Phase-out of smallpox vaccination and the female/male HIV-1 prevalence ratio: an ecological study from Guinea-Bissau.

OBJECTIVE: In Guinea-Bissau, West Africa, we observed that having a smallpox vaccination scar was associated with lower HIV-1 prevalence, more strongly for women than men. If this represents a causal effect, the female/male HIV-1 prevalence ratio would increase for birth cohorts no longer receiving smallpox vaccination due to the phase-out of this vaccine.
DESIGN: An ecological design using HIV surveys and information about smallpox vaccination coverage.
SETTING: Urban and rural Guinea-Bissau. PARTICIPANTS: Participants in HIV surveys were grouped into an age group with decreasing smallpox vaccination coverage (15-34 years) and an age group with steady smallpox vaccination coverage (≥35 years).
INTERVENTIONS: The exposure of interest was the phase-out of the smallpox vaccine in Guinea-Bissau. PRIMARY AND SECONDARY OUTCOME MEASURES: HIV-1 prevalence.
RESULTS: At both sites, the female/male HIV-1 prevalence ratio increased by calendar time for the age group with decreasing smallpox vaccination coverage; the combined female/male HIV-1 prevalence ratio among people aged 15-34 years was 1.00 (95% CI 0.17 to 5.99) in 1987-1990, 1.16 (95% CI 0.69 to 1.93) in 1996-1997, 2.32 (95% CI 1.51 to 3.56) in 2006-2007 (p value for no trend=0.04). There was no increase in the female-to-male HIV-1 prevalence ratio for the age group >35 years with steady smallpox vaccination coverage; 1.93 (95% CI 0.40 to 9.25) in 1987-1990, 1.32 (95% CI 0.83 to 2.10) in 1996-1997, 0.81 (95% CI 0.56 to 1.16) in 2006-2007 (p value for no trend=0.07).
CONCLUSIONS: Thus, data was compatible with the deduction that the phase-out of smallpox vaccination may have increased the susceptibility to HIV-1 relatively more for women than men. Hence, phasing out smallpox vaccination may have contributed to the global increase in the female/male HIV-1 prevalence ratio among young individuals. Due to the potential fallacies of ecological studies, the results should be interpreted carefully, and this hypothesis needs further assessment. If the hypothesis is true, studies of smallpox vaccination could inform HIV-1 vaccine research. © Author(s) (or their employer(s)) 2019. Re-use permitted under CC BY-NC. No commercial re-use. See rights and permissions. Published by BMJ.

The ecological design allowed us to assess a deduction of the hypothesis: smallpox vaccination has a protective effect against HIV-1 which is stronger among women than among men.

To increase the robustness of the results, we utilised parallel data from both urban and rural Guinea-Bissau.

Ecological studies should be interpreted carefully as spurious associations can arise, and thus triangulation with existing studies are necessary to further assess this hypothesis.

Introduction

Vaccination against smallpox infections was stopped globally in 1980 following the eradication of smallpox in 1977. It has been reported that smallpox vaccination reduced susceptibility to unrelated infectious diseases,1 and in immunological in vitro studies, smallpox vaccination was associated with an up to fivefold reduction in C-C chemokine receptor 5 (CCR5) tropic HIV-1 replication.2 Based on vaccination scar readings in Guinea-Bissau and school health records in Denmark, we have shown that smallpox vaccination (and BCG) was associated with a lower risk of HIV-1.3 The adjusted OR for HIV-1 infection was 0.52 (95% CI 0.32 to 0.84) for women and 0.77 (95% CI 0.48 to 1.24) for men. This association was stronger for women, who had received multiple smallpox vaccinations (OR of 0.18 (95% CI, 0.05 to 0.64)).

We hypothesised that smallpox vaccination has a stronger protective effect against HIV-1 in women than men. If this is the case, a logical deduction is that the female/male HIV-1 prevalence ratio should increase for age groups with decreasing smallpox vaccination coverage while there would be no change in the female/male HIV-1 prevalence ratio for age groups with steady smallpox vaccination coverage. By using a female/male HIV-1 prevalence ratio, we could to some extent disregard calendar time trends—such as the general spread of HIV-1 worldwide and the increased focus on prophylaxis and treatment—affecting both sexes and all age groups. We tested the hypothesis in two cohorts followed with sequential HIV surveys in Guinea-Bissau since the late 1980s.

Methods

In this ecological study, we compared the changes in smallpox vaccination coverage with the change in female/male HIV-1 prevalence ratio for the age groups that were between 15 and 34 (a decreasing smallpox vaccination coverage) and ≥35 years (a steady smallpox vaccination coverage) over a 30-year period. All analyses were based on aggregated data from different sources: We used reported HIV-1 prevalence surveys in Bissau, the capital of Guinea-Bissau, and Caió, a rural district of Guinea-Bissau.4 5 We used individually-based data from a smallpox vaccination scar survey in Bissau in 2005 to model smallpox vaccination coverage at the time of the HIV-1 surveys.

Patient and public involvement

As this study was based on previously published data,4–6 neither patients or the public were involved in conducting this research.

Estimates of smallpox vaccination coverage

Smallpox vaccination typically leaves a distinct vaccination scar. We used a cohort of individuals, who had both participated in a smallpox vaccination scar survey (2005–2007) and an HIV prevalence survey (2004–2006) conducted in Bissau (previously published6) to model the historical changes in smallpox vaccination coverage (see below). The smallpox vaccination scar prevalence is comparable between urban and rural Guinea-Bissau.7 8 In the smallpox vaccination scar survey, field workers examined vaccination scars and interviewed study participants. The field workers examined the study participants’ upper arms for vaccination scars and registered up to five scars. Scars were classified as BCG, smallpox vaccination or ‘uncertain’, based on size, colour and general appearance of the scar.

For each individual in the smallpox scar survey, we calculated the age the individual would have had in the different HIV survey years (1987, 1996, 2005 and 2016). We approximated the age-standardised smallpox vaccination coverage overall and by sex for the years 1987, 1996, 2005 and 2016 in each age group (15–34 and ≥35 years) by dividing the number of individuals with a smallpox vaccination scar by the total number of individuals in each group. The ‘15–34 years’ group was chosen as they have a declining smallpox vaccination coverage over the different HIV surveys. The ‘≥35 years’ group covered ages between 35 and 65 (oldest age registered) and had a steady smallpox vaccination coverage over the different HIV surveys. The smallpox vaccination coverage estimation for the age group ≥35 in 2016 was changed to ≥45 to ensure a steady smallpox vaccination coverage.

A small validation study based on a city register of smallpox vaccination from Bissau showed a sensitivity of 90% (95% CI 80% to 95%) by using smallpox scars as proxies for registered smallpox vaccinations (62 individuals had smallpox scars in community surveys out of 69 registered as smallpox vaccinated in the city register).7

Estimates of female/male HIV-1 prevalence ratios

Three HIV-1 prevalence surveys were carried out in parallel both in Bissau (1987, 1996, 2006)4 and Caió (1990, 1997, 2007).5 An additional survey was carried out in Bissau in 2016.9 In these surveys of randomly selected households, all individuals aged 15 years or older were interviewed and tested for HIV provided they accepted the informed consent. In Guinea-Bissau, injection drug use is virtually non-existent,10 and blood transfusions have been screened for HIV since 19874; thus, HIV-1 is almost exclusively sexually transmitted.

The results of the HIV-1 surveys were reported by sex and by 10 year age groups from 15 years of age. Based on these data, we constructed a dataset with the number of observed individuals by sex, age group (15–34; ≥35) and HIV-1 status for each of the HIV surveys. The reason for the age cut-off of 35 years was that the last smallpox vaccination campaign in Guinea-Bissau was in 1975 and pre-school children were rarely vaccinated7 resulting in a decreasing smallpox vaccination among people aged 15–34 years individuals across HIV survey years. The combined estimates for 2016 were only based on Bissau, as no HIV survey had been carried out in Caió; in this survey, the age range was changed to ≥45 to ensure a steady smallpox vaccination coverage.

The female/male HIV-1 prevalence ratio in two specific age groups was of interest in itself, but also, by using such as comparison, we could to some extent disregard calendar time trends such as the spread, prophylaxis and treatment of HIV-1, which would affect both sexes and all age groups.

Statistical analysis

We used R V.3.3.1 to estimate the female/male HIV-1 prevalence ratios among individuals aged 15–34 years ≥35 years for each HIV survey (CIs were calculated using the ‘epitools’ R package for risk ratios). Individual level-data sets were reconstructed for the surveys based on the summary tables in refs 4 5 9. To estimate the probability of data showing the observed trend in female/male HIV-1 prevalence by the combined HIV survey years (1987–1990, 1996–1997, 2006–2007) by chance, we fitted a logistic regression on HIV-1 status depending on HIV survey year as a linear and quadratic effect, sex and the interaction between a linear effect of HIV survey year and sex. The model was fitted separately for the individuals aged 15–34 and ≥35 years. We interpreted the p value for the interaction between survey year (assumed linear effect) and sex as a test for a homogeneous association between sex and a linear change in HIV-1 prevalence across survey years.

Results

For the age group ≥35 years (≥45 years in 2016), the estimated smallpox vaccination coverage was similar across all the HIV surveys (fluctuating between 66% and 77%, figure 1). As expected, the smallpox vaccination coverage decreased over HIV survey years for the age group 15–34 years (from 62% in 1987 to 0% in 2016, figure 1). There was no indication that the smallpox vaccination coverage differed between women and men (online supplementary figures 1 and 2). The general prevalence of HIV-1 among adults >15 years of age increased from 0% (0/649) in 1987 to 4.6% (118/2548) in 2006 in Bissau and from 0.5% (14/2770) in 1990 to 3.6% (105/2895) in 2007 in Caió. In 2016 in Bissau, the HIV-1 prevalence among adults over 15 was 4.0% (104/2601).

Figure 1

Estimated smallpox vaccination scar prevalence in different age groups for each sequential HIV survey. Based on data from Bissau, Guinea-Bissau, previously published.6 The estimation for the age group ≥35 in 2016 was changed to ≥45 to ensure a steady smallpox vaccination coverage.

As seen in table 1, there was an increase in the female/male HIV-1 prevalence ratio among individuals aged 15–34 years from the earliest to the latest conducted HIV surveys, the pattern being similar in Bissau and Caió. Combined, the female prevalence increased from 0.3% to 4.3% from 1987–1990 to 2006–2007, whereas the male prevalence increased from 0.3% to 1.9% in the same period. The female/male HIV-1 prevalence ratio was 1.00 (95% CI 0.17 to 5.99) in 1987–1990, 1.16 (95% CI 0.69 to 1.93) in 1996–1997 and 2.32 (95% CI 1.51 to 3.56) in 2006–2007. In a logistic regression, the interaction-test for a homogeneous association between sex and a linear change in HIV-1 prevalence across survey years for the individuals aged 15–34 years gave a p value of 0.04. Latest in Bissau in 2016, the female/male HIV-1 prevalence ratio was further increased to 5.41 (95% CI 2.15 to 13.61).

Table 1

The female/male HIV-1 prevalence ratio (PR) by age group, survey year and study site

Age group	15–34 years (decreasing smallpox vaccination rates in later years; see figure 1)				≥35 years (steady smallpox vaccination rate over time; see figure 1)
	Estimated smallpox coverage	HIV-1 prevalence (%)		Female/male PR (95% CI)	Estimated smallpox coverage	HIV-1 prevalence (%)		Female/male PR (95% CI)
Study site and survey year		Female	Male			Female	Male
Caió
1990	62%	0.3% (3/861)	0.4% (2/541)	0.94 (0.16 to 5.62)	72%	0.8% (7/907)	0.4% (2/461)	1.78 (0.37 to 8.53)
1997	27%	1.8% (17/958)	1.9% (14/738)	0.94 (0.46 to 1.89)	77%	4.4% (41/943)	2.8% (13/471)	1.58 (0.85 to 2.91)
2007	4%	3.2% (28/885)	1.5% (11/742)	2.13 (1.07 to 4.26)	71%	4.8% (41/850)	6.0% (25/418)	0.81 (0.50 to 1.31)
Bissau
1987	62%	0% (0/243)	0% (0/197)	NA	72%	0% (0/110)	0% (0/99)	NA
1996	27%	2.2% (19/881)	1.5% (10/680)	1.47 (0.69 to 3.13)	77%	3.3% (13/394)	3.5% (12/346)	0.95 (0.44 to 2.06)
2006	4%	5.3% (56/1056)	2.3% (16/705)	2.34 (1.35 to 4.04)	71%	5.4% (25/466)	6.5% (21/321)	0.82 (0.47 to 1.44)
2016	0%	4.2% (41/983)	0.8% (5/648)	5.41 (2.15 to 13.61)	66%	5.2% (13/252)*	5.0% (11/219)*	1.03 (0.47 to 2.25)*
Combined
1987–1990	62%	0.3% (3/1104)	0.3% (2/738)	1.00 (0.17 to 5.99)	72%	0.7% (7/1017)	0.4% (2/560)	1.93 (0.40 to 9.25)
1996–1997	27%	2.0% (36/1839)	1.7% (24/1418)	1.16 (0.69 to 1.93)	77%	4.0% (54/1337)	3.1% (25/817)	1.32 (0.83 to 2.10)
2006–2007	4%	4.3% (84/1941)	1.9% (27/1447)	2.32 (1.51 to 3.56)	71%	5.0% (66/1316)	6.2% (46/739)	0.81 (0.56 to 1.16)

Data are extracted from.4 5 9

*The estimation for the age group ≥35 in 2016 was changed to ≥45 to ensure a steady smallpox vaccination coverage.

The older age group with steady smallpox vaccination coverage had no increase in the female/male HIV-1 prevalence ratio. Combined, the female prevalence increased from 0.7% to 5.0% from 1987 to 1990 to 2006–2007, whereas the male prevalence increased from 0.4% to 6.2% in the same period. Thus, the female/male HIV-1 prevalence ratios were 1.93 (95% CI 0.40 to 9.25) in 1987–1990, 1.32 (95% CI 0.83 to 2.10) in 1996–1997, 0.81 (95% CI 0.56 to 1.16) in 2006–2007. The test of interaction for a homogeneous association between sex and a linear change in HIV-1 prevalence across surveys gave a p value of 0.07 and the direction trended towards the opposite direction than for the younger age group. The female/male HIV-1 prevalence ratio was 1.03 (95% CI 0.47 to 2.25) in 2016 in Bissau.

The combined female/male HIV-1 prevalence ratios are illustrated in figure 2. Relative to the female/male prevalence ratio among the older age group, the female/male prevalence ratio in the 15–34 years age group increased from 0.52 (95% CI 0.05 to 5.61) in 1987–1990 to 0.88 (95% CI 0.44 to 1.75) in 1996–1997 to 2.88 (95% CI 1.64 to 5.05) in 2006–2007 to 5.26 (95% CI 1.57 to 17.65) in 2016 (2016 estimates were only based on Bissau data) (ratios of ratios based on table 1).

Figure 2

Female/male HIV-1 prevalence ratios, Bissau and Caió. Visualisations of estimates from table 1. Circles and triangles represent point estimates and lines represent the 95% CIs. The estimation for the age group ≥35 in 2016 was only from Bissau and was changed to ≥45 to ensure a steady smallpox vaccination coverage.

Discussion

As we had hypothesised, the female/male HIV-1 prevalence ratio increased for the age group 15–34 years, as the proportion with smallpox vaccination scars decreased, whereas the female/male HIV-1 prevalence ratio remained unchanged for the age group ≥35 years which had a steady smallpox vaccination coverage over the HIV-1 survey years.

Strengths and limitations

This study was based on information from large HIV surveys carried out over 20–30 years in two different settings, urban and rural Guinea-Bissau. As no central smallpox vaccination register exists in Guinea-Bissau, we used smallpox vaccination scars as a proxy for the smallpox vaccination coverage. We have previously shown that smallpox scars have a sensitivity of >90% in correctly identifying smallpox vaccinated individuals (no specificity measure available).7 Some BCG vaccination scars and accidental wounds may have been misclassified as smallpox vaccination scars, but misclassification is unlikely to be sex-differential. Potential variation in false-positive and false-negative rates of scar across surveys would likewise not be expected to be sex-differential. Participation in the HIV surveys varied only slightly across the survey years in Bissau, being 86% in 1987, 85% in 1996, 79% in 2006 and 83% in 2016; furthermore, the HIV prevalence in participants, who were easy to reach, was similar to the prevalence in those who were difficult to reach.4 9 Hence, differential participation in different study years is unlikely to have caused selection bias.

The ecological design enabled us to investigate a potentially important hypothesis, but the results needs to be interpreted with caution since this design can be vulnerable to misinterpretations. By using the ratio of HIV-1 prevalence between sexes and within age groups, we could to some extent disregard calendar time trends such as the spread, prophylaxis and treatment of HIV-1 affecting both sexes and all age groups.

Female/male HIV-1 prevalence trends in sub-Saharan Africa

Consistent with our finding, cross sectional surveys from Malawi,11 Zambia12 and South Africa13 show that the birth cohorts who are too young to have been smallpox vaccinated have an increased female/male HIV prevalence compared with older birth cohorts, who are likely to have been smallpox vaccinated before the worldwide phase-out in 1980.

UNAIDS data for the female/male HIV-1 prevalence from 1985 to 2003 in Sub-Saharan Africa shows that in the age group of 15–49 years, the number of HIV-1 affected women began to increase over the number of men during the early 1990s.14 The female/male HIV-1 ratio increased at the same time as the smallpox vaccination coverage decreased after 1980. A multicountry study using repeated national representative demographic and health surveys on HIV prevalence in Sub-Saharan Africa during the 2000s did not find an increasing female/male HIV prevalence ratio.15 In contrast to the HIV surveys from Guinea-Bissau, where there was a clear increase in the prevalence of HIV-1, the reported HIV prevalence generally decreased between repeated surveys in other regions of Africa.15 The female/male HIV prevalence ratio may be influenced by multiple factors, and the introduction of HIV treatment, which only took place in the late 2000s in Guinea-Bissau,16 may have blurred the female/male HIV prevalence ratio trends in the 2000s surveys from Sub-Saharan Africa.

Potential causes of sex differences in HIV-1

The sex differences in the susceptibility to HIV-1 could theoretically be due to physiological, hormonal or local microbial differences, and higher prevalence of sexually transmitted diseases causing a higher male-to-female than female-to-male HIV transmission rate.14 These explanations would however not explain why young women in Sub-Saharan Africa did not have a higher HIV-1 prevalence than men when the HIV epidemic started. Our results showed that women and men in the age group of <35 years had similar HIV-1 prevalence in 1987–1990 of 0.3% but while men’s HIV-1 prevalence did not increase much in the younger age group, potentially due to more focus on availability and use of condoms over time, women’s HIV-1 prevalence continued to increase (table 1). It may be that females’ increased susceptibility were neutralised by the smallpox vaccination and became expressed when smallpox vaccination was stopped.

Alternative explanations for the sex-age-time pattern may be sought in social and cultural changes over time, including gender-power imbalances.14 Analyses of sexual mixing patterns from South Africa13 suggest that since young women often have sexual relations with older men, then as the prevalence of HIV-1 increases among older men the prevalence among young women will follow. We have no specific data to assess possible changes over time in the frequency of sexual relations across age groups. All ethnic groups in Guinea-Bissau has a taboo on intercourse while the mother is breast feeding for 1.5−3 years, which may have created a permissive attitude towards extramarital sexual relationships (possibly causing sexual relations between older men and younger women). We have documented such taboo on intercourse while breast feeding and a permissive attitude back to the 1980s in Guinea-Bissau,17 so it is clearly not a new phenomenon. While it is possible that behavioural patterns became increasingly permissive of extramarital sexual relationships in a setting with rapidly increasing urbanisation, it seems unlikely that the same change would have happened in a rural setting. We find it unlikely that the similar pattern of increasing female/male HIV-1 prevalence ratio in both an urban and a rural setting can be explained merely by changes in sexual behaviour patterns. During the study period, there may have been increased awareness and availability of condoms, but this would likely have affected the risk of acquiring HIV equally in both sexes or if anything diminished the risk in females relative to males.

The female/male HIV-1 prevalence ratio for the 15–34 years age group seems to continue to increase in Bissau with a ratio of 5.41 (95% CI 2.15 to 13.61) in 2016 compared with 2.34 (95% CI 1.35 to 4.04) in 2006, despite the prevalence of smallpox vaccination coverage for this age group was estimated to change from 4% to 0%, suggests that other factors continue to affect the susceptibility of HIV-1 differently for men and women.

Biological mechanisms

The CCR5 is fundamental for establishing HIV-1 infections.18 The CCR5-delta-32 deletion confers resistance to HIV-1 by preventing the expression of the CCR5 receptor; this allele provides almost complete resistance to HIV-1 in homozygous individuals.18 A recent immunological study found that cells from smallpox vaccinated individuals had up to fivefold reduction in CCR-5 tropic HIV-1 replication in vitro,2 which supports a role of the smallpox vaccine in HIV-1 prevention through heterologous immunity. A recent study did not show an association between smallpox vaccination scar and CCR5 expression on the surface of peripheral T-lymphocytes among HIV seronegative women old enough to have had a chance of being smallpox vaccinated19; this may be due to delay between smallpox vaccination and immunological testing of more than four decades or that the smallpox-unvaccinated control group had received another immunomodulator, the BCG vaccine.20

In animal models, administrating smallpox vaccination via skin scarification has been demonstrated to increase the immune response and survival compared with other modes of administration.21 Murine studies have shown that intradermal smallpox vaccination induced long-lived non-recirculating CD8+skin resident T-memory cells that resided within the entire skin and protected against reinfection.22 This indicates that vaccination can spread throughout the entire epithelial surface to create a ‘shield’ against infection.

Smallpox vaccine may also affect the innate immune system more broadly; in a very recent study, human monocytes trained with smallpox vaccine showed significantly increased interleukin-6 and TNF-α (tumor necrosis factor) production to stimulation with non-related stimuli, compared with non-trained monocytes.23

Overall, there is some immunological evidence to support that smallpox vaccination can provide cross-protection against HIV-1 infection. None of the above studies reported effects by sex, but it is plausible that an epithelial protection might be particularly protective against vaginally acquired HIV-1 infection.

Conclusion

Our hypothesis that the phase-out of smallpox vaccination may have increased the female/male HIV-1 prevalence ratio was compatible with our results. This hypothesis needs further assessments to determine if the relationship is causal, and we hope other research groups will test the hypothesis and other potential explanations for the change in female-to-male HIV prevalence ratios over time in individual-based data. If more support for the hypothesis that smallpox vaccine protected females against HIV can be obtained, from epidemiological and immunological studies, it could provide important information for HIV-1 vaccine research even though it may not be possible to reintroduce the smallpox vaccine.