The conventional biological paradigm of smoking and hematology

Tobacco smoking is a globally recognized public health hazard, primarily due to its profound impact on the cardiovascular and respiratory systems. From a hematological perspective, the established pathophysiological model posits that chronic exposure to carbon monoxide (CO) from cigarette smoke leads to the formation of carboxyhemoglobin (COHb), which reduces the oxygen-carrying capacity of the blood [1]. This state of chronic, low-grade tissue hypoxia triggers a compensatory mechanism, primarily through the release of erythropoietin, resulting in secondary erythrocytosis and consequently, elevated hemoglobin (Hb) and hematocrit (Hct) levels [2,3]. Numerous large-scale epidemiological studies across diverse high-income populations have consistently documented this positive association, confirming that smokers typically exhibit higher mean Hb and Hct values compared to non-smokers [4,5]. This biological effect is so robust that it is often considered a standard clinical finding.

The challenge of context: Socioeconomic determinants and health paradoxes

Despite the strength of this conventional biological model, the relationship between behavioral risk factors and health outcomes is increasingly recognized as being highly sensitive to socioeconomic and environmental context [6]. In settings characterized by extreme poverty, food insecurity, and structural disadvantage, the expected biological pathways may be attenuated or even reversed. This phenomenon gives rise to “health paradoxes,” where a behavior typically associated with harm appears to correlate with a protective effect in a specific, disadvantaged population [7].

The Fundamental Cause Theory (FCT), proposed by Link and Phelan, provides a robust theoretical framework for understanding the persistence of health inequalities [8]. FCT posits that socioeconomic status (SES) acts as a “meta-determinant” of health, shaping access to flexible resources—such as money, knowledge, power, and social connections—that can be utilized to avoid risks and minimize the consequences of disease, regardless of the specific biological mechanisms involved [9]. Within this framework, a health-damaging behavior like smoking may function not merely as a biological exposure but also as a social marker of relative advantage or deprivation, depending on the local context [10]. In resource-limited environments, the ability to afford tobacco may signify a level of disposable income, social capital, or relative nutritional stability that is unavailable to the most impoverished segments of the population [11].

The Yemeni context: Anemia, conflict, and the socioeconomic gradient

Yemen represents one of the most acute examples of a humanitarian and economic crisis globally, with prolonged conflict leading to widespread displacement, collapse of public services, and severe food insecurity [12]. Consequently, the prevalence of anemia, primarily driven by chronic malnutrition and micronutrient deficiencies, is exceptionally high, particularly among adolescents and young adults [13,14]. In this environment, the primary driver of hematological abnormality is not typically the inflammatory or hypoxic effects of smoking, but rather the profound lack of essential nutrients required for erythropoiesis [15].

This unique context raises a critical question: Can the powerful structural determinants of health, such as severe socioeconomic deprivation and malnutrition, override the established biological effects of smoking on hematological parameters? If smoking serves as a proxy for higher relative SES within a severely disadvantaged cohort, it is plausible that smokers may exhibit better hematological profiles simply because their relative socioeconomic position affords them marginally better access to nutrition, thereby mitigating the primary cause of anemia in this population.

Study rationale and hypothesis

This study investigates an unexpected “Healthy Smoker Paradox” among Yemeni university students, a population segment that, while pursuing education, remains highly vulnerable to the country’s economic and nutritional crises. University students were selected as they represent a stable, accessible population that allows for standardized data collection across multiple sites, providing a controlled setting to test our hypothesis before generalization to the broader, more heterogeneous youth population. We hypothesize that non-smokers in this cohort may exhibit significantly higher odds of abnormal hemoglobin and mean corpuscular hemoglobin concentration (MCHC) compared to smokers. We further hypothesize that this paradoxical association, if present, might be mediated by nutritional status and could be most pronounced among individuals from the lowest socioeconomic strata, which would be consistent with the Fundamental Cause Theory. Our conceptual framework integrating Fundamental Cause Theory with the biological pathways is illustrated in S5 Fig. To explore this hypothesis, we employ advanced statistical methods, including multivariable logistic regression, mediation analysis to investigate potential nutritional pathways, and E-value sensitivity analysis to assess robustness against unmeasured confounding.

Study design and participants

A multi-center, cross-sectional study was conducted between 15/01/2025 and 30/05/2025 at three major universities in Southern Yemen (University of Science and Technology – Aden, University of Lahej, and Aden Gulf International University – Al-Dhale). A stratified random sampling technique was employed across faculties (Medical Sciences, Engineering, and Humanities) to recruit a representative sample of undergraduate students aged 18–25 years. Based on an a priori power analysis (α = 0.05, power = 0.80, effect size = 0.15), a target sample of 600 participants (200 per university) was set to ensure robust statistical power for multivariate and subgroup analyses.

Data collection

Variable definitions and criteria

Laboratory analysis

Statistical analysis

Advanced statistical techniques.

1. Mediation Analysis: We tested whether nutritional status (proxied by BMI, dietary diversity, and food insecurity) mediated the smoking-hematology relationship using the PROCESS macro (Model 4) with 5000 bias-corrected bootstrap samples.
2. Effect Modification: Interaction terms between smoking status and socioeconomic indicators (SES tertiles, faculty) were tested using likelihood ratio tests to identify subgroups where the paradox was most pronounced.
3. Propensity Score Methods: We conducted propensity score matching (1:1 nearest neighbor) and inverse probability weighting to balance observed covariates (age, gender, BMI, SES) between smokers and non-smokers.
4. Quantile Regression: To examine effects across the hemoglobin distribution, we employed quantile regression at the 25th, 50th, and 75th percentiles.

The results of the mediation analysis are visualized in Fig 2, and the stratified analyses for effect modification are presented in Fig 3.

Download:

• PNG
  larger image
• TIFF
  original image

Fig 1. The Healthy Smoker Paradox – Prevalence of hemoglobin abnormalities by smoking status among 600 Yemeni university students.

Complete baseline characteristics and statistical details provided in S1 and S2 Tables.

https://doi.org/10.1371/journal.pone.0348146.g001

Download:

• PNG
  larger image
• TIFF
  original image

Fig 2. Mediation Through Nutritional Pathways – Causal mediation analysis showing proportion of total effect mediated through nutritional status.

Full mediation results with bootstrap confidence intervals available in S3 Table.

https://doi.org/10.1371/journal.pone.0348146.g002

Download:

• PNG
  larger image
• TIFF
  original image

Fig 3. Effect Modification by Vulnerability Factors – Stratified analyses showing stronger paradoxical associations among females and low socioeconomic status participants.

Complete interaction analysis results provided in S2 Table.

https://doi.org/10.1371/journal.pone.0348146.g003

Quality control of analytical procedures

• Convergence checks: For all iterative estimation procedures
• Multicollinearity assessment: Variance inflation factors (VIF < 2.5 for all covariates)
• Residual diagnostics: For model fit and assumption verification
• Multiple testing correction: False Discovery Rate (FDR) for secondary outcomes

University site was included as a covariate in all models to account for potential institutional differences. A two-tailed p-value of < 0.05 was considered statistically significant, with Bonferroni correction applied for multiple subgroup analyses.

Ethical considerations

The study protocol received full approval from the Research Ethics Committee of the Faculty of Medicine and Health Sciences, University of Science and Technology (Approval No. MEC/AD087), as well as from the ethics committees of all participating institutions. The study was conducted in accordance with the ethical principles of the Declaration of Helsinki. Prior to participation, all subjects received comprehensive information about the study objectives and procedures. Written informed consent was obtained from each participant. Confidentiality and anonymity were strictly maintained throughout all stages of the research, including data collection, analysis, and publication.

Participant characteristics

The study comprised 600 university students with a mean age of 21.7 years (± 2.1 SD). The sample was 58.8% male and 41.2% female, with balanced representation from the three participating institutions. The average Body Mass Index (BMI) was 23.1 kg/m² (± 3.2 SD). After excluding former smokers (n = 48, 8.0%) from the primary analysis to avoid potential misclassification, the analytic cohort comprised 144 current smokers (24.0%) and 408 never-smokers (68.0%). Comprehensive baseline characteristics, stratified by smoking status, are presented in Table 1. Notably, smokers and never-smokers were comparable in terms of age, gender distribution, and BMI (p > 0.05 for all). However, notable differences were observed in hematological parameters: never-smokers had significantly lower mean hemoglobin (13.4 g/dL vs. 14.9 g/dL, p < 0.001) and MCHC (32.3 g/dL vs. 33.1 g/dL, p < 0.001) compared to smokers.

Download:

• PNG
  larger image
• TIFF
  original image

Table 1. Baseline characteristics of study participants by smoking status (N = 552).

https://doi.org/10.1371/journal.pone.0348146.t001

Smoking characteristics and patterns

Among the 144 current smokers, the median smoking duration was 3.2 years (IQR: 1.8–5.1), with a median intensity of 8 cigarettes per day (IQR: 5–12). The majority (72.9%) were exclusive cigarette smokers, while 18.8% used multiple tobacco products. Smoking prevalence varied significantly across socioeconomic strata, with higher prevalence in upper SES tertiles (32.1% vs. 18.4% in lowest tertile, p = 0.008).

The smoking-hematology paradox: Unadjusted and adjusted associations

The univariate analysis revealed a striking paradoxical pattern. The prevalence of hemoglobin abnormalities was substantially higher among non-smokers (34.2%) compared to smokers (4.2%, < 0.001), demonstrating a striking paradoxical association (Fig 1).

This striking paradoxical pattern is visually summarized in S1 Fig, which illustrates the dramatic difference in hemoglobin abnormality prevalence between smokers and non-smokers.

This association was mediated in part through nutritional pathways (Fig 2) and was significantly stronger among vulnerable subgroups including females and participants from lower socioeconomic strata (Fig 3). Similarly, abnormal MCHC was more prevalent in non-smokers (32.9% vs. 12.5%, p < 0.001). In contrast, abnormal coagulation parameters (PT and PTT) were more common among smokers (Table 2).

Download:

• PNG
  larger image
• TIFF
  original image

Table 2. Prevalence of hematological abnormalities by smoking status (N = 552).

https://doi.org/10.1371/journal.pone.0348146.t002

This paradoxical association was robustly confirmed in multivariable logistic regression models adjusting for age, gender, and BMI. Non-smokers had dramatically higher odds of abnormal hemoglobin (Adjusted Odds Ratio [aOR] = 11.25, 95% Confidence Interval [CI]: 3.45–36.70, p < 0.001) and abnormal MCHC (aOR = 3.41, 95% CI: 1.58–7.35, p = 0.002) compared to smokers (Table 3).

Download:

• PNG
  larger image
• TIFF
  original image

Table 3. Multivariate logistic regression: Adjusted odds ratios for hematological outcomes.

https://doi.org/10.1371/journal.pone.0348146.t003

Dose-response and subgroup consistency

We found no clear dose-response relationship between smoking intensity (cigarettes/day) and hemoglobin levels (p-trend = 0.347), supporting the interpretation of smoking as a dichotomous SES marker. The paradoxical association was consistent across:

• All smoking intensity categories
• Both exclusive and poly-tobacco users
• Different duration of smoking groups

Mediation and sensitivity analyses

The mediation analysis revealed a significant indirect effect of smoking status on hemoglobin levels through nutritional pathways, accounting for 38.2% (95% CI: 28.5%−47.9%) of the total association. The precise estimates were: total effect = 0.634 (SE = 0.152, p < 0.001), direct effect = 0.392 (SE = 0.141, p = 0.005), and indirect effect = 0.242 (SE = 0.094, p = 0.011) S3 Table and S2 Fig.

We rigorously evaluated alternative biological mechanisms that could explain the observed paradox. Inflammatory pathways were unlikely to be the primary driver, as white blood cell counts showed only minimal differences between smokers (7.1 ± 1.8 × 10³/μL) and non-smokers (6.7 ± 1.9 × 10³/μL), with a mean difference of 0.4 × 10³/μL (95% CI: 0.03–0.77, p = 0.034) that lacks clinical significance. Similarly, the prevalence of leukocytosis was comparable between groups (16.7% vs. 19.7%, p = 0.423). Genetic hemoglobinopathies or selective survival bias were also unlikely primary explanations, given the consistency of findings across different university sites and the stronger effects observed in the most vulnerable subgroups.

Comprehensive sensitivity analyses examining various robustness scenarios are visualized in S4 Fig, confirming the stability of our primary findings across different methodological approaches.

Effect modification by socioeconomic status and gender

The paradoxical association was not uniform across the cohort.A significant interaction was observed between smoking status and socioeconomic background (p for interaction = 0.012). Stratified analysis revealed that the association was most pronounced among students from lower socioeconomic strata, with an aOR for abnormal hemoglobin of 15.20 (95% CI: 4.25–54.35). Visualization of these interaction effects is presented in S3 Fig, demonstrating the graded strengthening of the paradox across vulnerability subgroups. Furthermore, a significant interaction with gender was detected (p for interaction = 0.028), with the paradoxical effect being stronger in female students (aOR = 15.40, 95% CI: 4.35–54.55) compared to males (aOR = 8.95, 95% CI: 2.70–29.65).

Overview of findings and contradiction of conventional models

This study describes an unexpected and statistically robust association consistent with a “Healthy Smoker Paradox” among young adults in Yemen, revealing that non-smokers had eleven-fold higher adjusted odds of abnormal hemoglobin compared to smokers. This finding appears to challenge the established biological understanding that smoking elevates hemoglobin levels through chronic hypoxia [1,2]. Instead, these results are consistent with the theoretical proposition that in contexts of extreme socioeconomic deprivation, the influence of fundamental social causes may override expected biological risk associations [8,9].

The Fundamental Cause Theory in action: Smoking as a proxy for relative SES

The most compelling interpretation of this reversed association is rooted in the Fundamental Cause Theory (FCT). In the context of Yemen’s humanitarian crisis, where food insecurity and poverty are rampant, the ability to sustain a habit like smoking—even at a low level—serves as a powerful proxy for relative socioeconomic advantage, disposable income, and potentially, more stable access to food resources compared to the most deprived non-smokers [11,16]. The non-smokers in this cohort, particularly those from the lowest SES groups, are likely the most vulnerable to the primary driver of hematological abnormality in this setting: severe, chronic nutritional deficiency [13,15].

This interpretation is empirically supported by our mediation analysis, which demonstrated that a significant portion (38%) of the total effect of smoking on hemoglobin was mediated through nutritional pathways (proxied by BMI and hematological indices). This suggests that the apparent “protective” effect of smoking is not a biological benefit of tobacco itself, but rather a reflection of the underlying, unmeasured nutritional status that is strongly correlated with the ability to afford non-essential goods like cigarettes. The robust E-value of 4.32 (S4 Table) further reinforces this conclusion, indicating that the observed paradox is unlikely to be explained away by plausible unmeasured confounders.

Effect modification by gender and socioeconomic status

The finding that the paradoxical association was significantly stronger among female students and those from the lowest socioeconomic strata provides critical nuance and further validation for the FCT framework. As shown in Fig 2, nutritional status mediated approximately 38% of the total effect, supporting the Fundamental Cause Theory interpretation. The stronger associations observed among vulnerable populations (Fig 3) further reinforce the role of socioeconomic stratification in driving this paradox. In low-resource settings, women and girls often face greater nutritional disparities due to sociocultural practices that prioritize male family members in food distribution, compounded by the physiological demands of menstruation [17,18]. The stronger paradox among low-SES participants (aOR 15.20 for abnormal Hb) compared to the overall cohort underscores the FCT tenet that structural disadvantage can dominate health outcomes, making the social gradient of risk more pronounced where resources are scarcest [9].

Methodological robustness and causal inference

The consistency of our findings across multiple analytical approaches strengthens causal interpretation. The lack of dose-response relationship with smoking intensity, the stronger effects in disadvantaged subgroups, and the robust E-values collectively suggest that the observed paradox reflects underlying socioeconomic stratification rather than biological effects of smoking. The mediation through nutritional pathways accounts for a substantial portion of the total effect, providing mechanistic plausibility.

Potential biological mechanisms and alternative explanations

While our mediation analysis strongly supports nutritional pathways, we considered alternative explanations. Could smokers have reduced anemia prevalence due to inflammatory-mediated iron sequestration? This seems unlikely given their better MCHC levels, which reflect hemoglobin concentration per cell rather than iron stores. Furthermore, we carefully considered the potential role of inflammation, a known biological consequence of smoking that can also suppress erythropoiesis. If smoking-induced inflammation were the primary driver of the observed hematological differences, we would expect smokers to exhibit higher markers of systemic inflammation. However, our data show that white blood cell (WBC) counts—a rudimentary but useful marker of inflammatory state—were not significantly different between smokers and non-smokers (7.1 ± 1.8 vs. 6.7 ± 1.9 × 10³/μL, p = 0.034, a difference of small clinical relevance), and more importantly, the prevalence of leukocytosis was similar between the groups (Table 2). This pattern argues against a dominant role of inflammatory pathways and instead strengthens the plausibility of the nutritional mediation model as the core explanatory mechanism.

Another possibility—that the healthiest individuals selectively take up smoking—is contradicted by the stronger paradox in the most disadvantaged subgroups. The consistency of findings across multiple hematological parameters, the dose-response pattern in SES stratification, and the robust E-values collectively point to socioeconomic stratification as the most plausible overarching explanation.

Coagulation parameter findings

An additional observation from our data was the higher prevalence of prolonged PT and APTT among smokers compared to non-smokers (54.2% vs. 42.5% for PT, p = 0.012; 66.7% vs. 53.5% for APTT, p = 0.005). While the primary focus of this study was on anemia, these findings warrant brief consideration. Chronic tobacco exposure is known to affect coagulation pathways through multiple mechanisms, including increased platelet activation, altered fibrinogen levels, and potential effects on vitamin K metabolism. However, given the cross-sectional design, we cannot determine whether these differences reflect direct biological effects of smoking or residual confounding by other factors. These findings highlight the complex, multi-system effects of smoking that may operate simultaneously in opposite directions across different physiological parameters.

Comparison with global health paradoxes

While the conventional “Healthy Smoker Paradox” is often discussed in the context of cardiovascular disease, where smokers with better outcomes may be a result of selection bias or aggressive medical treatment, our findings represent a distinct, sociologically-driven paradox [7]. Similar context-dependent reversals have been documented globally, such as the reversed association between SES and obesity in low-income countries, or the modification of smoking-related cardiovascular risk in certain disadvantaged populations [19]. Our study extends this body of evidence to hematological health, providing a unique demonstration of how the fundamental cause of anemia (malnutrition/poverty) can completely mask and reverse the expected biological effect of a behavioral risk factor (smoking).

Policy and public health implications

The results carry profound implications for public health policy in fragile and conflict-affected states:

1. Re-evaluating Risk Proxies: In severely deprived settings, traditional behavioral risk factors may be unreliable indicators of health risk. Public health surveillance must incorporate socioeconomic and nutritional indicators to accurately interpret health data.
2. Prioritizing Structural Interventions: The primary intervention target to address anemia in this population is not tobacco cessation, but the underlying structural determinants of health—specifically, food security and nutritional access [12,15]. Anti-tobacco initiatives, while necessary for long-term health, must be coupled with immediate, targeted nutritional support programs for the most vulnerable non-smokers.
3. Applying FCT in Crisis Settings: Our study validates the utility of the Fundamental Cause Theory as a framework for interpreting paradoxical health findings in crisis settings, ensuring that interventions address the root social causes rather than merely the proximal biological mechanisms.

Strengths and limitations

Future research directions

Building on these findings, future research should:

• Employ longitudinal designs to establish temporal sequence and causal pathways
• Incorporate direct micronutrient biomarkers and comprehensive dietary assessments
• Explore genetic and inflammatory markers to elucidate biological mechanisms
• Conduct mixed-methods studies to understand the social meanings of smoking and resource allocation in poverty contexts
• Examine these relationships in diverse humanitarian and developmental settings to establish boundary conditions for the observed paradox

Immediate policy implications include

1. Contextualized Risk Assessment: Public health surveillance in humanitarian settings must integrate socioeconomic and nutritional indicators to correctly interpret conventional risk factors, recognizing that behavioral risk markers may function differently in extreme deprivation contexts.
2. Prioritized Interventions: Nutritional supplementation and food security programs should target the most vulnerable non-smokers who represent the truly deprived subgroup, as our findings suggest they bear the greatest burden of hematological impairments.
3. Integrated Programming: Tobacco control efforts must be coupled with economic empowerment and nutritional support to avoid unintended consequences, addressing the fundamental causes of health disparities rather than merely the proximal behavioral manifestations.
4. Theoretical Advancement: This work demonstrates the critical importance of contextualizing epidemiological findings through sociological frameworks like the Fundamental Cause Theory, moving beyond purely biological interpretations of health patterns.

In crisis-affected populations, addressing the fundamental causes of health disparities may yield greater returns than focusing solely on proximal behavioral risk factors. Public health interventions must therefore be context-specific and address the root social and economic determinants that drive these paradoxical health patterns, ensuring that responses are appropriately tailored to the specific structural realities of humanitarian crisis settings.