Exploring biocultural models of chewing tobacco and paan among reproductive-aged women: Self-medication, protection, or gender inequality?

Abstract

Objectives:

Tobacco and areca nut are two of the most widely used psychoactive plant substances worldwide, yet the biocultural factors that account for variation in use patterns are not well understood. Here we attempt to understand the high prevalence of, and variation in, tobacco and areca nut use among reproductive-aged women.

Methods:

Research was carried out in Mysore, Karnataka, India. First, we conducted a qualitative investigation where participants engaged in semi-structured interviews and focus group discussions to uncover cultural norms of women’s tobacco use. Findings informed the second stage of research which quantitatively tested three models of tobacco and areca nut use (N = 74).

Results:

The qualitative study found that women were more likely to chew “natural” forms of tobacco and areca nut products (kaddipudi and paan). Quantitative tests of our hypotheses revealed that kaddipudi and combined tobacco use were best explained by the self-medication model, with somatic and environmental stress as strong indicators of use. The presence of cotinine, our biological indicator of tobacco use, was best modeled by gender inequality. We also found that men and women reported approximately equal tobacco use, even though their preferred types of tobacco and areca nut products differed.

Conclusions:

Findings did not support the protection hypothesis as it relates to plant toxins. Instead, this study suggests that women might exploit neurotoxins such as nicotine and arecoline to offset the cognitive and energetic costs associated with iron deficiency in stressful environments.

INTRODUCTION

Tobacco and areca nut are two of the most widely used psychoactive plant substances in the world and they impose significant public health burdens, especially in developing countries. For instance, more than 80% of the estimated 1.1 billion tobacco smokers worldwide reside in low- and middle-income countries (LMICs) (Gaziano & Pagidipati, 2013; WHO, 2018) and, by 2030, 80% of the 8.3 million estimated tobacco-related deaths globally are expected to occur in LMICs (Mathers & Loncar, 2006). Areca nut, the fruit of the areca palm, Areca catechu, is a main constituent of masticatory substances such as paan (also referred to as pan masala). Areca nut is an increasing public health problem in parts of southern Asia and is linked to oral submucous fibrosis (Garg, Chaturvedi, & Gupta, 2014; Gupta & Ray, 2004). Although use of tobacco and areca nut and their associated disease burdens are increasing in LMICs, little is known about the biocultural factors involved in drug use, especially among traditional rural populations. In this paper, we investigate tobacco and paan chewing among the Jenu Kurubas, a Scheduled Tribal population residing in South India. Specifically, we develop and test three hypotheses for substance use among reproductive-aged women: gender inequality, protection, and self-medication.

Gender inequality

A consistent difference between developed and developing countries is the degree to which men and women use drugs. Cross-nationally, the prevalence of use is almost always higher for men than women, and the gender difference is especially pronounced in developing countries. Whereas more than 17% of women in high-income countries smoke tobacco, less than 4% of women in low-income countries smoke tobacco (Ng et al., 2014; WHO, 2018). Conventional explanations for low female use tend to focus on gender inequality: that is, in developing countries women tend to have lower social, economic, and/or political status than men (e.g., Hitchman & Fong, 2011).

A main limitation of the gender inequality model is that it cannot easily account for gender differences in smoking rates among highly egalitarian small-scale populations. For example, among groups of Congo Basin hunter-gatherers—populations noted for their high degree of egalitarianism—there is almost always a male-bias in smoking (Roulette & Hewlett, 2018), and in some cases, such as among the Aka, the differences are some of the largest recorded cross-nationally (Roulette, Hagen, & Hewlett, 2016). This raises the possibility that other factors also contribute to low female smoking rates.

Protection

Rather than socioeconomic factors, gender differences in drug use could be influenced by biological sex differences (Hagen, Garfield, & Sullivan, 2016; Hagen, Roulette, & Sullivan, 2013). Compared to men, women—especially pregnant women—have more fungiform papillae and taste buds and are better able to detect bitter tasting compounds (Bartoshuk, Duffy, & Miller, 1994; Prutkin et al., 2000). During pregnancy specifically, women are more vulnerable to the effects of toxins and pathogens due to immunological changes that promote placentation, organogenesis, and subsequent fetal growth (Fessler, 2002; Mor & Cardenas, 2010). Theory and evidence suggest that these biological shifts are associated with psychobehavioral changes, such as heightened aversions and subsequent avoidance of toxic plants that could harm the fetus (Hook, 1978; Placek & Hagen, 2015; Placek, Madhivanan, & Hagen, 2017; Tierson, Olsen, & Hook, 1985). Many plants contain compounds with teratogenic effects, and some, like nicotine (the principal psychoactive compound in tobacco) and arecoline (a psychoactive compound related to nicotine that is found in areca nuts), are neurotoxins that disrupt central nervous system (CNS) development (Hagen et al., 2013). Observations among active and former hunting-gathering populations support this hypothesis, but also indicate that avoidance of toxins is influenced by socially-transmitted information (Placek et al., 2017; Roulette et al., 2016). Hagen et al. (2016), however, found support for both the gender inequality and fetal protection models: in multiple regression models of female smoking prevalence in 173 countries, gender inequality and total fertility rates were significant negative predictors.

Another biological shift in pregnancy, the maternal-fetal transfer of iron, is hypothesized to be a protection mechanism that limits pathogen proliferation in pregnant women (Denic & Agarwal, 2007; Miller, 2016). The likelihood of becoming anemic, however, increases with high fertility rates in populations where women lack the birth spacing and resources needed for iron repletion (Miller, 2010). Although this transfer of iron leaves women vulnerable to developing anemia, it is nonetheless considered part of the repertoire of traits associated with pathogen avoidance in pregnancy and during women’s reproductive years (Fessler, 2002). Iron deficiency anemia, for example, commonly co-occurs with the consumption of clay in pregnancy. Clay consumption, or geophagy, sometimes provides bioavailable iron to anemic consumers, but is more often a source of protection from ingested toxins and pathogens (Young, Sherman, Lucks, Pelto, & Rowe, 2011). Researchers have also found that tobacco consumption in pregnant women is associated with lower hemoglobin levels (Ganganahalli, Pratinidhi, Patil, & Kakade, 2015; Gupta & Sreevidya, 2004; Subramoney & Gupta, 2008), suggesting the possibility that tobacco use could function to limit pathogen exposure; however, the biological mechanism between iron bioavailability and tobacco is unclear, suggesting the possibility of a third, unknown factor.

In summary, the protection hypothesis predicts that women will avoid tobacco and other toxic plants during periods of heightened vulnerability, such as pregnancy, as a toxin-avoidance strategy, and that avoidance will increase in women residing in resource-scarce environments who suffer from low iron and have high-fertility. On the other hand, immunologically and nutritionally constrained women might seek tobacco as a strategy to limit pathogen exposure, perhaps as an immediate defense mechanism that comes with future costs on reproduction, such as lower fertility or adverse fetal health outcomes.

Self-Medication

A complementary approach, called the neurotoxin regulation model of drug use, instead views sex (and age) differences in drug use as reflecting the relative fitness costs and benefits of neurotoxin exposure (Hagen et al., 2013). If the benefits outweigh the costs, then neurotoxin ingestion can be expected. The protection model, however, is founded primarily on the cost of teratogenesis, which is one major and obvious fitness cost associated with female consumption of neurotoxins, but does not consider the benefits of consuming plant neurotoxins during the reproductive life span.

Self-medication is one potential benefit of regulated exposure to neurotoxins. Several animal species, for example, are known to exploit plant defensive toxins to modulate infections with intestinal parasites (Huffman, 1997), and intriguingly, most globally popular recreational drugs have anthelmintic properties. Hence recreational drug use might serve a similar anthelmintic function in humans. Roulette et al. (2014; 2016) found support for this hypothesis in their investigation of tobacco and cannabis smoking among Aka foragers of the Congo Basin. Not only were cannabis and tobacco smoking levels inversely associated with helminth burdens in cross-sectional and longitudinal studies, but in a double-blind, placebo-controlled trial, smoking levels in the treatment group decreased two-weeks after treatment with a commercial anthelmintic relative to the control group, suggesting that infection moderates smoking levels.

In addition, many neurotransmitters within the mammalian nervous system, such as the acetylcholines and monoamines—which are frequently implicated in drug use—are evolutionarily ancient and ubiquitous across all eukaryotes including plants. By virtue of their similarity across taxa, when exploited, these compounds can interact with and even enhance the CNS of other organisms, including humans (Kennedy & Wightman, 2011). In effect, these compounds serve as neurotransmitter-analogs (Sullivan & Hagen, 2002). For example, in a meta-analysis, Heishman, Kleykamp, and Singleton (2010) found that nicotine enhanced performance in six domains related to motor abilities, attention and memory.

Monoamine and cholinergic neurotransmitters are also nutritionally constrained in the environment and are activated by physiological and psychological stress (Sullivan & Hagen, 2002). Research shows that chronic deficits of bioavailable iron can have detrimental and long-lasting impacts on the brain, particularly the dopamine-opiate and cholinergic systems (Youdim & Yehuda, 2000). Low iron also impairs monoamine metabolism, leading to long-term deficits in memory, learning, and motor function (Kim & Wessling-Resnick, 2014). Furthermore, iron deficiency can impair immune function and cause fatigue, which can negatively impact work performance (Basta, Soekirman, Karyadi, & Scrimshaw, 1979; Beard, 2001). Thus, exploitation of psychoactive plants might help alleviate constraints on brain-signaling processes and prevent further neurotransmitter depletion during times of nutritional stress, particularly among individuals suffering from iron deficiency.

Indeed, most drugs used by indigenous populations reportedly increase tolerance for fatigue, hunger, and thermal stress in aversive situations (for review, see Sullivan & Hagen, 2002). Relatedly, some recreational neurotoxins, such as caffeine and nicotine, are potentially ergogenic in that they have been demonstrated in lab settings to enhance physical performance, stamina, memory and/or recovery (Elrod, Buccafusco, & Jackson, 1988; Pesta, Angadi, Burtscher, & Roberts, 2013; Thiel, Zilles, & Fink, 2005). Given the effects iron deficiency has on the brain and the role nicotine has in improving cognitive deficits, perhaps those women who suffer from iron deficiency anemia seek out plant toxins such as nicotine or arecoline to enhance their energy and other cognitive faculties.

The effects of iron deficiency on tobacco and areca nut use are likely to be exacerbated by environmental stress. Stress is induced by environmental harshness and unpredictability, which in life history theory refers to high morbidity and mortality and its spatial-temporal variation, respectively (Ellis, Figueredo, Brumbach, & Schlomer, 2009). There are abundant data indicating that humans respond to environmental stress by shifting investments in somatic maintenance and reproductive effort, and these have long-lasting biobehavioral effects. When harshness and/or unpredictability are high, individuals are expected to display a “fast” life history strategy, which is a set of correlated traits that increase timing of puberty, age at sexual debut and first birth, and parental investment strategies (e.g., Placek & Quinlan, 2012; Quinlan, 2007). Substance use is often associated with environmental harshness and unpredictability, a factor associated with fast life history strategies (Decker & Flinn, 2011; Hill & Chow, 2002; Richardson & Hardesty, 2012; Richardson et al. 2014); however, the most comprehensive study to date found very little evidence that substance use itself is an adaptation for mating success (Richardson et al., 2017). One potential explanation for the association between substance use and life history characteristics is the neurotransmitter-analog hypothesis we described above (Sullivan & Hagen, 2002). When both environmental risk and associated nutritional stress are high, substance use might be expected as a means to defend against neurotransmitter depletion.

The current study used mixed-methods to investigate the factors that give rise to tobacco and pan masala (also referred herein as paan) chewing among reproductive-aged women in South India. Specifically, we first conducted a qualitative study to understand emic perspectives of tobacco and paan use, which helped formulate hypotheses that were tested in the second part of the study, where we tested the following hypotheses: gender inequality, protection, and self-medication.

Hypothesis 1, Gender inequality:

The gender inequality model predicts that women who have social and economic equality with men will have rates of tobacco use that are comparable to men. We first tested this by comparing women and men’s tobacco and paan use. We predicted that women will also be more likely to use tobacco when they are in control of household finances, have higher education, and work outside of the home.

Hypothesis 2, Protection:

The protection hypothesis predicts that women will avoid plant neurotoxins when they are more vulnerable to the effects of consumption, for example, during periods of high fertility and pregnancy (Hook, 1978; Miller, 2010, 2016). We predicted that tobacco and areca nut avoidance would be predicted by high fertility and low iron, two indicators of somatic vulnerability whereby women need enhanced protection from toxins. Social learning also plays an important role for fetal protection strategies, in that maternal relatives often inform reproducing women of what they should avoid in order to enhance fetal health outcomes (Placek et al., 2017). In the current study, we predict that tobacco and paan use would not be acquired from maternal relatives if tobacco avoidance functions as a toxin-protection strategy. On the other hand, if tobacco and areca nut function as pathogen-avoidance strategies, then findings should reveal a positive relationship with iron deficiency and tobacco and areca nut use, within the context of high fertility. However, we are presented with an unavoidable confound given that existing research has found an association between tobacco use and low hemoglobin among pregnant women (Ganganahalli, Pratinidhi, Patil, & Kakade, 2015; Gupta & Sreevidya, 2004; Subramoney & Gupta, 2008), and might, therefore, be the baseline relationship between tobacco use and hemoglobin in reproductive-aged women as well.

Hypothesis 3, Self-Medication:

The self-medication hypothesis predicts that neurotransmitter depletion brought on by somatic and environmental stressors leads individuals to seek out substances that enhance CNS functioning, particularly in terms of performance, attention, and memory (Heishman et al., 2010; Sullivan & Hagen, 2002). We predicted that reproductive-aged women faced with heightened somatic and environmental stress would be more likely to use tobacco and paan. In contrast to the toxin-protection hypothesis, but similar to pathogen avoidance, low iron, our measure of somatic stress, would be positively associated with tobacco and paan use, particularly because nicotine and arecoline could function to enhance memory, attention, and work performance in individuals who experience the cognitive deficits and CNS impairment associated with chronic iron deficiency (Lozoff, 2011; Youdim & Yehuda, 2000). This relationship would be situated in the context of environmental risk and energetic stress, and qualitative data will suggest that women use tobacco and areca nut to enhance work performance.

Study Population

The current study was conducted in South India with the Jenu Kurubas (“honey collectors”), a Scheduled Tribe occupying the border areas of Karnataka, Tamil Nadu, and Kerala. Scheduled Tribes, the most marginalized group in India with the greatest health disparities, are among the highest users of tobacco and other substances (Mohindra & Labonté, 2010; Rani, Bonu, Jha, Nguyen, & Jamjoum, 2003). They are former hunter-gatherers who were displaced from their residential forest approximately 25–30 years ago and were forced into government housing where they were given small plots of land to grow crops for consumption or sale (Roy, Hegde, Bhattacharya, Upadhya, & Kholkute, 2015). As a consequence of their displacement, Jenu Kurubas have changed their primary mode of subsistence; as hunter-gatherers, they collected honey and hunted wild game; however, now they receive monthly provisions from the government consisting of rice, ragi wheat, eggs, jaggery, oil, and pulses. Despite assistance from the government, Jenu Kurubas remain food insecure (Placek et al., 2017). For income, they engage in daily-wage agricultural work, where many of them spend several months away working on distant tobacco and coffee plantations.

Jenu Kuruba women engage in agricultural work, but are the primary homemakers and child caregivers. Furthermore, they do not use any form of traditional or modern contraception, and girls often begin reproduction after their first menses and continue sexual activity throughout their reproductive life span.

METHODS

Research took place from October 2016 to July 2017. This study was conducted in collaboration with The Public Health Research Institute of India (PHRII). This research was reviewed and approved by the Institutional Review Boards at Florida International University and PHRII. Literate participants provided written informed consent, and the others provided verbal consent and thumbprints to satisfy PHRII IRB requirements. Herein we present two separate studies, a qualitative and quantitative investigation. They are presented as separate studies because results from the qualitative study were used to generate measures for the quantitative study.

Part I: Qualitative investigation

The first phase of this study aimed to uncover the types of substances used by women and men in the community. For this phase, we conducted semi-structured interviews with 30 community members, consisting of both men and women. This sample size was determined through an iterative process; data were continuously analyzed and recruitment ceased once thematic saturation was reached, meaning that no new information was being generated during the interview process (Galletta, 2013).

We used findings from the semi-structured interviews to generate a list of in-depth questions for focus group discussions (FGDs). FGDs were conducted with 3 groups of women and 3 groups of men for a total N = 41. Participants for both the semi-structured interviews and FGDs were ages 18–35 and were recruited using convenience sampling. Twelve (29.3%) of the participants who participated in the semi-structured interviews also participated in the FGDs. FGD questions were geared toward understanding gender differences in substance use, how individuals learn about consuming each substance, and reasons why substances are used. Similar to semi-structured interviews, recruitment ceased once we reached thematic saturation (Galletta, 2013; Rabiee, 2004).

Analyses

Free-listed substances from semi-structured questionnaires were analyzed by computing salience scores in Microsoft Excel 2017. Salience analysis is a method used to identify the main items in a cultural domain by computing a score for each item (in this case, substances) that reflects both the frequency and order of mention (Quinlan, 2005). The composite salience score is computed by averaging the scores across participants. FGD data were analyzed using an iterative coding process (LeCompte & Schensul, 2012). The first author, C.P., first developed a codebook that consisted of key terms relevant to the study, such as substance, purpose of use, benefits, and negative effects. She held a training session for N.H., A.K., and K.R. to blindly code the FGDs according to the codebook. Once the coders were finished, C.P. found consensus among their codes, and then themes were identified by N.H. and C.P.

Results

Semi-structured Interviews.

Participants (N = 30) were an average age of 23.77 and had an average 8.9 years of education. Participants worked as agricultural laborers (90.1%), ASHA workers (3.3%), domestic homemakers (3.3%), and cooks for the local school (3.3%). Participants were most likely to state that “alcohol” is most frequently consumed by Jenu Kurubas, followed by a series of substances that fell within the “tobacco” and “paan” classifications (see Table 1 for a description of each). Cannabis (bhang) had low salience; many participants claimed that bhang is expensive and difficult to acquire in the rural setting.

Table 1.

Descriptions of tobacco products identified from qualitative investigation. Jenu Kuruba women’s self-reported use is also included.

Substance	Description	Number of women users N (%)
Beedis	Tobacco-filled leaves that are rolled and smoked	0 (0%)
Cigarettes	N/A	0 (0%)
Hans	Processed, chewing, tobacco	6 (8.1%)
Hogesoppu	Crushed and fresh tobacco	7 (9.5%)
Kaddipudi	Loose and dried tobacco	32 (43.2%)
Paan/Pan masala	Betel nut and slaked lime wrapped in betel leaves, often mixed with kaddipudi	47 (63.5%)
Panparag	Betel nut mixed with lime and spices	1 (1.4%)
Vimal/Gutkha	Crushed areca nut, tobacco, paraffin wax, slaked lime and added flavorings	3 (4.1%)

Focus Group Discussions.

Jenu Kuruba FGD participants consisted of 21 women and 20 men. Their average age was 25.07, and average years of education was 8.5. The majority of participants were agricultural workers (78.0%), and few held alternative occupations (12.2%; e.g., ASHA, Anganvadi). Only four were domestic homemakers (9.8%).

Results from the FGDs revealed that when Jenu Kurubas refer to tambaku (tobacco), they are referring to products that are smoked or chewed and contain tobacco, areca nut, or some combination of the two. Jenu Kurubas stated that kaddipudi is a common tobacco product, which consists of tobacco leaves chewed in a natural, powdered form, but is also included with paan, which consists of betel leaves, betel nut, and slaked lime. Hogesoppu is fresh tobacco that is chewed without any additional ingredients, but is sometimes mixed with paan as well. Other commonly used substances include commercial betel nut and tobacco products, such as Vimal, Gutka, Panparag and Hans. Beedis and cigarettes are also available in the community.

Respondents indicated that tobacco use varies according to cultural norms pertaining to gender and age. When asked why these differences exist, one participant simply stated, “That is their practice.” Another participant highlighted that the community enforces gender differences in tobacco use, for example, women do not chew or smoke the same substances commonly chewed and smoked by men. Women are more likely than men to chew kaddipudi and paan, and sometimes chew hogesoppu. The other tobacco items, such as commercial tobacco and betel nut products, are most often used by children, men, and the elderly. Girls, for example, reportedly chew commercial tobacco products, but switch to more “acceptable” forms of tobacco once they get older (i.e., kaddipudi, paan, and hogesoppu). Participants stated that the age of onset for tobacco chewing for girls is between 6 years and late adolescence. Elderly women sometimes smoke beedis; although this is rare, and young girls and women have not been seen smoking this particular form of tobacco.

Participants reported that women mainly acquired their tobacco habits through observing others. Young girls, for example, learn by observing mothers, elders, community members, and peers (once they become adolescents). One participant stated: “Elders use beedis, hans (commercial chewing tobacco), paanparag and kaddipudi in front of children, who will then think it is alright to use all of these things. It is in our hands [adults]; we have to stop using them and the kids will stop.” Participants also stated that paan, specifically, is a traditional substance with deep roots in the community, which explains why young girls start chewing it. Aside from observational learning of cultural traditions, one participant stated that children are given these substances to make them stop crying.

Paan was the only item used by women for medicinal reasons, and is sometimes used to cure nausea and vomiting or to alleviate toothache. Participants were more inclined to discuss the performance enhancing effects of the substances, such as the ability of kaddipudi to improve work performance. Confirming this purpose of kaddipudi, one participant stated, “While working, women have [tobacco] all the time.” Chewing kaddipudi can also lead to cognitive changes, such as “giddiness,” or vertigo (“they feel giddiness, when they eat, so they want to use it”). Participants indicated that chewing paan was useful for digestion. The final use for substances by women was habit; participants simply indicated that, over time, using the products becomes habitual and addictive since people often use them to pass time.

Overall, we identified the following themes of tobacco use for women: teaching and learning, medicine, work enhancement, cognitive impact, digestive aid, and habit. Participants did not mention any general threats to health for using paan, kaddipudi, and hogesoppu, nor did they mention any harms if used by women during pregnancy, in the post-partum period, or during their reproductive years. According to participants, however, beedis can cause sleepiness, which can make working difficult, as well as cause cancer and asthma.

Part II: Quantitative and Biological Investigation

After completing the qualitative investigation, the research team began recruiting participants for the quantitative phase of the study, where we tested our three hypotheses.

PHRII research staff used convenience sampling to locate participants because many Jenu Kuruba adults were engaged in agricultural work at distant farms. Our final sample size included 75 women and 75 men, ages 18–35. Data were collected from each participant in their respective village (haadi) community center, which allowed for privacy. Consenting participants completed a structured questionnaire that included the study measures. In addition, they filled in a chart that contained 23 substances derived from the qualitative investigation. A counselor showed each participant the image of the substance on an iPad and asked if they used it (yes or no), how frequently (Likert scale 1–9; 1 = never, and 9 = everyday), age of onset, and from whom they acquired the habit. Finally, a PHRII nurse and lab member set up a temporary clinic and lab, where participants provided a sample of whole venous blood (3 mL) to assess hemoglobin and Herpes-Simplex 2 virus; urine to assess cotinine, a nicotine metabolite, and THC, the main psychoactive cannabinoid found in cannabis. In total, study procedures took approximately one and a half hours per participant to complete. Participants were compensated an amount in accordance with local norms.

Measures

Outcome variables Outcome measures were selected by plotting data to determine which tambaku products women reported chewing or smoking (see Fig. 1 for a comparison between self-reported use for men and women). Items were included as outcome variables if self-reported use was 10% or higher. The only items that met these criteria were kaddipudi and paan, which align with findings from the qualitative study. Inspection of these variables revealed that they were not normally distributed; each variable had both a high proportion of zeros (paan: 36%; kaddipudi: 56%) and a high proportion of women reported being daily users (paan: 39%; kaddipudi: 31%). Consequently, new variables were created that consisted of three levels to assess these varied patterns of use: never = 0, sometimes = 1 (very seldom to nearly every day, or 1–7 on the frequency range), and daily = 3 (once a day to multiple times a day, or 8–9 on the frequency range). An additional “combined tobacco” variable was generated that included kaddipudi and paan (0 = neither, 1 = either kaddipudi and paan, and 2 = both). Finally, qualitative cotinine (yes = 1, no = 0) was used as a biological indicator of tobacco use. Qualitative cotinine was measured with a One-Step Rapid Nicotine Urine Test (Craig Medical, Vista, CA). This test has a detection sensitivity of 200 ng/ml or higher, which is sensitive enough to distinguish environmental tobacco smoke from tobacco use.

Fig. 1:

Jenu Kuruba female and male preferences of tobacco and frequency of use.

Models

Gender inequality:

First, frequency of tobacco and paan use along with the presence/absence of cotinine, were compared between men and women (Fig. 1). Then, a model was fitted to include the following indicators of social and economic equality: women’s control over household finances (yes = 1, no = 0), education (years), and whether or not the woman works outside of the home (yes = 1, no = 0; note that this was inclusive of any occupation).

Protection:

Fertility was measured by asking women to report their total number of living offspring. Anemia was measured with hemoglobin. Hemoglobin levels were assessed with blood samples using the Hemoglobin (Cyanomethemoglobin) Beacon Test (Beacon Diagnostics Pvt. Ltd. Kabilpore, Gujarat, India). For the protection hypothesis, we predicted that lower hemoglobin would be correlated with avoidance of tobacco and paan. We assessed the effects of learning on women’s tobacco and paan use in two ways. First, we plotted women’s self-reported learning of tobacco and paan use. Then, women’s self-reported teachers were recoded into a dichotomous variable measuring members of her matriline (mothers and grandmothers). Participants were also asked to list anyone in their household who has a history of substance use. Responses from this variable were recoded into a dichotomous variable to capture the presence of mothers and grandmothers who had a history of using any form of tobacco (yes = 1, no = 0).

Self-medication:

First, a total environmental stress score was computed. Environmental stress was measured with an 8-item scale developed during the qualitative investigation. Questions were framed within a 12-month timeline (e.g., “In the past 12 months, how often have you experienced stress from people using substances in your community?”). Participants answered according to a 9-point Likert scale (0 = never, 8 = at least once a day). Hemoglobin was included as a measure of somatic stress, and was predicted to be inversely associated with tobacco use, such that anemic women would be more likely to use tobacco in order to enhance energy or cognitive performance. Occupation as an agricultural worker was included as a measure of energetic stress, since this work requires long hours, traveling to a far distance, and low pay, and work performance might be impaired by chronic low iron.

Age was included as a control variable in each model.

Analyses

Data were analyzed in R version 3.3.3 for Macintosh. Summary statistics were computed for all variables. Table 2 provides summary statistics for raw values of predictor and outcome measures. A heatmap was used to explore social learning among participants. Heatmaps are visual matrix representations of the relationship between variables. The Euclidean metric and Ward agglomeration algorithm was used to cluster rows and columns. Density of values are displayed by variation in color scheme (Fig. 2). All continuous and ordinal measures were scaled and centered.

Table 2:

Descriptive statistics for outcome and predictor variables for quantitative models

Outcome Variables	Presence/Yes	Mean	Median	SD	Range
Cotinine	57 (77%)
Kaddipudi frequency (recoded)		0.74	0.00	0.91	0.00 – 2.00
Paan frequency (recoded)		1.03	1.00	0.88	0.00 – 2.00
Combined tobacco		1.07	1.00	0.78	0.00 – 2.00
Predictor Variables
Woman manages finances	16 (22%)
Woman works outside home	47 (64%)
Woman works in agriculture	43 (58%)
Hemoglobin (g/dL)		10.95	11.10	2.05	5.30 – 15.90
Current Stress		5.27	3.00	4.88	0.00 – 18.00
Age		26.95	26.00	5.54	18.00 – 35.00
Number of children		1.91	2.00	1.15	0.00 – 6.00
Education		6.10	6.00	3.43	0.00 – 12.00

Fig. 2:

A heatmap displaying the modes of learning for each tobacco and areca nut product for Jenu Kuruba women.

Ordered logistic regression was used to model frequency of kaddipudi and paan use, along with combined tobacco use. A binary logistic regression modeled the presence/absence of cotinine. Where appropriate, odds ratios are reported, and all models include 95% confidence intervals. Support for hypotheses was determined by examining individual relationships between predictors and outcome variables, and the “best” explanatory models were selected by using Aikake’s Information Criterion corrected for small samples (AICc).

Results

Participants included 75 non-pregnant women, who were an average age of 26.95 years and had completed 6.1 years of education. The average fertility was 1.9. Summary statistics for the remaining study variables are presented in Table 2. Women reported chewing kaddipudi (n = 32; 43%) and paan (n = 47; 64%). Male and female cotinine values were similar: 59 men (79%) were positive, compared to 57 (77%) women. Only 5 (6.8%) participants reported a matrilineal history of tobacco or paan use in their respected households. Of the 5 cases, grandmothers were not mentioned. Only 2 of the 32 (6.3%) self-reported kaddipudi users stated that they acquired the habit from their mother or grandmother, whereas 3 of the 47 self-reported paan users reported acquisition from their mother or grandmother. Given these low rates of matrilineal influence, this measure was not included in the protection model. The heatmap presents all of the teachers reported by women: kaddipudi was learned from relatives, mothers-in-law, coworkers, mothers, grandmothers, sisters, neighbors, and self, whereas paan chewing was learned from relatives, coworkers, mother, parents, grandmother, sisters, neighbors, and self (Fig. 2).

Kaddipudi.

The self-medication model had the lowest AIC (Table 3). In this model, frequency of kaddipudi chewing was negatively and significantly associated with hemoglobin (OR = 0.40, p < 0.0001; 95% CI = 0.22, 0.71), positively and significantly associated with being an agricultural worker (OR = 1.90, p < 0.02; 95% CI = 1.11, 3.42), and positively and significantly associated with current stress (OR = 1.83, p < 0.02; 95% CI = 1.09, 3.17).

Table 3.

Results from a priori tests of kaddipudi, paan, combined tobacco, and presence/absence of cotinine in Jenu Kuruba women.

Model: Protection
Variables	Kaddipuddi OR (95% CI)	Paan OR (95% CI)	Combined Tobacco OR (95% CI)	Cotinine OR (95% CI)
Number of children	0.82 (0.46, 1.45)	1.43 (0.84, 2.52)	1.08 (0.64, 1.82)	1.41 (0.73, 2.84)
Hemoglobin	0.38 (0.21, 0.64)**	0.62 (0.37, 1.00)†	0.48 (0.29, 0.77)**	0.62 (0.32, 1.14)
Age	1.38 (0.80, 2.43)	1.50 (0.92, 2.50)	1.43 (0.87, 2.36)	1.01 (0.53, 1.93)
Observations	74	74	74	74
Null deviance (df)				79.77
Residual deviance (df)	124.45	147.71	147.97	75.66
AIC	134.45	157.72	157.97	83.67
Model: Self-Medication
Variables	Kaddipuddi OR (95% CI)	Paan OR (95% CI)	Combined Tobacco OR (95% CI)	Cotinine OR (95% CI)
Hemoglobin	0.40 (0.22, 0.71)**	0.62 (0.37, 1.00)†	0.50 (0.30, 0.82)**	0.59 (0.29, 1.07)†
Current stress	1.83 (1.09, 3.17)*	1.20 (0.74, 1.97)	1.82 (1.12, 3.05)*	0.85 (0.46, 1.59)
Agricultural worker	1.90 (1.11, 3.42)*	1.19 (0.76, 1.86)	1.63 (1.03, 2.62)*	1.29 (0.73, 2.28)
Age	0.99 (0.59, 1.65)	1.60 (0.99, 2.62)†	1.18 (0.73, 1.91)	1.16 (0.62, 2.23)
Observations	74	74	74	74
Null deviance (df)				79.77
Residual deviance (df)	115.39	148.41	138.84	75.63
AIC	127.39	160.41	150.84	85.63
	Model: Gender Inequality
Variables	Kaddipuddi OR (95% CI)	Paan OR (95% CI)	Combined Tobacco OR (95% CI)	Cotinine OR (95% CI)
Woman - finances	0.82 (0.43, 1.50)	1.04 (0.61, 1.80)	0.82 (0.48, 1.37)	2.16 (0.94, 7.70)
Woman - work	1.49 (0.90, 2.58)	1.10 (0.70, 1.73)	1.30 (0.83, 2.06)	1.31 (0.71, 2.43)
Education	0.31 (0.16, 0.55)***	0.66 (0.39, 1.08)†	0.46 (0.27, 0.74)**	0.29 (0.10, 0.66)*
Number of children	1.31 (0.75, 2.34)	1.65 (0.98, 2.90)†	1.39 (0.84, 2.37)	2.03 (0.98, 4.63)
Age	0.77 (0.38, 1.45)	1.18 (0.68, 2.03)	1.07 (0.62, 1.85)	0.56 (0.26, 1.16)
Observations	74	74	74	74
Null deviance (df)				79.77
Residual deviance (df)	118.88	148.24	145.96	62.70
AIC	132.88	162.24	159.96	74.70

^†p ≤ 0.10

^*p ≤ 0.05

^**p ≤ 0.01

^***p ≤ 0.001

In the protection model, hemoglobin was also significantly associated with kaddipudi chewing; however, the direction of the relationship between variables is in line with the self-medication model, such that lower hemoglobin increased the odds of chewing kaddipudi (OR = 0.38, p < 0.0001; 95% CI = 0.21, 0.64). In the gender inequality model, chewing kaddipudi was negatively and significantly associated with education (OR = 0.31, p < 0.0001; 95% CI = 0.16, 0.55).

Paan.

The protection model had the lowest AIC (Table 3); however, no variables were significant at the 0.05 alpha level.

Combined tobacco use.

The lowest AIC for combined tobacco use was the self-medication model (Table 3). Combined tobacco was negatively and significantly associated with hemoglobin (OR = 0.50, p < 0.01; 95% CI = 0.3, 0.82), positively and significantly associated with being an agricultural worker (OR = 1.63, p < 0.04; 95% CI = 1.03, 2.62), and positively and significantly associated with current stress (OR = 1.82, p < 0.02; 95% CI = 1.12, 3.05).

In the protection model, hemoglobin was also significantly associated with combined tobacco; however, the direction of the relationship between variables is in line with the pathogen-protection or self-medication model, such that lower hemoglobin increased the odds of using both tobacco substances (OR = 0.48, p < 0.0001; 95% CI = 0.29, 0.77). In the gender inequality model, combined tobacco use was negatively and significantly associated with education (OR = 0.46, p < 0.0001; 95% CI = 0.27, 0.74).

Cotinine.

In contrast to self-reported tobacco and paan use, cotinine was best modeled by gender inequality (Table 3); however, the relationship between education and cotinine was not in the predicted direction: lower education predicted higher odds of urinary cotinine (OR = 0.29, p < 0.01; 95% CI = 0.1, 0.66).

DISCUSSION

Smokeless tobacco and areca nut chewing are widely used in developing countries, yet the biocultural factors contributing to their use among reproductive-aged women are unclear. Results from the qualitative study revealed that women are more likely to chew “natural” forms of tobacco including kaddipudi, paan, and hoggesoppu. Participants stated that women acquire the habit of using tobacco by observing others or through experiential learning, such as when parents give children tobacco with the intent to soothe them. Participants also stated that women used tobacco as a means to improve their work performance.

In the quantitative investigation, the self-medication model was the best fit for kaddipudi and combined tobacco use. The relationship between hemoglobin and tobacco use supported our prediction that women exploit tobacco under heightened somatic stress, perhaps in order to enhance CNS functioning in those who experience cognitive deficits from chronic iron deficiency. Working as an agricultural laborer and exposure to environmental stress were also meaningful predictors for kaddipudi and combined tobacco use, providing further support that women use these substances to enhance cognition or work performance.

Although we cannot pinpoint the biological mechanism underlying smokeless tobacco use and hemoglobin levels, our results point to the possibility that iron deficient women might be exploiting tobacco and areca nut in order to enhance cognitive functioning and work performance. These substances could also function as a pathogen avoidance strategy, as revealed in the kaddipudi and combined tobacco self-medication models; however the other predictors for pathogen protection were not significant. Further investigations are warranted, however, because reproductive-aged women might exploit the teratogenic and toxic properties of psychoactive drugs for other reasons. First of all, in the current study, the relationships between tobacco, anemia, and agricultural work could be reflective of labor exploitation. This is supported by the fact that the Jenu Kurubas are a marginalized population that have been exploited by farmers and other organizations to conduct agricultural work since they became resettled. One key informant in Mysore, an anthropologist who studied the Jenu Kurubas after they first became settled, highlighted that different groups incentivized the tribe members with tobacco and alcohol in exchange for labor (Shanthi Vasanthagopalan, personal communication). The possibility of labor exploitation is in line with other research on indigenous populations who have been induced by neighboring groups with greater socioeconomic power to perform labor in exchange for addictive substances (Jankowiak & Bradburd, 2003; Jankowiak & Bradburd, 1996; Mintz, 1986).

Furthermore, women might exploit plant neurotoxins in order to control their fertility. Neurotoxins, such as nicotine and arecoline, could function as natural birth control through the process of fertility suppression or serving as abortifacients (Holloway, Kellenberger, & Petrik, 2006; Kumar & Shankar, 2015). This might be especially important for mothers during times of energetic and nutritional constraints when having additional dependents could decrease the amount of parental investment allocated to older, dependent children. In the current study, this hypothesis is supported by the relatively low fertility rates of this population compared to other resettled hunter gatherers (Bentley, Goldberg, & Jasieńska, 1993), where one might expect to see less plant toxin exploitation in association with higher fertility. Finally, neurotoxins—especially in resource-scarce and pathogen-dense environments—could be used as a social bargaining tool to signal a need for increased social support by “threatening” to consume dangerous substances (e.g., Placek, 2017).

In support of the gender inequality model, men and women reported approximately equal tobacco use, which was expected given that Jenu Kurubas are traditionally an egalitarian society. Interestingly, however, we found gender differences in tobacco preferences. Men are more likely to smoke beedis and chew commercial tobacco and paan products, whereas women were less likely to report smoking and reported chewing “natural” forms of tobacco. Perhaps these differences are due to perceived gender differences among the Jenu Kurubas themselves, but was not uncovered in this study.

Although the protection model was the best fit for paan chewing, none of the variables were statistically significant. Further research is needed to understand the underlying factors leading women to use paan and related areca nut products, particularly because the rates among South Indian women are high (Gupta & Warnakulasuriya, 2002).

Our prediction for social learning, that women would acquire the habit of tobacco from their matrilineal relatives, was not supported. One explanation is that the Jenu Kurubas in this region might rely less on matrilineal social learning for reproductive and health knowledge due to egalitarian values of autonomy, similar to a prior study in this region (Placek et al., 2017).

The current study has several noteworthy implications for public health. The relationship between iron deficiency and tobacco use deserves further attention, particularly because reproductive-aged Indian women and girls suffer from high rates of anemia and smokeless tobacco use among them is on the rise (Gupta & Ray, 2004), and health care providers often overlook women’s use (Bhaumik et al. 2018). These findings highlight the need for public health efforts to tackle anemia in conjunction with tobacco use in women, rather than investigating them as isolated health concerns. Furthermore, public health interventions need to consider gendered nuances in tobacco preference in order to tailor gender-specific anti-tobacco messages. Currently, anti-tobacco efforts in India largely target commercial products, which are not typically used by Jenu Kuruba women. By not including “natural” products in anti-tobacco use interventions, women might perceive these substances as being safe to consume (Banerjee et al., 2014).

One limitation of our study is that the biological measure for tobacco use could only provide the presence or absence of cotinine, and not the amount of cotinine in the participants’ body. Similarly, although we asked participants to identify frequency of tobacco and paan use according to a Likert scale, their responses were not normally distributed, which led us to re-categorize the variable subsequently leading to loss in explanatory power. A strength of this study, however, is that we were able to compare self-reported tobacco use with the presence/absence of cotinine, which gave us confidence that participants were not under-reporting their use.

Conclusion

This biocultural investigation of Jenu Kuruba women’s smokeless tobacco and areca nut use found that self-medication is the most likely explanatory model for kaddipudi and combined tobacco and paan. These findings suggest that women might rely on these substances to enhance cognitive performance under heightened stress.