not correlate, and it may be that different priming mechanisms are involved. Furthermore, we did not observe a single common pattern for the four different pneumococcal serotypes measured, possibly suggesting that differing patterns of exposure and carriage within this Gambian environment can drive later Ab-specific responses. Comparison of the effects of month of administration of vaccines from other studies in such seasonally affected population groups may help explain this observation.
It is also possible that we are observing an ‘adjuvant’ effect on Ab response. Adjuvants, such as aluminium hydroxide (alum) and calcium phosphate, are intentionally used in vaccines to elicit an early, high and long-lasting immune response. It is plausible therefore that other seasonally dependent non-infectious environmental antigens are having an adjuvant effect on the vaccine, and hence boosting the Ab response at specific times of the year. The apparent rise in Ab response to the second dose of the rabies vaccine during the months of July and August in Lahore supports the hypothesis of an adjuvant effect; an environmental exposure occurring most strongly in these two months might be driving the pronounced effects on Ab responses at this time. Elucidating and understanding the cause and mechanisms for this could be helpful in vaccine development and usage in such population groups.
In the current analysis, we have demonstrated a significant variation in Ab response to vaccination according to month of vaccination in studies from rural Gambia and urban Pakistan. Despite consideration of the many seasonally varying environmental factors, the aetiology for this observation is still not clear. We have not yet followed up any of the subjects included in the current analysis to explore whether the observed variance in Ab levels shortly after vaccination is maintained in the future, but this would be of interest. It would also be useful to understand which other vaccines are related to month of administration, and whether this is just a phenomenon unique to countries where the pattern in seasonality is known to influence many factors in life. Indeed, the specific impact of a seasonally defined reduced response to vaccination would need to be considered on a case-by-case basis. If seasonally related deficits were detected, it might be appropriate to consider selective additional booster vaccine doses for children immunized in a month/season during which responses have been predicted to be low. However, and as the majority of vaccines used in vaccination programmes such as the EPI elicit an appropriate response from the majority of recipients, this finding may be most relevant in the interpretation of vaccine trial data from such population settings and in understanding in more detail mechanisms for potential seasonal interactions at an immunological level.
This work was funded by the UK Medical Research Council (Gambia) and the Nestlé Foundation (Pakistan). We thank the study participants for their willingness to participate in this research and the local field staff at each site for their help with the studies. We are grateful to Dr Tony Fooks and Trudie Goddard from the Central Veterinary Laboratories and Dr Shousun Chen Szu from NIH for their help and advice with the analysis of the rabies and Vi antibody levels. We also thank Dr Susanna Schlegel and her collaborators from the WHO Collaborating Center for Neonatal Vaccinology for the quantification of anti-tetanus and diphtheria antibodies.