## INTRODUCTION

Tobacco use is one of the leading causes of preventable disease, disability, and death1. While tobacco use is declining in high-income countries, it is increasing in certain low- and middle-income countries (LMICs)1,2. In 2017, the prevalence of tobacco use and tobacco cigarette smoking in Zambia was 15.8% and 12.3%, respectively3, and the most recent Demographic Health Surveys in Zambia suggest that the prevalence of tobacco use has been increasing especially among males, from 15% in 2000–2001 to 19% in 20184-7. In 2008, Zambia became a party to the World Health Organization Framework Convention on Tobacco Control (WHO FCTC) which provides evidence-based guidelines on how countries can implement and manage tobacco control measures8-10. The WHO FCTC outlines six proven measures toward tobacco demand reduction, including warning about the dangers of tobacco11. Article 11 of the WHO FCTC requires the adoption of health warnings on tobacco packages with specific recommendations that the warnings be large (i.e. cover at least 50% and no less than 30% of the principal display areas), clear, and legible, and include both pictures and words to enhance their effectiveness8.

Health warning labels on cigarette packages serve as one of the primary means of communicating the harmful effects of cigarette smoking12,13. Awareness of the health harms of cigarette smoking has been associated with increased thinking about quitting and quit attempts among people who smoke13. Some countries in the African region such as Uganda and Mauritius are addressing the tobacco epidemic by strengthening implementation of the WHO FCTC; in particular, introducing pictorial health warnings covering more than 65% of tobacco product packaging14,15. Several studies that assessed the impact of health warnings concluded that pictorial health warnings were more effective than single-text warnings in achieving increased awareness and knowledge of tobacco harms, encouraging thinking about quitting among people who currently smoke, and preventing smoking initiation among people who do not smoke13,16-22.

Although Zambia became a Party to the WHO FCTC in 2008, Zambia’s policy on health warnings does not yet comply with WHO FCTC recommendations: current legislation fails to specify any minimum size of health warnings with respect to the cigarette package, and no pictures are required2,23. A 2014 International Tobacco Control (ITC) country-wide survey found that only 69.0% of cigarette packages in Zambia had a visible text health warning in English, with the rest being either not visible or in another language23. In the 2017 WHO STEPwise approach to Surveillance (STEPS) survey, two-thirds (66.9%) of adults who currently smoked tobacco who saw text-only health warning labels on cigarette packages in Zambia thought about quitting smoking, and the survey report did not delineate the factors that influenced such thinking about quitting3. Prior studies from high-income countries have suggested that the responses to health warnings vary based on many factors, including socioeconomic or sociodemographic factors such as sex, age, education level, and income levels24,25; however, there is a paucity of literature from LMICs in general26, and Zambia in particular. As has been stated in a number of studies25-27, there is a clear need for local data to support policy and in-country tobacco control strategies, especially in developing countries. The purpose of this study was, therefore, two-fold: 1) to examine sociodemographic characteristics associated with awareness of health warnings on cigarette packages among Zambian adults who currently smoke; and 2) to examine sociodemographic characteristics associated with thinking about quitting smoking because of the health warnings among adults who currently smoke.

## METHODS

This study was a secondary analysis using data from the 2017 Zambia WHO STEPS survey. The aim of the survey was to establish a non-communicable disease (NCD) surveillance platform that collects baseline indicators on determinants of NCDs and their risk factors for policy and planning purposes. The survey integrated modified questions from the Tobacco Questions for Surveys (TQS), a subset of key questions from the Global Adult Tobacco Survey (GATS)28,29. The 2017 Zambia STEPS survey was the first nationally representative cross-sectional survey that collected comprehensive information on risk factors for NCDs, mental and oral health in adults aged 18–69 years in Zambia. A multi-stage cluster sampling technique was used to select a nationally representative sample of 5791 adults in Zambia, of which 4302 completed the survey, resulting in an overall response rate of 74.3%3. More details on the sampling methods of the 2017 Zambia STEPS survey can be publicly accessed3. The University of Zambia Biomedical Research Ethics Committee (UNZABREC) approved the STEPS survey, and informed consent was obtained from participants3. Because deidentified secondary data were used, this study was considered exempt from CDC institutional review board approval.

### Measures and variables

The study population for this analysis included adults aged 18–69 years in Zambia who currently smoked tobacco. Adults who currently smoked tobacco were defined as those who answered ‘Yes’ to the question: ‘Do you currently smoke any tobacco products, such as cigarettes, shisha, cigars or pipes?’. Two measures identified as key indicators in promoting thinking about quitting smoking16,18,19 were assessed in this study: 1) awareness of health warnings, and 2) thinking about quitting because of health warnings on cigarette packages. Health warning label awareness among adults who currently smoked tobacco was measured by the question: ‘During the past 30 days, did you notice any health warnings on cigarette packages?’. The responses were ‘Yes’, ‘No’, ‘Did not see any cigarette packages’, or ‘Don't know’. Responses other than ‘Yes’ were considered unaware of health warnings; these respondents were not questioned further regarding thinking about quitting smoking because of health warning labels as they were assumed to have no such intention per GATS/TQS guidelines28,29.

Sociodemographic variables included four age groups (18–29, 30–44, 45–59, 60–69 years), sex (male, female), ethnicity (Bemba, Chewa, Tonga, other), education level (no formal schooling, less than primary school, primary school completed, junior secondary school completed, secondary higher school completed or higher), marital status (currently married, not married), and annual household income (≤10000, >10000 and ≤20000, >20000 ZMW).

### Statistical analysis

Descriptive analyses, including percentages and 95% confidence intervals (CIs) of awareness of health warnings and thinking about quitting were computed by sociodemographic characteristics. Chisquared tests were performed to test for significant differences in the distribution of demographic characteristics between those who were and were not aware of health warnings and those who did and did not intend to quit because of the health warnings; two-sided p<0.05 was considered statistically significant. Logistic regression was performed to calculate crude (PR) and adjusted prevalence ratios (APR) for the association of sociodemographic characteristics with awareness of health warnings and thinking about quitting. Adjusted models included all sociodemographic characteristics of interest except for income, which was not included in the model due to a high proportion (approximately 24%) of missing data. Results were suppressed if the unweighted denominator was <2528,29. Analyses were conducted using SAS-callable SUDAAN version 11.0.

## RESULTS

### Characteristics of the sample

Overall, this study included a total of 475 adults who reported they currently smoked tobacco, which represents 916182 adults when survey weights were applied or 12.3% of adults in Zambia. Approximately 59.2% of tobacco smokers (unweighted n=240) reported awareness of cigarette health warning labels. Among all tobacco smokers, 37.4% (unweighted n=162) reported thinking about quitting because of health warning labels; among those with awareness of the labels, 66.9% of respondents reported thinking about quitting because of the labels.

### Distributions of sociodemographic characteristics by awareness of health warnings and thinking about quitting

Among adults who currently smoked, significant differences in the distribution of sex, ethnicity, education level, and income were observed between those who were aware of health warning labels and those who were not (Table 1). Almost 95% of adult current smokers (94.6%; 95% CI: 91.1–96.7) who were aware of warning labels were male compared to 88.1% (95% CI: 83.2–91.7) among those who were not aware. More than one-third (35.6%; 95% CI: 26.6–45.9) of those who were aware were of Bemba ethnicity, 11.1% (95% CI: 7.07–17.0) of Chewa ethnicity, and almost half (46.8%; 95% CI: 37.3–56.5) belonged to other ethnic groups while among those not aware 32.0% (95% CI: 24.2–40.9), 4.3% (95% CI: 1.9–9.3), and 54.9% (95% CI: 45.9–63.5) were of Bemba, Chewa, and other ethnicities, respectively. Most of those who were aware of health warnings had completed primary school (27.6%; 95% CI: 21.1–35.2) or junior secondary school (28.1%; 95% CI: 20.6–37.2), and secondary or higher (24.6%; 95% CI: 18.3– 32.2) while, among those who were not aware, less than 40% had completed primary school (19.4%; 95% CI: 14.1–26.1), or junior secondary school (10.2%; 95% CI: 6.49–15.8), and secondary or higher (9.7%; 95% CI: 4.94–18.3). Among those who were aware, 38.0% (95% CI: 30.6–46.0) had an annual income of ≤10000 ZMW and 36.4% (95% CI: 27.8–45.9) had an annual income of >20000 ZMW, while among those who were not aware, nearly half (48.0%; 95% CI: 40.1–56.0) and 18.6% (95% CI: 12.7–26.3) had an annual income of ≤10000 ZMW and >20000 ZMW, respectively. About 14.6% (95% CI: 9.8–21.3) and almost 3 in 10 (28.6%; 95% CI: 22.3–35.9) of those who were aware and not aware, respectively, did not have information on income. The distributions of age and marital status did not differ significantly between those who were aware and were not aware of health warnings.

Among all adults who currently smoked, a higher percentage of adults who currently smoked who self-reported thinking about quitting because of health warnings on cigarette packages (from here on referred to as ‘thinking about quitting’) were male (96.2%; 95% CI: 92.6–98.1), 29.1% (95% CI: 21.1– 38.5) had completed primary school and 29.4% (95% CI: 19.8–41.2) completed junior secondary school, with similar proportions of those who had income levels ≤10000 ZMW (37.8%; 95% CI: 28.9–47.6), and >20000 ZMW (37.9%; 95% CI: 27.1–50.2). Among those who were not thinking about quitting, 89% (95% CI: 84.6–92.3) were male, half completed less than primary school (37.5%; 95% CI: 31.5–44.0) or had no formal schooling (12.1%; 95% CI: 8.25– 17.4), and 45% (95% CI: 38.4–52.0) had an income level ≤10000 ZMW (Table 1).

### Prevalence and associated sociodemographic characteristics of awareness of health warnings and thinking about quitting

The prevalence of awareness of health warnings among adults who currently smoked was relatively high among the 30–44 years age group (61.5%; 95% CI: 52.8–69.5), men (57.7%; 95% CI: 51.0–64.2), those of Chewa ethnicity (76.5%; 95% CI: 59.3– 87.9), those who had completed junior secondary (77.8%; 95% CI: 65.5–86.6) or secondary high school (76.2%; 95% CI: 58.7–87.8), and those who were not married (64.1%; 95% CI: 52.9–74.0) (Table 2). In the unadjusted model, those of Chewa ethnicity were 1.31 times (95% CI for PR: 1.01–1.68) more likely to be aware of health warnings compared with those of Bemba ethnicity. Those who had no formal schooling were 0.52 times (95% CI: 0.26–1.03) less likely to be aware than those with less than primary school. Those who had completed primary school (PR=1.89; 95% CI: 1.38–2.58), junior secondary school (PR=2.17; 95% CI: 1.61–2.93) or secondary high school (PR=2.13; 95% CI: 1.52–2.98) were more likely to be aware than those with less than primary schooling. Those who were not married were 1.23 times (95% CI for PR: 1.00–1.52) more likely to be aware of health warnings compared to those who were married. Women (PR=0.63; 95% CI: 0.44–0.92) were less likely to be aware of health warnings compared to men. In the adjusted model, being in the 30–44 age group (APR=1.26; 95% CI: 1.02–1.54), Chewa ethnicity (APR=1.43; 95% CI: 1.17–1.74) and education: without formal schooling (APR=0.41; 95% CI: 0.20–0.82), primary school (APR=1.77; 95% CI: 1.31–2.93), junior secondary school (APR=2.10; 95% CI: 1.57–2.79) and secondary high school (APR=1.98; 95% CI: 1.41–2.78) remained significantly more or less likely to be aware of health warnings. Sex and marital status were no longer significant after adjusting for other sociodemographic variables.

##### Table 2

Prevalence ratios (PR) for the association of selected sociodemographic characteristics with awareness of health warnings and thinking about quitting among adults who currently smoked tobacco aged 18–69 years in Zambia, STEPS survey 2017

CharacteristicsAware of health warnings on cigarette packagesThinking about quitting because of health warnings on cigarette packages among all adults who currently smokedThinking about quitting because of health warnings on cigarette packages among those aware of health warning labels
%95% CIPR95% CIAPR95% CI%95% CIPR95% CIAPR95% CI%95% CIPR95% CIAPR95% CI
Age (years)
18-29 (Ref.)57.045.1-68.11140.128.6-52.91170.455.0-82.311
30-4461.552.8-69.51.080.85-1.361.261.02-1.54a41.933.4-51.01.040.72-1.521.110.79-1.5668.257.2-77.40.970.75-1.240.860.66-1.11
45-5946.636.1-57.40.820.60-1.111.060.81-1.3826.018.3-35.60.650.41-1.020.780.50-1.2055.839.5-71.00.790.56-1.130.720.50-1.04
60-6945.527.4-64.90.800.49-1.291.190.85-1.6529.014.1-50.40.720.36-1.451.030.62-1.72~~~
Sex
Male (Ref.)57.751.0-64.21139.332.8-46.21168.159.5-75.611
Female36.624.9-50.10.630.44-0.920.780.60-1.0117.18.70-30.80.430.23-0.84b0.510.27-0.96b~~~
Ethnicity
Bemba (Ref.)58.648.1-68.41139.728.8-51.81167.851.7-80.511
Chewa76.559.3-87.91.311.01-1.68b1.431.17-1.74b58.441.7-73.51.470.98-2.221.551.03-2.35b76.451.8-90.71.130.80-1.591.110.77-1.60
Tonga46.628.0-66.20.790.50-1.270.790.53-1.1830.915.3-52.60.780.39-1.550.830.44-1.56~~~
Other52.042.7-61.10.890.69-1.1410.81-1.2233.124.8-42.70.830.56-1.250.960.68-1.3763.851.6-74.40.940.71-1.250.970.74-1.27
Education level
No formal schooling14.87.4-27.40.520.26-1.030.410.20-0.82b11.25.0-23.10.590.25-1.410.560.23-1.37~~~
Less than primary school (Ref.)32.924.4-42.61118.912.6-27.51157.639.0-74.311
Primary school completed64.353.1-74.21.891.38-2.58b1.771.31-2.39b45.334.0-57.22.391.49-3.84b2.271.43-3.62b70.554.3-82.81.220.83-1.791.270.87-1.85
Junior secondary school completed77.865.5-86.62.171.61-2.93b2.101.57-2.79b54.339.6-68.42.871.78-4.64b2.761.72-4.41b69.954.3-82.01.210.84-1.761.210.84-1.74
Secondary high school completed or higher76.258.7-87.82.131.52-2.98b1.981.41-2.78b50.536.3-64.72.671.66-4.30b2.431.49-3.98b66.349.9-79.51.150.78-1.691.180.80-1.73
Marital status
Currently married (Ref.)52.044.9-59.11136.329.6-43.51169.759.9-78.011
Not married64.152.9-74.01.231.00-1.52b1.170.96-1.4339.928.7-52.21.100.78-1.550.980.70-1.3962.247.2-75.20.890.69-1.160.810.60-1.08

PR: prevalence ratio. APR: adjusted prevalence ratio.

a Calculated in logistic regression models adjusted for all covariates in table.

b Statistically significant, p<0.05. ~ Insufficient sample size; estimates where the unweighted denominator was <25 have been suppressed.

Among all adults who currently smoked, thinking about quitting had similar findings to those of awareness of health warnings (Table 2). The prevalence was relatively high among those aged 18–29 and 30–44 years (40.1%; 95% CI: 28.6– 52.9, and 41.9%; 95% CI: 33.4–51.0, respectively), men (39.3%; 95% CI: 32.8–46.2), those of Chewa ethnicity (58.4%; 95% CI: 41.7–73.5), those who had completed junior secondary and secondary high school (54.3%; 95% CI: 39.6–68.4, and 50.5%; 95% CI: 36.3–64.7, respectively) and those who were not married (39.9%; 95% CI: 28.7–52.2). In the unadjusted model, those with any formal education (PR range: 2.39; 95% CI: 1.49–3.84 [primary school completed] to 2.67; 95% CI: 1.66–4.30 [secondary high school completed]) were more likely to report thinking about quitting than those with no formal schooling (PR=0.59; 95 CI: 0.25–1.41). Women (PR=0.43; 95% CI: 0.23–0.84) were less likely to report thinking about quitting than men. In the adjusted model, sex and education level remained significant; additionally, those of Chewa ethnicity (APR=1.55; 95% CI: 1.03–2.35) were more likely to report thinking about quitting than those of Bemba ethnicity.

When restricted to current smokers who were aware of health warning labels, the prevalence of thinking about quitting was 70.4% (95% CI: 55.0–82.3) among those aged 18–29 years, 76.4% (95% CI: 51.8–90.7) among those of Chewa ethnicity, 70.5% (95% CI: 54.3–82.8) and 69.9% (95% CI: 54.3–82.0), respectively, among those who had completed primary and junior secondary schools, and 69.7% (95% CI: 59.9–78.0) among those who were married (Table 2). There were no sociodemographic characteristics that were significantly associated with thinking about quitting in unadjusted or adjusted models.

## DISCUSSION

Our analyses found that among adults who currently smoke, sociodemographic differences exist in awareness of health warning labels on cigarette packages, particularly in sex, ethnicity, and education level, but no sociodemographic differences were found in thinking about quitting among tobacco smokers aware of the health warnings on cigarette packages. Generally, those who had completed primary school or above were more likely to be aware of health warnings, with the magnitude of the association increasing with higher education level. Findings from this study may highlight the need for clear and easy to understand pictorial health warnings on cigarette packages that could be understood by all adults of any level of education, sex, or ethnicity in Zambia.

We found that certain sociodemographic characteristics were significantly associated with awareness of health warnings and thinking about quitting among all adults who currently smoked. Even though the proportion of female current tobacco smokers in Zambia was relatively low, our finding that women were less likely to notice health warnings and think about quitting than men is like that reported in some Asian countries such as Vietnam, Malaysia and Thailand20,21,26,30-32. Women, especially those of low socioeconomic status, may be generally less responsive to tobacco control interventions than men not only because fewer women attain higher levels of education compared to men in Zambia7, but there has also been limited integration of gender considerations into multiple elements of tobacco control policies, including messages on the harms of tobacco use33-35. This may, in part, explain our findings that suggest that women may be less likely to have awareness of health warning labels or thinking about quitting. We also found that those of Chewa ethnicity were more likely to be aware of health warnings and thinking about quitting than were those of Bemba ethnicity. Most adults of Chewa ethnicity reside in the Eastern Province, where smoking prevalence is one of the highest in Zambia32 and tobacco farming is common36; however, there is very little research on differences in tobacco use, awareness of health warnings, and thinking about quitting across ethnic groups, particularly in Zambia. While more research is required to understand these findings, it is important to ensure tobacco control interventions and campaigns are implemented equitably across sex and ethnic groups.

Previous studies have shown that smoking prevalence is highest among those with lower education level, decreasing as the level of education increases7,25,32. Not only are those with less education more likely to smoke, but they are also less likely to be aware of health warnings7,25,32,37, a finding like those observed in our study. We observed low levels of awareness of health warnings among those with no formal schooling (14.8%); those who had some levels of education were 1.89 to 2.13 times as likely to be aware of health warnings, as opposed to those who had no formal schooling. A study conducted in 2014 showed that 40.0% of Zambian adults who currently smoked could not read the health warning at all, and another 14.0% read it with difficulty23. Several studies, including those from the United States and Europe, have concluded that people with less education were more responsive to graphic health warnings or perceived them as more effective than text-only warnings16,19,21,25,38. Furthermore, Zambian adults who currently smoked were reported to be less likely to be aware of several important smoking-related health effects such as lung cancer, heart disease, and stroke compared to smokers in other countries23.

We found similar patterns in the associations between sociodemographic characteristics and thinking about quitting as we did for awareness of health warnings; however, in our model that included only those who were already aware of health warnings, these associations were no longer observed. This could be due, in part, to the small sample size observed in our study. Alternatively, it could possibly demonstrate that, while there are disparities in awareness across sociodemographic groups, warning labels are similarly effective at causing adults who smoke to think about quitting once they are aware of warning labels25,39-42. While results varied by sociodemographic group, over half of those who were aware of health warnings also indicated thinking about quitting, highlighting the importance of increasing awareness. This may present an opportunity for using health warnings as a means of educating smokers about the harms of smoking with messages such as ‘smoking causes lung cancer’ or ‘smoking can lead to gangrene’ or implementing pictorial health warnings15,23,43. Article 11 of the WHO Framework Convention on Tobacco Control (FCTC) requires the adoption of health warnings on tobacco packages and recommends, in part, that the warnings are large, clear, visible, legible and include pictures to enhance their effectiveness8.

Several studies from developed countries such as the USA, Japan, United Kingdom, Canada and Australia have reported that large text and pictorial health warnings are associated with an increased likelihood of quitting as opposed to single text-only warnings, regardless of the age or education of those who smoke13,16,18,19,21,38,43-45. Another study showed that while effectiveness of health warnings on cigarette packages varied, pictorial warnings with personal testimonials were the most effective across more than 100 countries41. Pictorial warnings have several advantages relative to text-only warnings, including ability to attract and hold attention, elicit more negative smoking attitudes, and increase thinking about quitting37,46. Some Sub-Saharan African countries such as Uganda and Mauritius have adopted pictorial health warnings14. For example, Mauritius adopted large pictorial warning labels covering over 60–70% of cigarette packages, which resulted in increased awareness of health warnings from 58.0% to 83.4% over the course of two years post-implementation15. Zambia may have a similar opportunity to increase the awareness of health warnings and thinking about quitting among those who smoke and deter initiation among those who do not smoke. Implementing the WHO-FCTC recommendation of large and pictorial health warnings may mitigate the sociodemographic differences related to awareness and thinking about quitting.

### Limitations

This study has several limitations. First, this was a secondary analysis of a cross-sectional survey so causality cannot be determined between sociodemographic factors and awareness of health warnings and thinking about quitting. Second, selfreport responses could be subject to recall or social desirability biases, leading to underreporting of smoking or overreporting of thinking about quitting (e.g. some participants may intend to quit smoking for reasons other than noticing health warnings and thus would not be captured), although several studies have reported a high correlation between self-reports and biomarkers of smoking47-49. Third, we were not able to measure important variables with STEPS data such as language, region, or knowledge of health harms of tobacco, therefore we were unable to account for these factors in our analyses. For example, although English is the official language in Zambia, it is not the first language for most Zambians. Evidence suggests that about 69% of cigarette packages in Zambia carry a single text-only warning in English, 14% are in another foreign language, and 18% have no visible warnings23, underscoring the significant role language could play in responsiveness to health warnings. Additionally, we opted not to include income in our models due to a high proportion of missing data; thus, were unable to assess any associations between household income with awareness of health warnings and/or thinking about quitting. Despite these limitations, our study is unique in that, to our knowledge, this is the first time a study describing sociodemographic characteristics associated with awareness of health warnings and thinking about quitting has been conducted in Zambia. Another strength of this study lies in the ability to generalize its findings as it was based on the nationally representative 2017 Zambia STEPs survey.

## CONCLUSIONS

Sociodemographic characteristics such as sex, ethnicity, and education level were significantly associated with awareness of cigarette health warnings. Among cigarette smokers aware of health warnings, no sociodemographic differences in thinking about quitting were found. Findings from this study indicated that current tobacco smokers who had primary school or above education were more likely to be aware of health warnings, which suggests the need for clear, easy to understand, pictorial health warnings on cigarette packages in Zambia. Implementation of stronger policies requiring health warnings with these characteristics may increase awareness of health warnings and thinking about quitting smoking, particularly among those with lower levels of education. Additionally, while a small percentage of women currently smoke tobacco in Zambia, there may be a need for targeted campaigns among women as they were less likely to be aware of health warnings and thinking about quitting. Equitable implementation of tobacco control campaigns may need to target regions or ethnicities (e.g. Chewa) in which smoking prevalence is higher compared to other regions of the country.