Twin births, which are associated with an elevated risk of preterm birth, low birth weight, cesarean section, and numerous other health complications, represent a significant public health challenge (Bjerregaard-Andersen et al., Reference Bjerregaard-Andersen, Biering-Sørensen, Gomes, Bidonga, Jensen, Rodrigues, Christensen, Aaby, Beck-Nielsen, Benn and Sodemann2014; Blondel, Reference Blondel2009; Boubkraoui et al., Reference Boubkraoui, Aguenaou, Mrabet and Barkat2016; Pison, Reference Pison2000; Schenker et al., Reference Schenker, Yarkoni and Granat1981). This situation is of particular concern in sub-Saharan Africa (SSA), as this region has the highest rate of twin births worldwide (Smits & Monden, Reference Smits and Monden2011). Furthermore, the mortality rate for children in SSA is higher than in other regions (World Health Organization [WHO], 2018). This situation necessitates a more comprehensive understanding of the twinning rate in SSA countries, including its trends and the factors influencing this variation.
There is considerable variation in the twin birth rate, or twinning rate, between continents. The current twinning rate in SSA is approximately 17 per thousand (per 1000), as reported by Monden et al. (Reference Monden, Pison and Smits2021) and Ouedraogo (Reference Ouedraogo2020). In the 1980s and 1990s, the twinning rate in SSA was 4 to 5 times higher than in Asia and almost twice as high as in Europe (Pison, Reference Pison, Van de Walle, Pison and Sala-Diakanda1992). These disparities are currently being narrowed, although they remain significant (Monden et al., Reference Monden, Pison and Smits2021).
Several factors contribute to the spatial and temporal variations in twin birth rates. For instance, the twinning rate in developed countries increased significantly between 1970 and 2010, from less than 8 to almost 16 per 1000 (Pison et al., Reference Pison, Monden and Smits2015). This significant increase can be attributed to two factors: an increase in fertility treatments and the age of motherhood (Pison & Couvert, Reference Pison and Couvert2004; Pison et al., Reference Pison, Monden and Smits2015; Terzera, Reference Terzera2002). In developing countries, particularly in SSA, where twin birth rates are exceptionally high, it is presumed that advanced fertility treatments, such as in vitro fertilisation (IVF), are not widely utilized. However, other factors, such as a high fertility rate, a high number of births at later ages, and genetic and ethnic characteristics (Bomsel-Helmreich & Al Mufti, Reference Bomsel-Helmreich, Al Mufti, Blickstein, Keith, Keith and Teplica2005; Mbarek et al., Reference Mbarek, Gordon, Duffy, Hubers, Mortlock, Beck, Hottenga, Pool, Dolan, Actkins, Gerring, Van Dongen, Ehli, Iacono, Mcgue, Chasman, Gallagher, Schilit, Morton and Martin2024; Nylander, Reference Nylander1971), may contribute to maintaining these high twin birth rates (Ouedraogo, Reference Ouedraogo2020).
Statistics on twinning in SSA are scarce, and the extent of variation between regions and countries remains poorly known. There have been studies on the geographical and spatial variations in twinning rates in SSA (Monden et al., Reference Monden, Pison and Smits2021; Smits & Monden Reference Smits and Monden2011). However, few have conducted comprehensive analyses in multiple countries to examine these geographical and temporal variations. Additionally, research on the main factors influencing SSA’s twinning rates is scarce. Furthermore, although some researchers (Bomsel-Helmreich & Al Mufti, Reference Bomsel-Helmreich, Al Mufti, Blickstein, Keith, Keith and Teplica2005; Nylander, Reference Nylander1971) have analyzed the relationship between ethnicity and twinning rates in SSA, their studies focused only on specific ethnic groups, such as the Yoruba in Nigeria (Nylander, Reference Nylander1971). It remains to be demonstrated whether these situations are specific to certain groups or whether they can be observed throughout SSA. Finally, examining the future dynamics of twinning rates is essential in the current context of rapid population growth in sub-Saharan Africa.
The first objective of this study is to estimate the prevalence of twin births in 42 countries on the African continent between 1986 and 2016. This will be achieved by analysing the spatial, temporal and ethnic variations in twin birth rates. The second objective is identifying social and demographic factors associated with a high probability of twin births in SSA. Finally, the third objective is to provide a baseline projection of twin birth rates at the SSA continental level up to 2050.
Literature Review
Biology of Twinning: Two Types of Twin Births
There are two types of twins: monozygotic (MZ) or identical twins and dizygotic (DZ) or fraternal twins (Hall, Reference Hall2003). MZ twins result from fertilization of a single egg by a single sperm, with the egg splitting in two in the first few days after fertilization. MZ twins are necessarily of the same sex and have a similar genotype. The MZ twinning rate is constant at around 3.5 to 4 per 1000, independent of maternal age, birth order number, geographic or ethnic origin, and medically assisted reproduction (MAR; Long & Ferriman, Reference Long and Ferriman2016; Pison et al., Reference Pison, Monden and Smits2015).
DZ twins are the product of fertilizing two different eggs by two different sperm. In contrast to MZ twins, DZ twins are not inherently of the same sex. The frequency of DZ births is subject to the influence of the aforementioned factors (Bulmer, Reference Bulmer1970; Pison, Reference Pison, Van de Walle, Pison and Sala-Diakanda1992).
This article does not distinguish between the types of twins, as this information is unavailable.
Twinning Rate Variation Factors
Maternal age
A substantial body of research has demonstrated that the likelihood of a twin birth increases with maternal age (Blondel, Reference Blondel2009; Bulmer, Reference Bulmer1970; Gabler & Voland, Reference Gabler and Voland1994; Pison et al., Reference Pison, Monden and Smits2015; Satija et al., Reference Satija, Sharma, Soni, Sachar and Singh2008; Sear et al., Reference Sear, Shanley, Mcgregor and Mace2001). In a study conducted by Pison et al. (Reference Pison, Monden and Smits2015), the maternal age range of 35—39 years had the highest twinning rates in Japan, England & Wales, France and the US. This was observed before the spread of MAR.
Bomsel-Helmreich and Al Mufti (Reference Bomsel-Helmreich, Al Mufti, Blickstein, Keith, Keith and Teplica2005) posit that follicle-stimulating hormone (FSH) is necessary for follicle development and triggers ovulation. As the average FSH level increases, so does the likelihood of double ovulation and double fertilization in the same cycle (Couvert, Reference Couvert2011; Mbarek et al., Reference Mbarek, Gordon, Duffy, Hubers, Mortlock, Beck, Hottenga, Pool, Dolan, Actkins, Gerring, Van Dongen, Ehli, Iacono, Mcgue, Chasman, Gallagher, Schilit, Morton and Martin2024; Mbarek et al., Reference Mbarek, Steinberg, Nyholt, Gordon, Miller, McRae, Hottenga, Day, Willemsen, de Geus, Davies, Martin, Penninx, Jansen, McAloney, Vink, Kaprio, Plomin, Spector and Boomsma2016; Nylander, Reference Nylander1981). It has been demonstrated that FSH concentration increases with age, which explains the rise in the twin birth rate with maternal age.
Birth order number (parity)
The birth order number is another maternal characteristic associated with the probability of twin births. This link has been the subject of several studies, including that by Duncan (Reference Duncan1865). His work demonstrated that the number of twin pregnancies in women increases with maternal age and birth order number. These results were confirmed by Bulmer (Reference Bulmer1970), who found that despite the apparent correlation between maternal age and birth order number, each factor has an independent effect on the probability of twin births. Daguet (Reference Daguet2002) and Couvert (Reference Couvert2011) also observed that at the same maternal age, women with a high birth order number are more likely to give birth to twins than nulliparous women or women with a small number of births.
Medically assisted reproduction (MAR)
The phenomenon of MAR represents a significant and novel factor influencing the prevalence of twinning rates across the globe (Smits & Monden, Reference Smits and Monden2011). The likelihood of multiple births is significantly increased by MAR (Terzera Reference Terzera2002; Pison & Couvert Reference Pison and Couvert2004; Vitthala et al., Reference Vitthala, Gelbaya, Brison, Fitzgerald and Nardo2009). In industrialized countries, the advancement of human reproductive technology is currently the primary driver behind the significant increase in the twinning rate associated with delayed maternity (Pison et al., Reference Pison, Monden and Smits2015). In SSA, the development of MAR is still in its infancy (Bonnet, Reference Bonnet2016). Although not yet documented, it is likely that its impact on twinning rates is minimal.
Geographical and ethnic factors
There is considerable geographical variability in the frequency of twin births between subregions in Africa. Pison (Reference Pison, Van de Walle, Pison and Sala-Diakanda1992) demonstrated that the twinning rate was higher in countries bordering the Gulf of Guinea, with an increase from inland to the coast. Additionally, Smits and Monden (Reference Smits and Monden2011) have shown that this African area with a high incidence of twin births extends to some central and eastern African countries. They also demonstrated that Benin had the highest national twinning rate, estimated at 28 per 1000, while Madagascar had the lowest rate, at approximately 10.6 per 1000.
The high twinning rates observed in SSA have been attributed to a genetic predisposition of women from particular ethnic groups. The geographical distribution of these ethnic groups could thus explain the regional disparities in twinning rates. Bomsel-Helmreich and Al Mufti (Reference Bomsel-Helmreich, Al Mufti, Blickstein, Keith, Keith and Teplica2005) demonstrated that Yoruba women had a significantly higher concentration of FSH in their blood than in Aberdeen (Scotland), which may explain the higher rate of twin births among Yoruba women. In this article, we will consider the role of ethnicity. However, we will not analyze the effect of genetics due to the lack of information in the surveys studied.
The preceding literature review indicates that the twinning rate of the SSA is the highest. Furthermore, the literature contains studies examining the factors contributing to the variation in the twinning rate within the SSA context. In addition to corroborating certain factors (e.g., geographical variations in the twinning rate), this study extensively examines the temporal variations in the twinning rate across 42 countries on the continent. Moreover, while research has demonstrated ethnic variations in the twinning rate in SSA, this association is most often investigated at the localized scale of one or two countries, predominantly in Nigeria. Consequently, an additional novel aspect of our methodology is the examination of ethnic disparities in the twin birth rate in SSA, encompassing a broader geographical scope. Another area of interest in our research is the study of socio-demographic factors associated with the likelihood of having twin births. This represents a novel contribution, as it extends beyond existing descriptive analyses. It proposes an investigation at the sub-Saharan level, which will mobilize several factors to reach more robust conclusions. It is also observed that there is a paucity of research on the projected twinning rate in Africa. Consequently, the baseline forecast of the twinning rate up to 2050, proposed by the present approach, will contribute to the demographic analysis of the twinning rate.
Methods
Data Source
The data used in this study were obtained from 174 Demographic and Health Surveys (DHS) and Multiple Indicator Cluster Surveys (MICS) (see Appendix A) conducted in 42 SSA countriesFootnote 1 (Figure 1) between 1986 and 2016. These are retrospective cross-sectional surveys with national coverage. The DHS (https://www.dhsprogram.com/) and the MICS (https://mics.unicef.org/) are national surveys collecting information on a range of topics, including the fertility of women aged 15−49. They employ a multistage stratified sampling design and are nationally representative. DHS and MICS data are typically divided into several datasets, including the ‘Births’ file, which records the complete reproductive history of every woman aged 15−49. A specific variable on twin births is available in almost all datasets (see an extract of the questionnaire in Appendix B). Nevertheless, to guarantee the quality of this information, we created another variable entitled ‘twin’, using a matching method that considers the woman’s identification number (ID) and the child’s date of birth.
Figure 1. Geographical coverage of the study.
One hundred seventy-four surveys were analyzed, with data extracted and pooled from all. The data included deliveries in the 10 years preceding each survey (between t and t−10 years, where t is the survey year). This 10-year selection was made to compensate for the low annual number of twin births. A pooled sample of 2,479,385 deliveries was obtained, of which 44,035 were multiple births and 2,435,350 were single births. This pooled sample was used to estimate twinning rates using surveys, ethnic groups, countries, subregions, and an SSA-aggregated scale.
The analysis of factors associated with twin births was restricted to a survey period covering as many countries as possible (2000–2010). One survey per country was considered, with only those countries that collect information on ethnicity included, as this is a variable of interest in the analysis. The final subsample consisted of 25 surveys, encompassing 488,083 deliveries, 9160 multiple births, and 478,923 single births.
Additional data from the 2019 United Nations World Population Prospects (WPP; United Nations, Department of Economic and Social Affairs, Population Division, 2019) (https://population.un.org/wpp/) — specifically, the projections of deliveries by maternal age group — were used to draw up the projection of the twinning (only for the SSA-aggregated scale) up to 2050. The three main UN projection scenarios were considered: low fertility, medium fertility, and high fertility.
Statistical Analysis
The twinning rate was calculated by applying the following formula:
$${\rm{Twinning\;rate\; = }}\left( {{{{\bf{Number}}\;{\bf{of}}\;{\bf{multiple}}\;{\bf{births}}}}\over{{{\bf{All}}\;{\bf{births}}\;}}} \right)*1000$$
Given that the twinning rate depends on maternal age (Smits & Monden, Reference Smits and Monden2011), it was standardized using the standard age distribution of births from women aged 15–49 in SSA, as provided by the United Nations, Department of Economic and Social Affairs, Population Division (2019).
A weighting was applied to estimate the twinning rate for all 42 countries and its distribution by subregion, considering the share (weight) of each country’s births in the total births of the 42 countries.
As triplets and more are relatively uncommon (0.21 per 1000 in our data), they were included in the analysis alongside twin births.
Bivariate tabulations explored the unadjusted relationship between the response and each independent variable. Subsequently, a multivariate logistic regression was employed to construct an adjusted model. Logistic regression is a semiparametric method that can be employed to identify the factors that influence the probability of having twins, assuming that all other variables are held constant. The response was a categorical variable indicating whether the delivery was a twin birth or not. The independent variables included Maternal age, Birth order number, Maternal ethnic group, Household wealth quintile, African subregion, and Year of childbirth. To ascertain which of the two factors, maternal age or birth order number, was the most strongly associated with twin births, we also evaluated their respective contributions to the model’s parsimony.
The twinning rate projection method was straightforward and elementary. The twinning rate for each maternal age group (estimated from DHS and MICS) was multiplied by the projected number of deliveries (obtained from WPP) in that age group. Then, we calculated the projected number of twin births by maternal age group and subsequently determined the total projected number of twin births by country by summing these values. The number of twin births in the SSA region was estimated by aggregating the data from the 42 countries included in the study. The projected twinning rate is calculated by dividing the number of twin births by the total number of deliveries.
For each country, we assumed:
• p: the projection period
• G 15–19, G 20–24, …, G 40–49: the twinning rates by age group calculated from DHS and MICS
• N 15–19, N 20–24, …, N 40–49: projected births by age group from UN World Population Prospects
• TRp: the projected twinning rate for the period p.
$$\small{T{R_p} = {{{\left( {{G_{15 - 19}}*{N_{15 - 19}}} \right) \!+\! \left( {{G_{20 - 24}}*{N_{20 - 24}}} \right) \!+\! \ldots \! +\! \left( {{G_{40 - 49}}*{N_{40 - 49}}} \right)}}\over{{\left( {{N_{15 - 19}} + {N_{20 - 24}} + \; \ldots + {N_{40 - 49}}} \right)}}}}$$
Results
Twinning Rate
The average standardized twinning rate for all 42 countries between 1986 and 2016 was 17.4 per 1000 (95% CI [17.2, 17.6]), with a median of 18.2 per 1000. The twinning rate is typically higher than the global (world) average of 11.3 per 1000 in 2010 (Pison et al., 2017), except for Madagascar (10.6 per 1000), Burundi (10.6 per 1000) and Somalia (5.5 per 1000). Benin is the African country with the highest twinning rate, estimated at more than 27 per 1000 (25–30). Table 1 presents the mean twinning rate by subregion in SSA. The twinning rate in West Africa is the highest, 20 per 1000 (19.5–20.1), while that in Southern Africa is the lowest, 13 per 1000 (12.1–13.7). Further details regarding the twinning rate by survey and country can be found in Appendix A. The map in Figure 2 also illustrates the mean twinning rates for each country and subregion. The region exhibiting the highest twinning rate is situated near the Gulf of Guinea, extending in a band traversing Africa from the Democratic Republic of the Congo in the west to Tanzania and Mozambique in the east.
Table 1. Variation of the twinning rate by subregion in sub-Saharan Africa
Note: aSahel: Burkina Faso, Mali, Mauritania, Niger, Senegal, Chad; Gulf of Guinea: Angola, Benin, Cameroon, Congo, Cote d’Ivoire, Gabon, Gambia, Ghana, Guinea, Guinea-Bissau, Liberia, Nigeria, DR Congo, Sao Tome, Sierra Leone, Togo; East Africa: Burundi, Comoros, Kenya, Malawi, Mozambique, Uganda, Rwanda, South Sudan, Tanzania, Zambia; Southern Africa and Madagascar: Lesotho, Madagascar, Namibia, Swaziland, Zimbabwe. Source: Demographic and Health Surveys (DHS) and Multiple Indicator Cluster Surveys (MICS); authors’ calculation.
Figure 2. Map of the twinning ratea (average) in sub-Saharan Africa.
Note: aMaternal age-standardized twinning rates.
Figure 3 illustrates the fluctuations in the twinning rate at the continental level and for each country over time (from 1986 to 2016). The observed trends differ according to the countries under consideration but appear relatively stable in almost all of these countries, starting in the 2000s.
Figure 3. Trends of the standardized twinning rate in sub-Saharan African countries.
Source: Demographic and Health Surveys (DHS) and Multiple Indicator Cluster Surveys (MICS); authors’ calculation.
Ethnic Variation in the Twinning Rate in SSA
Figure 4 illustrates the variability in the twinning rate across 102 ethnic groups in some countries in SSA for which the DHS and MICS collected data on ethnicity. These findings corroborate the existence of notable disparities in the twinning rate across different ethnic groups. However, the confidence intervals are large due to the limited number of twin births. The ethnic groups with the highest twinning rates (per 1000) are Fon (29) from Benin; Bamileke (28), Beti (27.6) and Grassfields (27.5) from Cameroon; Mbochi (28) and Teke (25) from Congo; ‘Burkinabe’ (28) of Cote d’Ivoire; Kota-Kele (30) from Gabon; Ewe (25) from Ghana; Balanta (25) from Guinea-Bissau; Tonga (27), Tumbuka (25) and Yao (25) from Malawi; Ndau (30) and Tswa (28.6) from Mozambique; Fulani (26) from Nigeria; Kuranko (26,4) and Kono (25,6) from Sierra Leone; Kabye (26) from Togo; Baganda (25) from Uganda; and Lunda (25,6) from Zimbabwe.
Figure 4. Variation in the standardized twinning rate by selected ethnic groups in some sub-Saharan African countries.
Source: Demographic and Health Surveys (DHS) and Multiple Indicator Cluster Surveys (MICS); authors’ calculation.
Twin Births Associated Factors in SSA
In our sample, 47% of births were from women under 25 years of age (Table 2), 43% were from women who had already given birth three or more times, nearly 50% were from women living in countries around the Gulf of Guinea, almost 50% were from women in the poorest or poorest squared households, and 32% were from women belonging to the aggregated ethnic Bantu group.
Table 2. Factors associated with twin births: A univariate, bivariate and multivariate analysis
Note: OR, odds ratio; CI, confidence Interval; ***p value < .001, **p value < .01, *p value < .05. a The household wealth quintile employed in this study is based on the standardised wealth index, which permits comparisons across time and countries. bThe construction of these aggregated ethnic groups is an adaptation of a primary linguistic division (Malherbe, Reference Malherbe2000) of the peoples of sub-Saharan Africa: http://www.cosmovisions.com/Afrique-Carte-Langues.htm Source: Demographic and Health Surveys (DHS) and Multiple Indicator Cluster Surveys (MICS); authors’ calculation.
Bivariate analysis indicates that the primary factors associated with twinning are maternal age and birth order number. The twinning rate increases with maternal age until it reaches a maximum around 39–43 years of age, with a rate of approximately 32 per 1000 births. An independence test of Rao-Scott’s chi-square demonstrates the association between maternal age (recoded into an age group) and twinning, with a p value less than .001. Similarly, a positive correlation was found between twinning rate and birth order number (p < .001). The rate is greater than 34 per 1000 for births at order 10 and above, compared to 8 per 1000 for first births.
The bivariate results also demonstrated associations between the twinning rate and other explanatory variables, including the country’s geographical sub-region (p value < .001), the household wealth quintile (p value = .02) and the mother’s ethnic group (p value < .001).
The logistic regression results (Table 2) demonstrate that assuming all other variables remain constant, the probability of giving birth to twins is significantly higher among older women, regardless of the birth order number. Compared to women aged 20–25, the likelihood of twin births is 1.16 times higher (p value < .01) among women in the 35 and over age group. Concerning the birth order number, the adjusted odds ratios indicate that the probability of twin births is 2.92 times higher (p value < .01) for births of the sixth order or higher than first births.
The influence of birth order number on the probability of twin births appears to be greater than that of maternal age, as evidenced by its higher odds ratio (OR) and its more significant contribution to lowering the Akaike information criterion (AIC) in the adjusted model.
For the remaining covariates, the probability of twin births is significantly higher among women belonging to the aggregated Bantu ethnic group compared to women from the following ethnic groups: Arabs and Related (OR = 0.75; p < .001), Fulani and Related (OR = 0.82; p < .001), Saharans (OR = 0.66; p < .01), Mandes (OR = 0.84; p < .01), and Ubangian-Adamaouans (OR = 0.8; p < .05). The probability of twin births increases with the household’s wealth quintile. The following ORs are observed: 2nd quintile (OR = 1.1, p < .01); 3rd quintile (OR = 1.15, p < .001); 4th quintile (OR = 1.17, p < .001); 5th quintile (OR = 1.2, p < .001). Additionally, a high probability of twin births is observed in Gulf of Guinea countries (OR = 1.3, p < .001). Finally, there is a low positive correlation between twin births and the year of birth (OR = 1.013, p < .01), indicating a slight increase in the probability of a twin birth over time.
Projected Twinning Rate in SSA up to 2050
For aggregated SSA, our primary projections indicate that between 2015 and 2050, the twinning rate is not expected to change appreciably (Figure 5). The low-fertility scenario projects that between 2015 and 2050, the twinning rate in SSA will increase from 17.4 to 18.4 per 1000, representing a growth of less than 0.16% per year. Considering the medium-fertility scenario, the twinning rate in SSA is projected to increase from 17.4 per 1000 in 2015 to 18.1 per 1000 by 2050, representing an annual growth of only 0.11%. Concerning the high-fertility scenario, the twinning rate is anticipated to increase by approximately 0.08% annually between 2015 and 2050 (from 17.4 to 17.9 per 1000).
Figure 5. Projected Twinning Rate in sub-Saharan Africa.
Source: UN World Population Prospects, Demographic and Health Surveys (DHS), and Multiple Indicator Cluster Surveys (MICS); authors’ calculation.
Discussion
Confirming the High Twinning Rate in Africa
This study corroborates the high twinning rates observed in SSA. The twinning rate of 17.4 per 1000 we found is consistent with the findings of other researchers, including Monden et al. (Reference Monden, Pison and Smits2021), Gebremedhin (Reference Gebremedhin2015), Smits and Monden (Reference Smits and Monden2011), and Pison (Reference Pison, Van de Walle, Pison and Sala-Diakanda1992). The spatial variations observed are consistent with the results of previous studies, which have shown a higher rate of twin births around the Gulf of Guinea and in some central and eastern African countries, including Southern Sudan, Malawi, Mozambique, Comoros, Zambia, and Tanzania. Our study has also examined the temporal trends in the twinning rate. While Monden et al. (Reference Monden, Pison and Smits2021) and Smits and Monden (Reference Smits and Monden2011) have addressed this aspect, they have not done so systematically. This study has mobilized more national surveys for each country, enabling us to construct significant and robust trends. This approach allowed us to highlight the slow variation in the twinning rate in SSA. This slight variation could be explained by a kind of equilibrium resulting from the decline in the fertility rate (including births at older ages), which draws the twinning rate down, and a slow increase in the average maternal age, which slowly pulls the twinning rate upwards. If the decline in the fertility rate on the continent outweighs the increase in the average age of motherhood, an absolute reduction in the twinning rate is expected in the coming years. MAR, which could increase the twinning rate, is limited to a small proportion of the sub-Saharan population and is unlikely to have affected the twinning rate (Bonnet, Reference Bonnet2016).
Significant Variations in Twinning Rates by Ethnic Group in SSA
A significant disparity in the twinning rate has been observed between ethnic groups in SSA. While for some ethnic groups, the twinning rate aligns with the national rate, this is not the case for the following groups: the Bamileke, Beti and Grassfields (Cameroon); Mbochi (Congo); ‘Burkinabe’ (Cote d’Ivoire); Kota-Kele (Gabon); Ewe (Ghana); Balanta (Guinea-Bissau); Tonga (Malawi); Ndau and Tswa (Mozambique); Fulani (Nigeria); Kuranko and Kono (Sierra Leone); and Baganda (Zimbabwe). In comparison to the national rates, the twinning rates for these ethnic groups are significantly higher. Excepting West African countries, the majority of ethnic groups with high twinning rates belong to the Bantu ethnic group, which is one of the largest in Africa. Conversely, among the Biu-Mandara (from Cameroon), Somalis (from Kenya), and Makua (from Mozambique), the twinning rates observed were significantly lower than the national rates.
For example, our findings are consistent with Pollard (Reference Pollard1996), who reported an extraordinarily high twinning rate among the Tumbuka from Malawi. However, our study did not find an exceptionally high twinning rate among the Yoruba of Nigeria. This is in contrast to Nylander (Reference Nylander1971). Nevertheless, this study has allowed us to examine the relationship between ethnicity and twinning rates in SSA more broadly than previously. Indeed, we conducted a comprehensive analysis of the variation in the twinning rate across 102 ethnic groups, whereas existing studies have focused on specific ethnic groups. However, it should be noted that SSA comprises over 2000 ethnic groups (Elamé, Reference Elamé2016), and therefore, our analysis was limited to the most represented ethnic groups in our data set. Additionally, it is important to acknowledge that almost half of the SSA countries no longer collect ethnicity information in their surveys. In light of the above, our approach has not exhausted the distribution of the twinning rate according to ethnic groups in SSA.
Researchers have put forth two principal explanations for the potential correlation between the probability of twin births and ethnicity. The first is genetic. There is evidence to suggest that women of certain ethnic groups may have a genetic predisposition to experiencing twin pregnancies. This was demonstrated by Bomsel-Helmreich and Al Mufti (Reference Bomsel-Helmreich, Al Mufti, Blickstein, Keith, Keith and Teplica2005) in their study of the Yoruba in Nigeria. The present study did not specifically address this issue due to the lack of consistent data. However, our findings that most ethnic groups with very high twinning rates belong to the large family of Bantu ethnicities could provide further evidence to support the genetic link hypothesis. The second reason given was the lifestyle, particularly dietary. In this context, Creinin and Keith (Reference Creinin and Keith1989) identified a correlation between yam consumption and an elevated risk of twin pregnancies. Steinman (Reference Steinman2017) highlighted the potential role of dairy products in this process. The authors posit that certain foods contain hormones that induce multiple ovulations. Whereas these studies offer a novel perspective, the high twinning rates observed in numerous ethnic groups across the continent do not confirm a specific dietary pattern as a causal factor in twin births. However, further investigation is needed to test this hypothesis, which was not the focus of our study.
Birth Order Number is a Leading Factor More Associated With Twin Births Than Maternal Age
Our findings indicate that the birth order number is the primary factor associated with twin births, in contrast to the maternal age observed by Couvert (Reference Couvert2011) for France. It is postulated that the discrepancy between the results of this study and those of Couvert (Reference Couvert2011) can be attributed to differences in the fertility rates of the countries under consideration. Given that fertility rates in SSA are three to four times higher than in France, it can be concluded that the birth order number is more strongly associated with the probability of twin births in countries with high fertility rates (Ouedraogo & Jean Simon, Reference Ouedraogo and Jean Simon2021). Nonetheless, the impact of the birth order number on the likelihood of twin births may be attributable to the fact that the genes that predispose mothers to higher fecundity are likely to be the same genes that predispose them to the occurrence of DZ twin births (Couvert, Reference Couvert2011; Mbarek et al., Reference Mbarek, Gordon, Duffy, Hubers, Mortlock, Beck, Hottenga, Pool, Dolan, Actkins, Gerring, Van Dongen, Ehli, Iacono, Mcgue, Chasman, Gallagher, Schilit, Morton and Martin2024).
The Twinning in SSA Will Probably not Increase Much by 2050
One of our study’s original features is the twinning rate projection. However, this preliminary analysis demonstrated that in the optimal scenario (low fertility scenario), the twinning rate would increase from 17.4 per 1000 in 2015 to 18.4 per 1000 in 2050. The observation that the low fertility scenario predicts higher growth in the twinning rate than the other projection scenarios is intriguing. A reduction in fertility at both younger and older ages is anticipated under low fertility scenarios, with the decline being more pronounced at younger ages. This will result in a reconfiguration of the maternal age structure, with the average age at childbirth increasing slightly, thereby exerting a slight growth effect on the twinning rate. It should be noted, however, that this projection does not consider the potential expansion of MAR on the continent.
The World Population Prospects (WPP) considers changes in the maternal age structure but does not incorporate assumptions about the possible expansion of MAR in SSA, which is an essential factor in the future dynamics of the twinning rate. The effective utilization of MAR by 2050 would significantly increase the projected twinning rate, exceeding the predicted levels. However, by 2050, it remains to be seen whether MAR will expand to such an extent in SSA that it will significantly impact the twinning rate. Given the statistics on MAR on the continent, this possibility seems unlikely to become a reality. Indeed, Dyer, Archer et al. (Reference Dyer, Archary, de Mouzon, Fiadjoe and Ashiru2019), Dyer, Chambers et al. (Reference Dyer, Chambers, de Mouzon, Nygren, Zegers-Hochschild, Mansour, Ishihara, Banker and Adamson2016) and the ICMART (International Committee for Monitoring Assisted Reproductive Technologies) have demonstrated that Africa recorded fewer than 2000 pregnancies resulting from MAR in 2010. Concurrently, our projections indicate that an average of 260 million births per year is anticipated in SSA by 2050, including approximately 4.7 million twin births per year. Consequently, even if we assume that 500,000 births will result from MAR pregnancies by 2050, this will not alter the projected twinning rate by 2050.
The conclusions drawn in this study regarding the projected twinning rate on a continental scale by the year 2050 are consistent with those previously outlined by Lee and Barclay (Reference Lee and Barclay2025). By extending the projection to the year 2100, these authors have underscored the potential for substantial increases in the twinning rate in certain countries, including Benin, Malawi, Mozambique, Niger, Nigeria, and Togo. However, it is essential to note that projections of the twinning rate at the country level and over a prolonged period (to 2100) are potentially unreliable. Consequently, in this study, we have opted to estimate the twinning rate on a continental scale and have limited our analysis to the year 2050. Still, the findings of Lee and Barclay (Reference Lee and Barclay2025) and the results of this article are in accordance with the conclusion that if there is to be a substantial increase in the twinning rate in SSA in the forthcoming decades, it will be attributable to the combined impact of two factors: the extensive recourse to MAR and the maintenance of fertility at relatively high levels with an increase in late motherhood.
Conclusion
Given the high prevalence of twin births in the region, this study’s findings can potentially inform health policies in SSA. The increased fragility of twin children, due to their lower birth weight and higher frequency of prematurity, poses significant obstetric challenges and contributes to elevated risks of foetal and neonatal mortality. It is, therefore, crucial to implement effective monitoring strategies for pregnancies in large multiparous women, with the aim of early diagnosis and prevention of multiple pregnancies and their associated adverse outcomes. This health challenge was analyzed in another article that addressed the excess mortality among twins in SSA.
Limitations
The key limitation of this study is the potential for reliability issues in the data utilized, with the DHS not accounting for stillbirths, which could lead to an underestimation of twin births due to the heightened probability of stillbirth in twin pregnancies. However, extant studies have concluded that this bias’s impact is negligible or limited. Another potential source of reliability in DHS data related to twins is the existence of voluntary under-reporting of twin births by some families, either because they have ambivalent or negative perceptions of twins or because they do not want to evoke painful memories related to the (more frequent) death of their twin babies.
Added Value of this Study
The principal contribution of this work concerns its geographical scope, in that it reports on the level and evolution of the twinning rate in 42 countries, thus providing a broad ‘geographical view’ and allowing for various comparisons (e.g., ethnic groups, countries, subregions). A notable strength of the study lies in its extensive coverage of twinning rates across a diverse array of 102 sub-Saharan ethnic groups, a feat that is particularly significant given the paucity of studies that have hitherto focused on such a wide range of ethnic groups in the region. Furthermore, the study’s (summary) projections of the twinning rate in SSA by the year 2050 represent a new and valuable addition to the existing body of knowledge on the subject.
Acknowledgments
We acknowledge the Demographic and Health Surveys program (https://www.dhsprogram.com/) for free access to the surveys used in this study.
Declaration of interest statement
This article is based on the doctoral research of Adama Ouedraogo, conducted between October 2017 and September 2020 at the French Institute for Demographic Studies (INED) and the Research Centre of Paris 1 Pantheon-Sorbonne University’s Institute of Demography (CRIDUP).
Financial support
None.
Appendix A. Twinning rate in 42 countries of sub-Saharan Africa — Data: standard Demographic and Health Surveys (DHS), Malaria Indicators Survey (MIS), AIDS Indicator Survey (AIS) and Multiple Indicator Cluster Surveys (MICS)
Note: *Data with possible bias: short period (less than 10 years) and reproductive histories limited to 5 entries (5 deliveries) per woman. aCountries not included because of lack of data: Botswana, Cape Verde, Djibouti, Equatorial Guinea, Eritrea, Mauritius, and Seychelles. bBy dividing the sum of a country’s standardized rates by the number of its surveys. cBold type indicates the number of double births per 1000 deliveries. dBold type indicates that given the positive correlation between twinning and maternal age, rates were standardized using the age distribution of births of women aged 15–49 in SSA from 2000–2010 (source: United Nations). Source. DHS and MICS; authors’ calculations
Appendix B. Extract from the birth section of the Women’s Questionnaire 15–49 years old Source. Demographic and Health Survey Ghana 2014
Source. Demographic and Health Survey Ghana 2014.
Appendix C. Variation in the crude and standardized twinning rates by selected ethnic groups in some sub-Saharan African countries
Open access
