A multi-country, multi-year, meta-analytic evaluation of the sex differences in age-specific pertussis incidence rates

Background Pertussis is frequently reported to be more common in females than in males. However, the variability of the sources of these observations makes it difficult to estimate the magnitude and consistency of the sex differences by age. To address this question, we used meta-analytic methods to analyze pertussis national incidence rates by sex and age group from nine countries between the years 1990 and 2017. Methods For each age group, we used meta-analytic methods to combine the female to male incidence rate ratios (RRs) by country and year. Meta-regression was performed to assess the relative contributions of age, country and time-period to the variation in the incidence RRs. Results The pooled female to male incidence RRs (with 95% CI) for ages 0–1, 1–4, 5–9 and 10–14, were 1.03 (1.01–1.06), 1.16 (1.14–1.17), 1.18 (1.15–1.22), 1.15 (1.11–1.18) respectively. For the ages 15–44, 45–64 and 65+ they were 1.65 (1.58–1.72), 1.59 (1.53–1.66), 1.20 (1.16–1.24), respectively. While there were some differences between the countries, the directions were consistent. When including age, country and time in meta-regression analyses, almost all the variation could be attributed to the differences between the age groups. Conclusions The consistency of the excess pertussis incidence rates in females, particularly in infants and very young children, is unlikely to be due to differences in exposure. Other factors that impact on the immune system, including chromosomal differences and hormones, should be further investigated to explain these sex differences. Future studies should consider sex for better understanding the mechanisms affecting disease incidence, with possible implications for management and vaccine development.


Introduction
Despite the availability of an effective vaccine, pertussis caused by Bordetella pertussis (B. pertussis) remains a public health problem in both developing and developed countries [1]. Clinical manifestations of the disease can be mild, severe with occasional fatal outcomes, especially in infants [2]. The appearance of new cases could be, for example, due to low immunization rates [3] inadequate immune responses to vaccine [4] or waning immunity following immunization [5].
Reports from individual countries often mention higher pertussis incidence rates (IR) in females [6][7][8][9][10][11], not always specifying age groups. These observations have usually based on data from individual countries or using case series data, without denominators for calculating incidence rates (IR). If the excess in females is consistent, it could be due to a number of factors.
These include response to vaccine, behavioral and social factors [12], chromosomal factors, or an interaction between sex hormones and immune function affecting the clinical manifestations of the disease [13]. Establishment of the magnitude and consistency of the sex differences in the disease can provide important clues to the mechanism of disease. In order to study this further, we carried out an in-depth study of the sex differences in pertussis incidence rates at different ages, in different countries and over a number of years, based on national data.

Source of data and search strategy
In order to guarantee the data quality, we restricted our search strategy to all countries in Europe, North and South America, Australia and New Zealand, with established diagnostic tools and well-organized mandatory reporting systems, that provide data by age and sex for a number of years. National data were obtained either from official internet sites or by contacting representatives of the appropriate country health authorities. The original search was performed from March to June of 2018.
Netherlands from Statistics Netherlands' database [26], for New Zealand from Stats NZ, Infoshare, New Zealand [27], and for Spain from the Demographic Statistics Database [28].

Ethical considerations and informed consent
National, open access aggregative and anonymous data were used and there was no need for ethics committee approval.

Statistical analyses
The period under study was between 1990 and 2017. Due to the large amount of data, for presentation purposes, the years were grouped for the graphical presentations. Annual pertussis incidence rates (per 100,000) were calculated by sex and age group, for each country and group of years using the number of reported cases divided by the respective population size and multiplied by 100,000.
The age groups considered were <1 (infants), 1-4 (early childhood), 5-9 (late childhood), 10-14 (puberty), 15-44 or 15-39 (young adulthood), 45-64 or 40-59 (middle adulthood) and 65+/60+ (senior adulthood) years. The surveillance systems in Canada, England, Finland, Netherlands, and New Zealand used similar age groups except for the following: 15-39, 40-59 and 60+. For Australia and Finland, data are missing for ages <1 and 1-4 separately. We made an informed decision not to combine these age groups since there is a difference between infants <1 year old and early childhood. The female to male incidence rate ratio (RR) was calculated by dividing the annual incidence rate in females by that of males, by age group, country, and time periods.
The data were analyzed using meta-analytic methods and meta-regression STATA software version 12.1 (Stata Corp., College Station, TX). For the purpose of applying meta-analytic methods, the national data sets for each age group by country and year were considered as separate "studies" and the outcome variable was the female to male incidence RR. After obtaining pooled incidence RRs separately for each age group, by country and time period, pooled incidence RR's for each age group were obtained for all countries and time periods together. The results are presented in forest plots. Heterogeneity was evaluated using Cochran's Q statistic, and Tau 2 and I 2 (to estimate between-study variance) [29]. Where significant heterogeneity was present (if I 2 �50% and/or the Q test yielded a p-value <0.1) the random effects model [30] was used to estimate pooled RRs and 95% confidence intervals (CI). Otherwise, the fixed effects model was used.
We performed leave-one-out sensitivity analysis in order to determine how each country and group of years affected the outcome following the recomputed pooled pertussis female to male incidence RR. In order to determine whether there were countries or time periods that are outliers, we created funnel plots and used Egger's test. In order to explore the contributions of age, countries and time periods to the heterogeneity of the incidence RRs, meta-regression analyses were performed, with incidence RR as the dependent variable.

Results
The summary of male and female pertussis incidence rates (per 100,000 populations) in different countries for each age group and for number of years is presented in Table 1.
Age-specific rates by sex were highest in female infants, in 1-4-year-olds and in age groups of 5-9 and 10-14. There was a decrease in the incidence rate of pertussis in adults, in both groups of males and females. Results of the study are presented in the forest plots presented by age group in Figs 1-7 (with CI = 95% confidence interval, RR = rate ratio. The right side of the X-axis indicates a higher IR for females and the left side for males).
The forest plot for infants (age <1) is shown in To evaluate the effect of individual country and the group of years on the pooled RR, we performed leave-one-out sensitivity analysis and recomputed the pooled RRs. After omitting one country at a time, the pooled RRs remained very similar (Table 2).
For the funnel plot (Fig 8), Egger's test p value for asymmetry was not significant for all age groups except middle adulthood (from infancy, young childhood, late childhood, puberty, young adulthood, and senior adulthood p value were p = 0.711, p = 0.427, p = 0.217, p = 0.176, p = 0.055 and p = 0.076 respectively). Evidence of asymmetry was observed only for middle adulthood with p = 0.036.

Discussion
In this study, we examined the sex differences in pertussis incidence rates by age group in nine countries over a period of six to 27 years. These results revealed higher pertussis incidence rates in females than in males in all age groups from infancy to older adults. The pooled results varied by age from a 3% excess in infants to an excess of 65% in in young adulthood. These findings were consistent over countries and time periods. The meta-regression results revealed that among the variables, age group contributed almost all the variation in the incidence RRs. The results of this study contrast with the perception that males suffer more than females from infectious diseases [13,31].
Surveillance data from 1995 onwards in former West German states showed a higher pertussis incidence in females (overall 60% of cases) than in males mainly due to a higher proportion of females among adult cases [4]. In England, slight differences were observed between

PLOS ONE
Female excess in pertussis incidence rates Forest plot of the female to male pertussis incidence rate ratios (RR) for different years in Canada, Czech Republic, England, Israel, Netherlands, New Zealand, and Spain in yearly childhood. The overall incidence RR in ages 1-4 was 1.16 (95% CI 1.14-1.17), which indicated a 16% excess incidence rates in females, with low heterogeneity, I 2 = 24.3%, and Tau 2 = 0.001. The subtotal incidence RRs varied from 1.08 in New Zealand to 1.30 in Czech Republic. The forest plot for age 5-9 is shown in Fig 3. https://doi.org/10.1371/journal.pone.0231570.g002

PLOS ONE
Female excess in pertussis incidence rates

PLOS ONE
Female excess in pertussis incidence rates pertussis were observed in males and females under the age 12 [10]. In Alberta, Canada, between 2004 and 2015, incidence rates by sex in children under the age 14 were similar between females and males [11]. Unlike in the present study, in general, the sex differences in incidence rates were not reported by age or whether they were consistent over time periods. In some, studies were based on hospital or local data, without population denominators [6][7][8][9][10]. This could be an important source of selection bias.
The current study is based on national data with very large populations, covering a number of years and consequently with large numbers of cases. Selection bias has been minimized by using national data over different time periods, which should be representative of each country. Relevant denominators were available to compute incidence rates as opposed to studies based on a case series. The inclusion of nine countries, with advanced health system, allowed us to evaluate the consistency of the findings over different populations and many years. There is no evidence to suggest that there is selective care or differences in vaccine coverage according to the sex of the child, in any of the countries in this study. Underreporting, as the result of non-specific clinical manifestations of the disease and the lack of laboratory confirmation, may be a source of information bias [32]. However, this is unlikely to be different for females and males. There may be a difference in use of health services by sex in the adult age groups [33], but is unlikely to be a factor in infants and children in the countries in this study. Surveillance systems as well as the diagnostic criteria and proportion of laboratory-confirmed cases are heterogeneous [34], but should not differ between females and males.
There is no clear evidence on differences in response to pertussis vaccine between males and females. Antibody levels have been found to be similar in males and females in infants, children and adults following immunization [35,36]. As regards exposure differences, in young and middle adulthood, women may have more exposure to cases of pertussis while Similar results were obtained after dropping one group of years at a time (Table 3). https://doi.org/10.1371/journal.pone.0231570.t002

Fig 8. Funnel plots: A) for infants, B) for early childhood, C) late childhood, D) for puberty, E) for young adulthood, F) for middle adulthood and G) for senior adulthood.
In the meta-regression analyses, including age group, country and year, age group contributed almost all the variation in the incidence RRs. For infants the incidence RR was lower than for other age groups and in young and middle adulthood, the incidence RRs were significantly higher than in the other age groups (P < .0001). There was no significant association with reporting time-periods, aside from a borderline negative trend among puberty and middle adulthood groups. In those groups, as the timeperiods increased, the incidence RR values decreased (P = 0.05 and P = 0.05 for puberty and middle adulthood respectively). In this group, as the time-periods increased, the incidence RR values decreased, (P = 0.02). https://doi.org/10.1371/journal.pone.0231570.g008

PLOS ONE
caring for their own children [12], or through exposure to sick children while working in daycare centers. Adults are a potential reservoir for exposure to pertussis in very young infants [2], although the exposure should be the same regardless of the infant's sex. Such possible sex differences in exposure, vaccination rates [37] or medical services utilization are not relevant explanations for the excess pertussis incidence rates observed in infants and young children.
While this study cannot provide information on the mechanisms underlying the excess incidence rates in females, we can explore some possible explanations. Sex differences in pertussis incidence rates can be due to factors such as biological differences between sexes, such as sex chromosomes and sex hormones. It could be postulated that genetic and/or hormonal differences explain, at least partly, increased pertussis incidence rates in females. Studies indicate that infection with B. pertussis results in an immune response mediated through expansion of Th17 cells [38,39]. These cells may induce tissue immunopathology [40] via the production of inflammatory cytokines and the creation of an environment contributing to inflammation of the upper respiratory tract, duration of lung tissue pathology and prolonged cough [40,41].
Differences in the immune responses between males and females are in part attributed to the X chromosome, which contains a high number of immune-related genes and regulatory factors that are involved in both the innate and adaptive immune responses [42,43]. X-linked mosaicism encourages a highly polymorphic gene expression that could enhance the immune response more in females [43], with consequent more symptomatic pertussis. Thus, a stronger immune response in females could result in more clinical manifestations of pertussis.
Sex hormones may also be implicated in the higher incidence rates of clinical pertussis in females. Higher pertussis incidence and immune response may also be due in part to an estrogen mediated enhanced pro-inflammatory response to B. pertussis invasion via IL-17 and a cytokine storm phenomenon. Progesterone and estrogen are lead to more severe inflammation in respiratory diseases [44,45] and an increased expression of IL-17, whereas testosterone [46] reduces the generation of Th17 cells. Kuwabara T et al [47] showed that IL-17 plays an important role in chronic inflammation that occurs during the pathogenesis of autoimmune diseases such as human rheumatoid arthritis and multiple sclerosis (MS). Pertussis toxin served as adjuvants to induce sensitization to neural antigens in experimental autoimmune encephalomyelitis, the principle animal model of MS, which is more common in female. It appears that IL-17, the cytokine that is involved in pertussis pathogen eradication [38] and autoimmune diseases pathways [47], along with associated chemokines IL-1β, IL-23R, IL-6 and many others [44,47] is significant and may be linked to an estrogen-regulated immune overresponse to pertussis infection in female. The impact of sex hormones on the immune response prior to puberty especially in infancy, is not clear. Maternal hormones that pass through the placenta affect male and female fetuses equally [48]. The mini-puberty phenomenon in infancy results in higher endogenous estrogen levels in female infants [49], which could explain the higher incidence of disease in female infants. In addition, maternal hormones may persist in the infant's circulation for some months after birth and will affect females and males equally [48]. This could mitigate the sex difference in disease and explain the lower female to male incidence ratios in infants than those seen at older ages.
Differences in sex hormone levels continue in childhood [50] and in pre-pubertal children [51].
It is conceivable that, in young adulthood, hormonal and genetic differences continue to exist, but the excess pertussis incidence rates need to be viewed in the context of possible different exposure. It has been noted that IL-17 blood levels increase in pregnancy [52,53].The available literature [54] indicates that the immune response of aged women may be preserved to a greater extent than in aged men and may contribute to prolonged inflammatory responses and tissue damage in respiratory airways. This could be the reason why women exhibit a higher pertussis incidence rate even in older ages.

Conclusions
This study has provided strong evidence that while the excess female incidence rates for pertussis observed in all age groups differ in magnitude, they are consistent over a number of countries and over different time periods. The mechanism underlying the excess in females is still largely unknown. Behavioral factors may contribute to some of the differences seen in the postpubertal age groups. However, in infants and children, genetic factors, as well as sex hormones could play a part. Our findings suggest the need to explore further the role of sex differences in the mechanism of pertussis infection, when evaluating the efficacy of pertussis vaccine dosing and schedules especially in adult females for disease prevention and public health promotion.