Introduction
Consanguinity is a common practice in most of the developing world, whereas, with few exceptions, it is generally not noteworthy in many of the developed nations (Bittles, Reference Bittles, Speicher, Antonarakis and Motulsky2010). Rapid demographic change is affecting developing nations, and new patterns of fertility, parenthood, and family formation are emerging (Bongaarts et al., Reference Bongaarts, Sathar, Mahmood, Sathar, Royan and Bongaarts2013a; Ranganathan et al., Reference Ranganathan, Swain and Sumpter2015; Goujon et al., Reference Goujon, Wazir and Gailey2020).
Numerous demographic features have been linked to changes in the pattern of marriages and the decline in consanguinity. Consanguineous unions (CU), for example, typically decline as a result of social modernization, which involves a number of changes from a traditional, rural, and extended family, agro-based society to a secular, urban and nuclear family, industrial society. These changes have a significant positive impact on employment, education, and health (Bittles and Black, Reference Bittles and Black2010). Furthermore, the rate of CU is projected to decline due to rising marriage age, postponed motherhood for women, and declining teenage fertility (Bongaarts et al., Reference Bongaarts, Mir, Mahmood, Sathar, Royan and Bongaarts2013b; Goujon et al., Reference Goujon, Wazir and Gailey2020).
In many developing nations particularly the Muslin counties, there is a clear shift in marital union patterns. Examples include the alignment of first cousin unions, the transfer from CUs to non-CU, and patrilineal types to matrilineal unions (Hamamy et al., Reference Hamamy, Jamhawi, Al-Darawsheh and Ajlouni2005; Al-Kandari, Reference Al-Kandari2006; Al-Arrayed and Hamamy, Reference Al-Arrayed and Hamamy2012; Schellekens et al., Reference Schellekens, Kenan and Hleihel2017). A demographic research conducted in Jordan by Hamamy et al. (Reference Hamamy, Jamhawi, Al-Darawsheh and Ajlouni2005) shows a drop in CU over time. In a recent study, Islam (Reference Islam2021) observed a decline in CU in Jordan and identified the factors that contributed to this decline as rising female education levels and marriage-age women, husbands with higher levels of education, declining family sizes, rising rates of urbanization and female employment, media exposure, and higher socio-economic status. Likewise, Al-Kandari (Reference Al-Kandari2006) witnessed that the level of education and the expansion of the ‘marriage circle’ were two important causes in Kuwait’s apparent reduction in CU after over two decades. Al-Arrayed and Hamamy (Reference Al-Arrayed and Hamamy2012) witnessed that in Bahrain, between 1990 and 2009, the percentage of cousin marriage sharply declined from 24% to 7%. A marked decrease in CU has also been witnessed in Muslims in Palestine and Israel (Assaf and Khawaja, Reference Assaf and Khawaja2009; Schellekens et al., Reference Schellekens, Kenan and Hleihel2017). The primary causes of the drop in CU were determined to be school enrolment and educational achievement (Schellekens et al., Reference Schellekens, Kenan and Hleihel2017).
Furthermore, in Morocco, the prevalence of CU declined by 4 percentage points and the researchers showed that improving economic conditions, declining fertility, rural-to-urban migration, and higher female educational levels are all contributing factors (Anwar et al., Reference Anwar, Khyatti and Hemminki2014). A recent study showed a significant overall decline in CU in India. However, the degree of change varied in different states, with major declines in CU being seen in southern areas and moderate increases being seen in certain northern regions (Kumari et al., Reference Kumari, Bittles and Saxena2020). Grjibovski et al. (Reference Grjibovski, Magnus and Stoltenberg2009) showed that there is a decrease in the proportion of consanguineously related parents of children born to women of Pakistani origin in Norway. In the Rahim Yar Khan population of Pakistan, Riaz et al. (Reference Riaz, Mannan and Malik2016) investigated consanguinity and its socio-biological factors. According to this study, CU decreased from 63% of participants of higher age to 55% of participants of younger age.
On the other hand, CU are still common in some nations in Asia and Africa. Consanguinity was found to be prevalent in several rural and socioeconomically challenged Muslim populations in Kerala, India, and there has been no decline in subsequent generations. Educational and economic backwardness and being strict in Islamic norms are contributing factors to this situation (Lekshmi and Sudhakaran, Reference Lekshmi and Sudhakaran2012). Studies from two Pakistani communities, Malakand and Bajaur Agency, found an increase in the rate of CU as a result of civil unrest at the Pakistan-Afghanistan border and turbulent security conditions (Sthanadar et al., Reference Sthanadar, Bittles and Zahid2014; Ahmad et al., Reference Ahmad, Rehman and Malik2016a, Ahmad et al., Reference Ahmad, Rehman and Malik2016b; Rehman et al., Reference Rehman, Ahmad, Zaman and Malik2016). However, as demonstrated by a recent study in India (Kumari et al., Reference Kumari, Bittles and Saxena2020), these data might simply represent localized trends and the overall picture is probably different.
Pakistan faces a huge burden of congenital and hereditary anomalies majority of which are autosomal recessive in nature and are rendered by cousin marriages and CU (Bibi et al., Reference Bibi, Naqvi, Syed, Zainab, Sohail and Malik2022; Shaheen et al., Reference Shaheen, Humayoon, Malik and Mumtaz2023). The high frequency of CU in Pakistan is due to a number of socio-demographic factors, including the usual practice of marrying a blood-related or kin. This study aimed to identify the factors that influence consanguinity in four areas of northwest Pakistan and to clarify any temporal changes in the level of consanguinity.
Methods
Study design and sample collection
The study protocols were approved by the Ethical Review Committees of the Quaid-i-Azam University and Hazara University. Ever-married men/ever-married women (referred to as respondents or individuals) from the study districts of Haripur, Muzaffarabad, Mansehra, and Shangla were recruited during 2018–2021. The majority of the population lives in rural and mountainous areas, and the data collection was possible after the availability of local resource persons and logistical arrangements. The respondents were approached in their homes, public places and community centres.
The sampling strategy was primarily convenience sampling, which ranged from cluster random sampling to door-to-door surveys depending upon the logistical, language, and cultural barriers and the COVID situation. There had been no prior selection based on the ethnicity or socio-demographic characteristics. Because the population of four districts ranged from 0.13 to 1.0 million, hence, sample size of >385 from each district deemed appropriate, considering a background consanguinity rate of 50%, precision of +/− 0.05 (5%) and 95%CI (Calculator.net). In order to reduce sampling error and maximize the likelihood of obtaining a representative sample, we expanded the sample size significantly and used cluster sampling where possible. The random walk method was used in scattered mountainous communities, with randomly selecting a direction to walk, a random starting point and sampling contiguous households. In plain urban neighbourhoods, the area was divided into geographic zones, randomly selected a zone and randomly selected a starting point within the zone.
Only those with permanent residency in a particular district and granted formal approval to their volunteer participation in the study were recruited. A structured proforma was utilized to gather information in face-to-face interviews on marital union types, demographic, and household variables. We removed the few responders who could not recall their exact marital union type.
Definitions
Four types of marital unions were considered as consanguineous, i.e., double-first-cousin (DFC), first cousin (FC), first-cousin-once-removed (FCOR), and second cousin (SC) marriages. The FC marriages were further resolved into four subtypes, i.e., father’s-brother’s-daughter (FBD), father’s-sister’s-daughter (FSD), mother’s-brother’s-daughter (MBD), and mother’s-sister’s-daughter (MSD) (Bittles, Reference Bittles, Speicher, Antonarakis and Motulsky2010; Jabeen and Malik, Reference Jabeen and Malik2014). The inbreeding coefficient F (ICF) was estimated from the weighted proportion of individual CU types multiplied with respective coefficient of inbreeding (Bittles, Reference Bittles, Speicher, Antonarakis and Motulsky2010). Hence, ICF was the aggregate of four estimates: (0.125 × DFC unions) + (0.0625 × FC unions) + (0.03125 × FCOR unions) + (0.015625 × SC unions). The demographic variables included District of residence, rural/urban origin, and mother tongue (linguistic group). Data were obtained on self-identified caste systems such as Awan, Sawati, Syed, Rajput, among others. Information was also gathered from the participants regarding their literacy and years of schooling. Literate participants were defined as those who had basic ability to read or write or were able to sign their names. For the occupational categories, self-identified occupations of participants (husbands) were documented which were later recoded to the closest categories established in the Pakistan Demographic and Health Survey (NIPS, 2013).
The household variables included family type defined as either ‘nuclear’, ‘two couples’, or ‘extended’. Nuclear and two-couples families comprised of 1 couple and 2 couples households, respectively, and extended families comprised 3 or more overlapping generations dwelling as a unit. Data on parental marriage type (consanguineous or non-consanguineous), exchange marriage (reciprocal or non-reciprocal), age of individual, year of marriage, and age at marriage were also collected (Zaman, Reference Zaman2010; Riaz et al., Reference Riaz, Mannan and Malik2016; Tufail et al., Reference Tufail, Rehman and Malik2017).
Statistical analyses
All data were maintained in Excel and analysed through GraphPad Prism (ver.5) and STATA (ver.11). Descriptive summaries were generated; the CU and ICF were calculated across the socio-demographic variables. Chi-square test and Fisher’s exact test were used to check the independence between the categorical variables (Garstman, Reference Garstman2006). Chi-square test for trend was utilized for testing the association between a nominal variable with two levels (like CU and non-consanguineous) and an ordinal variable (like age intervals). T-test was utilized to check the differences in the distributions of continuous variables. Bivariate logistic regression was employed to observe the relationship between dependent variable (consanguinity) and a single independent variable (socio-demographic). In order to observe whether two or more variables were correlated with dependent variable bivariate and multivariable logistic regression analyses were performed. The dependent variable (CU) was taken as dichotomous and the socio-demographic factors were coded as independent variables and the results were depicted in odd ratios (OR) (Garstman, Reference Garstman2006). The OR was calculated as the fraction of CU in respective category compared/fraction of CU in Reference category. In each variable, the category with the lowest prevalence of CU and sample size >10% was taken as the Reference. For the multivariable regression analyses, a stepwise logistic regression was performed and the dependent variables were added one by one. Only the significant variables were retained in the final model.
Results
Sample characteristics
A total of 6,323 respondents consented to participate in the survey out of an anticipated 6,700 people that were approached throughout the study period, making the response rate 94%. The major reasons for non-responsiveness include worries about security and privacy, social, and cultural issues, or a refusal to provide information while their family heads are not present. The number of respondents in the sample ranged from 1300 (from Mansehra) to 2023 (from Muzaffarabad). The respondents ranged in age from 15 to 80 years, with mean ages (StdDev) for men and women of 38.4 ± 10.9 and 33.8 ± 10.0, respectively.
The CU were calculated to be 55% for the total sample and the ICF was estimated to be 0.029 (Table 1). The percentages of DFC, FC, FCOR, and SC unions were 1.5%, 37.4%, 7.7%, and 8.4%, respectively. The rate of CU varied by district, from 46% in Shangla to 62% in Muzaffarabad. ICF was calculated to be 0.031, 0.030, 0.028, and 0.026 in Muzaffarabad, Haripur, Mansehra, and Shangla, respectively.
Table 1. Consanguineous Unions, Odds of Consanguinity, and ICF across Demographic Variables
* χ2 = p < 0.05;
** χ2 = p < 0.0001; OR = odd ratios depicting bivariate regression; Ref. = reference category.
FC unions were the most prevalent marriage type, occurring in 37% of respondents. FBD, FSD, MBD, and MSD types made up 39%, 17%, 21%, and 23% of this category, respectively. As a result, 56% of marriages were patrilineal as opposed to 44% of matrilineal kinds.
Consanguinity and its determinants
Consanguinity rates were significantly higher among respondents from rural areas (vs. respondents from urban areas), those who spoke Pahari and Hindko (only first two stated), and belonged to the Abbasi, Rajpoot, and Choudhary caste systems (only first three stated) (Table 1).
Consanguinity was found to be statistically significantly higher in literate respondents (compared to the illiterate group), employed in sales, unskilled manual labour, and skilled manual labour, and extended family types (compared to nuclear families) (Table 2). Additionally, the rates of CU were much greater in people with parental consanguinity and exchange marriage. The distribution of CU was statistically non-significant when factors including wife literacy, wife occupation, and family type were taken into account.
Table 2. Consanguineous Unions, Odds of Consanguinity, and ICF in Socio-economic and Household Variables
** χ 2 = p < 0.0001; OR = odd ratios depicting bivariate regression; Ref. = reference category.
In multivariable logistic regression analyses, all the variables were included, and significant variables were kept while omitting the non-significant ones (marriage year was not included due to its high collinearity with age). Thus, eight variables – district, rural origin, literacy (of husband), occupational group (of husband), family type (extended), parental consanguinity, exchange marriage, and age of husband – emerged as significant predictors of consanguinity (Table 3). All these variables were positively associated with CU. Exchange marriages contributed most to the overall effect of these factors (OR: 2.00). The individuals from Mansehra, Haripur and Muzaffarabad had 28%, 35% and 62% higher odds of having CU, respectively, compared to those from Shangla; the individuals from rural areas had 44% higher odds of CU compared to those from urban areas; individuals with parental consanguinity had 87% more odds of having CU compared to those without parental consanguinity. Exchange marriage increased the odds of CU by 100%; the individuals belonging to extended family types had 46% higher odds of having CU compared to those with nuclear family. Mother tongue appeared to be significant when district was not included in the analyses. When the analyses were stratified for marriage year, six variables emerged as significant predictors: district, rural origin, occupation of husband, parental marriage (consanguinity), exchange marriage, and extended family; all variables were positively associated with most prominent were exchange marriage, parental consanguinity and rural origin (odd ratios 1.97, 1.91, and 1.36, respectively); and model was highly significant.
Table 3. Significant Predictors of Consanguinity in the Multivariate Model
Coef. = Coefficient; Ref. = reference category; Std. Err. = Standard error.
We further hypothesized that the combination of variables potentially influencing CU among the district communities might be different from sample as a whole. Hence, the multivariable analyses were repeated in the district-wise samples (Table 4). These analyses indeed revealed that different variables were significant predictors of CU among four districts.
Table 4. Predictors of Consanguinity in Four Districts #
# Reporting odd ratios and 95% CI; NA, not applicable; –, not included due to small data size.
* p-value was statistically significant.
For instance in Shangla, four variables were significant, i.e., age of husband, literacy of husband, parental marriage type, and family type. In Muzaffarabad district, eight variables were significant predictor of CU: rural origin, age of husband, age at marriage (wife), occupation of husband, parental marriage type, exchange marriage, and family type. Three variables were observed to be significant predictors of CU in all districts, i.e., age of the husband, parental marriage type, and family type.
Temporal decline in CU rate
The rate of CU decreased with decreasing age categories of the respondents; consanguinity was substantially higher in respondents belonging to higher age groups compared to respondents in lower age groups (chi-square test for trend: p < 0.0001) (Table 2).
With regard to marriage years, the distribution of CU was evaluated, and five groups of 10-year intervals were developed. In addition to the rate of CU dropping through time, there was also a notable drop in ICF, from 0.030 in marriages ‘before 1980’ to 0.027 in marriages ‘after 2010’ (Fig. 1a). Three of the four districts – Haripur, Muzaffarabad, and Mansehra – showed a temporal drop in CU (Fig. 1b). Up to 1991–2000, there was a minor increase in CU in the Shangla district, but the rate then remained largely steady.
Figure 1. (a) Bars Showing the Percentages of CU at Y-axis and Line Depicting ICF at Y-secondary-axis; (b) Temporal Changes in the Prevalence of CU in Four Districts; (c) Temporal Trend of Subtypes of FC Unions; Increase in the Rate of MSD and Decrease in FBD are Evident; (d) Temporal Shift between Matrilineal and Patrilineal Marriages is Evident; (e) Temporal Increase in the Mean Age at Marriage; (f) Literacy Rates of Husband and Wife are Rising Over the Time, and the Rate of Exchange Marriages is Declining.
The rate of MSD marriages increased significantly among FC unions, while FBD types decreased (Fig. 1c). As a result, matrilineal marriage rates rose over time, whereas patrilineal kinds decreased (Fig. 1d). Further, an increase in the average age at marriage was evident (Table 2; Fig. 1e), that both husband and wife’s literacy rates significantly increased (Fig. 1f) and that the frequency of exchange marriages decreased over time.
Discussion
The objective of this study was to examine the factors influencing consanguinity in four populations located in northwest Pakistan. A total of 6323 individuals were recruited, and the data analysis revealed that the prevalence of consanguinity stood at 55%, with an estimated coefficient of inbreeding (ICF) of 0.029. The study identified eight variables, namely district, rural origin, age of husband, occupational group (of husband), literacy (of husband), parental consanguinity, exchange marriage, and extended family type, as significant predictors of consanguinity.
Surprisingly, rural origin – a strong predictor in the entire sample – seemed to be significant only in two districts, namely Mansehra and Muzaffarabad, and not in Shangla and Haripur. This may be explained by the fact that in many areas of Pakistan, the metropolitan areas are expanding rapidly by overtaking rural settlements. The urban conglomerations have grown significantly beyond the city borders to include the nearby ‘rural’ areas and the ‘peri-urbanization’ phenomenon has been caused (Jabeen and Malik, Reference Jabeen and Malik2014; Ullah, Reference Ullah2022). As a result, there are less differences in the demographics between urban and rural populations including the marital union types.
Exchange marriage was the most significant predictor of CU in multivariate model. The practice of exchange marriage has been deeply rooted in the cultural and social traditions of many communities of Pakistan. Such unions are often influenced by social and economic factors. They can be a way to maintain or enhance family honour and status, forge alliances between families, and distribute assets, including dowries and property (Zaman, Reference Zaman2011). Curiously however, among the district-wise analyses exchange marriage appeared to be non-significant in Shangla and Mansehra districts. These differences highlighted the cultural and social heterogeneity in the marital union decisions. Our analyses show that different communities and regions in northwest Pakistan may have varied customs regarding exchange marriages.
Nonetheless, parental consanguinity was the second most significant predictor of CU in the study sample. The type of parental marriage can indeed influence a child’s decision regarding consanguineous marriage (marriage between close blood relatives), though it is just one of many factors that can play a role in this decision. Parents who have had consanguineous marriages may have certain expectations for their children to also marry within the family or the same ethnic group. These expectations can influence a child’s decision of marrying within the family (Shenk et al., Reference Shenk, Naz and Chaudhry2021).
However, in the study sample, parental consanguinity was the second most important predictor of CU. The type of parental marriage can indeed have an impact on a child’s decision, albeit there are numerous other factors that may also come into play. Consanguineous parents may have expectations for their children’s future spouses to be from the same family or ethnicity as them. According to Shenk et al. (Reference Shenk, Naz and Chaudhry2021), these expectations may have an impact on a child’s decision to marry within the family.
Contrary to earlier studies, it is interesting to note that women’s literacy did not seem to affect the occurrence of CU in this study. Women’s education is widely argued to have a negative impact on CU frequency (Wahab and Ahmad, Reference Wahab and Ahmad1996; Fuster and Colantonio, Reference Fuster and Colantonio2004; Hamamy et al., Reference Hamamy, Antonarakis, Cavalli-Sforza, Temtamy, Romeo, Ten Kate, Bennett, Shaw, Megarbane and van Duijn2011). On the other hand, our analyses revealed that increasing consanguinity was linked to the husband’s literacy. It has been suggested that the strong correlation between consanguinity and literacy may be indirectly related to the subjects’ socio-economic status, with subjects from lower socio-economic strata not only having lower literacy levels but also having a tendency to marry more frequently outside of their close kinships (Jabeen and Malik, Reference Jabeen and Malik2014).
Additionally, the results show a considerable decline in the frequency of CUs among these populations in Northwest Pakistan, from 60% prior to 1980 to 50% between 2010 and subsequent years. The study observed a decrease in the consanguinity rate as the age categories of the respondents decreased. Additionally, it was found that over time, the average age at marriage increased, literacy rates improved, and there was a decline in exchange marriages. When comparing respondents to their parents’ marital union types, a decrease in consanguinity rate was evident. The parental consanguinity was 63% compared to the 55% of respondents. Furthermore, FC unions have decreased in frequency throughout time, from 39% to 34%. On the other hand, marriages between people who were distantly related grew from 15% to 19%, and marriages between unrelated people went from 24% to 31% (data not shown). There has been a noticeable shift in marriage patterns among FC unions, particularly with an increase in MSD marriages and a decrease in FBD types. Matrilineal marriages have seen a rise in frequency, while patrilineal marriages have experienced a decline. This trend of increasing matrilineal marriages is noteworthy, as it suggests a more influential role for elder females or mothers in marriage decisions and domestic affairs. This shift challenges the perception of Pakistani society as traditionally portrayed male-dominated. The significant increase in matrilineal marriages may also indicate a transition from patriarchy to matriarchy, as supported by a similar study conducted in South Punjab, Pakistan by Zaman (Reference Zaman2011). As expected, transitions are evident in many of the demographic variables in the study populations. For instance, both the husband and wife’s mean age at marriage increased (Fig. 1e). Similar to this, the sampled population’s literacy rates for the husband and wife both significantly increased. However, the frequency of exchange marriages has decreased over time. The rate of CU is expected to continue to drop in northwest Pakistan as a result of these demographic shifts. Previous research has revealed that Pakistan’s average marriage age has increased over time, especially in metropolitan areas (Ahmed and Rukanuddin, Reference Ahmed and Rukanuddin1987; Aziz, Reference Aziz1994). The increased accessibility of education also significantly contributed to the rise in marriageable age. Higher-educated men are more likely to be exposed to modernization pressures and have a larger pool of potential spouses to choose from outside of their immediate family (Khoury and Massad, Reference Khoury and Massad1992). The drop in CU is similarly related to rising female school enrolment and educational achievement (Schellekens et al., Reference Schellekens, Kenan and Hleihel2017).
Study limitations
This study has several limitations. The findings of this study may only apply to Northwest Pakistan and may not be generalizable to Pakistan’s total population. Second, just one generation data of marital union types have been used to compute ICF. In cultures where CU are traditionally and historically encouraged, the computation of ICF from one generation may be underestimated (Tufail et al. Reference Tufail, Rehman and Malik2017). Furthermore, the results were based on convenience sampling because of the COVID situation at the time. This study did not mention any of the numerous negative health impacts of CU, including morbidity, death, and fertility. For the occupational groups, we also used data from self-reported. Due to the wide range of occupational types and several competing scales, it was frequently difficult to identify a certain occupation. Additionally, a lot of people switch employment based on the season. So, we used the self-reported occupation types for our analysis.
Conclusion
It is concluded that continued drop in CU in Northwest Pakistan, a phenomenon that is also common in many other Middle Eastern nations and India, may be caused by rising literacy, a delayed marriageable age, urbanization, and economic improvements. Furthermore, it is very likely that the burden of recessive genetic defects will decrease, and the morbidity and death associated with consanguinity may also decrease. However, it is anticipated that the genome homozygosity that has accumulated through the generations as a result of persistent inbreeding would not decrease significantly in just one or two generations. Despite an obvious decline in the rate of CU in the study population, which is mainly due to demographic transition, there is still a lack of awareness about the health risks associated with consanguinity. Therefore, community-based counselling and education initiatives should be started to raise public knowledge of the effects of consanguinity and related risks to maternal, child, and public health.
Acknowledgments
The participation of the individuals in this study is highly acknowledged. The helpful comments of Dr. Ijaz Hussain, Department of Statistics, Quaid-i-Azam University, are highly appreciated.
Funding statement
HEC-Pakistan and PSF-Islamabad.
Competing interests
None declared.
Ethical standard
The study protocols were approved by the Ethical Review Committees of the Quaid-i-Azam University and Hazara University.
