Trends and correlates of hardcore smoking in India: findings from the Global Adult Tobacco Surveys 1 & 2

Background: Data on the prevalence of hardcore smoking (HCS) among different socioeconomic status (SES) groups in low- and middle-income countries are limited. We looked at the prevalence and pattern of HCS in India with the following objectives: 1) to analyse the association between SES and HCS, 2) to find trends in HCS in different SES groups and 3) to find state-wide variations in hardcore smoking. Methods: Data of individuals aged ≥25 years from the Global Adult Tobacco Survey (GATS) India 2009-10 (N= 9223) and 2016-17 (N= 7647) were used for this study. If an individual met all the following criteria: (1) current smoker, (2) smokes 10 or more cigarettes/day, (3) smokes first puff within 30 minutes after waking up, (4) no quit attempt in the last 12 months, and (5) no intention to quit at all or in the next 12 months, s/he was identified as a hardcore smoker. Multiple regression analysis was done to find the factors associated with HCS. Results: Prevalence of HCS decreased from 3% in GATS 1 to 2.1% in GATS 2: males from 5.6% to 3.9% and females from 0.3% to 0.2%. Compared to the richest group the poorest, poor and those who belonged to the middle-income group were more likely to report HCS in GATS 1 and 2. However, only in the poorest SES group, there was an increase in the proportion of hardcore smokers in GATS 2 compared to GATS 1. Other factors that were significantly associated with HCS in both surveys were male gender, working adults, those with lower education, and households without any rules for smoking inside the home. Conclusions: Tobacco control and cessation efforts need to focus on individuals of poor SES groups, particularly in the high prevalence Indian states.

Any reports and responses or comments on the article can be found at the end of the article.

Introduction
There were more than one billion smokers globally in the year 2019 consuming 7.4 trillion cigarettes 1 . About 7.7 million deaths and 200 million disability adjusted life years (DALYs) were attributed to smoking and was the leading risk factor for death among males in that year 1  There is an argument that when smoking prevalence decreases a greater proportion of the remaining smokers are likely to be hardcore smokers (HCS). However, the recent editorial in the journal Tobacco Control stated that this argument is not supported by evidence. As per this editorial, softening (when smoking prevalence decreases, a greater proportion of the remaining smokers are likely to be non-HCS) was happening instead of hardening 3 . In high income countries, most evidence point towards softening rather than hardening 4,5 .
Although the definition of hardcore smoking varies across different studies, the common hardening indicators include measures of nicotine dependence and quitting behavior such as past experience of quitting and intention to quit. Even in low-and middle-income countries, a recent study reported a reduction of hard-core smokers in seven of the 10 countries included in the study and no change in one country (Mexico) 2 . In the remaining two countries (China and Turkey) hardcore smoking increased. In India, hardcore smoking decreased from 2.5% to 1.6% as per the above study 2 . Another study from low-and middle-income countries reported that the proportion of hardcore smokers varies from country to country and comprehensive and efficient implementation of the current tobacco control policies are likely to reduce the number of hardcore smokers 6 . A recent systematic review on hardcore smoking argued that there was no evidence for the occurrence of hardening (i.e. when smoking prevalence decreases, the remaining smokers likely to be HCS), and tobacco control policies based on hardening should be challenged 7 . A study from Australia reported that there was a greater accumulation of hardcore smokers in the low socioeconomic status group compared to the high socioeconomic status group indicating that hardcore smokers in low SES groups are less likely to quit 8 . Data on the prevalence of hardcore smokers among different socioeconomic status groups in low-and middle-income countries are limited. In order to address this issue, we looked at the change in hardcore smoking during the period from 2010 to 2017 using the Global Adult Tobacco Survey (GATS) data in India with the following objectives: 1) to look at the association between socioeconomic status (SES) and HCS in India.
2) to find out the trend in HCS in different SES groups in India and 3) to find state-wide variations in hardcore smoking in India. Study design and participants GATS India was the first nationally representative survey in which electronic handheld devices were used as survey instruments. A stratified multistage probability sampling technique was used in GATS and a standardized questionnaire was used for the survey. Households were selected randomly from selected locations and within household eligible persons were interviewed randomly. Privacy was maintained during the interview. The survey method and survey instrument are described in detail elsewhere 9 . To identify the hardcore smokers and factors associated with hardcore smoking in India, GATS India survey data of 2009-10 and 2016-17 were used. The data for GATS 1 & 2 surveys in India were accessed from CDC's GATS survey data webpage. We included smokers aged 25 years and above because majority of the studies on hardcore smoking included smokers aged 25 years and above and, in many studies, hardcore smoking definition included at least 5-6 years of smoking history 7 . Data from individuals aged 25 years and above (GATS 1= 56006, GATS 2= 60837) from both GATS India surveys were extracted to be included in the study 10 .

Variables in the study
Hardcore smoking was the outcome variable, and it was defined using several indicators based on previous studies 11,12 . An individual was categorized as a Hardcore smoker based on

Amendments from Version 2
In the revised version, we have removed the knowledge variable from the definition of "hardcore smoking". The current definition of a Hardcore smoker is based on the following criteria: (1) current smoker, (2) smokes 10 or more cigarettes/day, (3) smokes first puff within 30 minutes after waking up, (4) no quit attempt in last 12 months, and (5) no intention to quit at all or in the next 12 months.
In the revised analysis, we found that urban adults were more likely to be hardcore smokers compared to their rural counterparts. Similarly adults who initiated smoking above 20 years were more likely to be hardcore smokers compared to those who initiated below 20 years of age.
Any further responses from the reviewers can be found at the end of the article REVISED following criteria: (1) current smoker, (2) smokes 10 or more cigarettes/day, (3) smokes first puff within 30 minutes after waking up, (4) no quit attempt in last 12 months, and (5) no intention to quit at all or in the next 12 months. Several other researchers have used all six criteria or a few out of the six to define hardcore smoking 11,12 . All five indicators are captured as individual variables in GATS 1 & 2 datasets. In our study, we considered hardcore smoking if the individuals met all five criteria. Interviewees who answered 'yes' to the question that smoking causes serious illnesses (e.g., heart attack, lung cancer, stroke) were considered as knowing the harmful effect of smoking. Those who answered 'no' or 'don't know' were defined as not having knowledge on harmful effects of smoking.
The other variables included in the analyses were: gender (male or female), type of residence (urban or rural), marital status (separated/divorced, single, married, widowed), occupation (not working, working), education (no formal schooling, education up to higher secondary, graduation and above), age group in years (25-35, 36-45, > 45years), caste (scheduled caste, scheduled tribe, other backward caste, others), age of smoking initiation (<20 years, ≥20 years), smoking practices inside home (no rules, allowed, not allowed but exceptions, never allowed) and wealth index. The wealth index was calculated as a composite index of household's ownership of assets such as television, radio, refrigerator, fixed telephone, cell phone, flush toilet, car, scooter/moped/motorcycle, washing machine, computer/laptop etc 13 . Based on wealth index the sample was divided into five quintiles (poorest, poor, middle, rich and richest) 14 . As self-reported income is not a reliable indicator, wealth index calculated based on asset ownership is used as a proxy indicator for economic status 15 .
Statistical analysis IBM SPSS Statistics for Windows, Version 25.0 Armonk, NY: IBM Corp was used for data cleaning, preparation, and analysis. State-wise prevalence of current daily smokers and hardcore smokers was calculated from GATS 1 and 2 datasets using univariate analysis. Country sample weights were applied to balance for the complex sampling design adopted in GATS, to approximate prevalence rates and 95% confidence intervals (95% CI). Percentage distribution of hardcore and non-hardcore smokers among the survey population was analyzed using frequency tables. Binary logistic regression analysis was done to find the factors associated with hardcore smoking. The significance level in this study was fixed at p-value<0.05

Results
The sample characteristics for the first and second rounds of the GATS in India are described in Table 1. The proportion of adults aged more than 45 years was more in GATS 2 compared to GATS 1. Marital status and caste data were available only in GATS 2.
Change in current daily smoking and hardcore smoking prevalence between GATS 1 and GATS 2 in different states of India are given in Table 2. For India as a whole, current daily smoking decreased from 13.9% to 11.0% and hardcore smoking deceased from 3% to 2.1%. The relative reduction of hardcore smokers was 30% compared to 20.9% in current daily smokers. Current daily smoking prevalence in GATS 1 was highest in Mizoram (39.1%) and lowest in Goa (5.3%) while in GATS 2 the highest prevalence was in Mizoram (33.9%) and Meghalaya (32.4%) and the lowest in Goa (2.1%). Hardcore smoking prevalence in GATS 1 was highest in Mizoram (21.3%) and lowest in Goa (0%) whereas in GATS 2 the highest prevalence was in Meghalaya (6.3%) and the lowest in Goa (0%) and Sikkim (0%).
Change in current daily smoking and hardcore smoking prevalence between GATS 1 and GATS 2 among males are given in Table 3. Current daily smoking prevalence among males in India decreased from 23.9% in GATS 1 to 19.6% in GATS 2 and that of hardcore smoking from 5.6% to 3.9%. Relative reduction of current daily smokers was 18% and that of hardcore smokers was 30.4%. Current daily smoking prevalence among males in GATS 1 was highest in Meghalaya (64.7%) and lowest in Jharkhand (6.3%) and Goa (8.2%). In GATS 2 also the highest and lowest prevalence was in the states Meghalaya (58.3%) and Goa (4.3%), albeit decreased rates in both states. In GATS 1 the highest prevalence of hardcore smoking among males was in Mizoram (37.5%) and lowest was in Goa (0%) and Sikkim (0%) while in GATS 2 the highest prevalence was in Meghalaya (12.7%) and lowest in Goa (0%) and Sikkim (0%).
Change in current daily smoking and hard-core smoking prevalence between GATS 1 and GATS 2 among females is given in Table 4. Current daily smoking prevalence among females in India decreased from 3.3% in GATS 1 to 2.2% in GATS 2. The hardcore smoking among females aged 25 years and above at the national level decreased from 0.3% to 0.2%. Relative reduction was 33% for current daily smokers 33.3% for hardcore smokers. Current daily smoking prevalence among females in GATS 1 was highest in Mizoram (19.0%) and lowest in states like Kerala (0.0%), Maharashtra (0.0%) and Puducherry (0.0%). In GATS 2 the highest prevalence of current daily smoking was in Mizoram (12.9%) and lowest prevalence of 0.0% was reported from Chandigarh. In GATS 1 the highest prevalence of hardcore smoking among females was in Manipur (5.1%) and hardcore smoking was not reported from 18 states and union territories, while in GATS 2 the highest prevalence was in Mizoram (3.2%) and hardcore smoking was not reported from 15 states and union territories.
Factors associated with hardcore smoking in India based on regression analysis of GATS 1 data are presented in Table 5. Males, urban residents, working adults and those without any formal schooling were more likely to report hardcore smoking compared to their counterparts. Compared to those aged 25-35 years, those aged 36-45 years were more likely to report hardcore smoking. Compared to the richest group as per the wealth index all other categories were more likely to report hardcore smoking. Adults who initiated smoking above 20 years were more likely to report hardcore smoking compared to those who initiated below 20 years. With regard to smoking practices inside the home, those who did not have      any rules and those allowed smoking inside the home were more likely to report hardcore smoking.
Factors associated with hardcore smoking in India based on regression analysis of GATS 2 data are presented in Table 6.
Males, working adults and those without any formal schooling were more likely to report hardcore smoking compared to their counterparts similar to GATS 1. However, the rural urban difference in hardcore smoking was not significant in GATS 2. With regard to the age group, those who were in the age group of more than 45 years reported hardcore smoking more than those in the age group of 25-35 years. Regarding the wealth index categories, those who belonged to the poorest, poor and the middle group categories were more likely to report hardcore smoking compared to the richest wealth index group. Adults who initiated smoking above 20 years were more likely to report hardcore smoking compared to those who initiated below 20 years. With regard to smoking practices inside the home, those who did not have any rules and those who allowed smoking inside their home were more likely to report hardcore smoking similar to GATS 1.
The percentage of hardcore smokers across wealth quintiles among individuals aged 25 years and above in GATS 1 and GATS 2 is given in Figure 1. Only in the poorest wealth quintile was there an increase in the proportion of hardcore smokers in GATS 2 compared to GATS 1.
The proportion of hardcore smokers among current daily smokers and the changes between GATS 1 and GATS 2 are given in Table 7. There was an overall reduction of 2.5% in the percentage of hardcore smokers among current daily smokers in India in GATS 2 compared to GATS 1. In 10 states and union territories there was increase in the proportion of hardcore smokers ranging from 1.8% in Odisha to 15% in Assam. In all other states and union territories there was decrease in the proportion of hardcore smoking. In Goa the proportion of hardcore smokers remained zero in both GATS 1 and 2. In Kerala, there was no change in propotion of hardcore smokers in both GATS 1 and GATS 2.

Discussion
There were reductions in the prevalence of current daily smoking and hardcore smoking in India during the period between GATS 1 and GATS 2. A relative reduction of hardcore smoking prevalence of 30% was more than the 20.9% reduction in current daily smoking indicating softening of smoking rather than hardening during the period from GATS 1 to GATS 2. Compared to the GATS 1 data hardcore smoking increased only in the poorest wealth group in GATS 2 while in all the other wealth groups, hardcore smoking decreased demonstrating a greater accumulation of hardcore smokers in the poorest group as reported from Australia 8 . Current smoking prevalence was also reported to be highest among the poorest groups in India 14 . While the proportion of hardcore smokers in the poorest group remained high after a period of seven years between GATS 1 and GATS 2, in all the other wealth groups, there was a decline. Tobacco control measures need to focus on people belonging to the poorest group and offer them help to quit tobacco since they are likely to be more resistant to comply as reported recently in a US study 16 .
In GATS 1 the association between hardcore smoking and the wealth groups did not show a gradient. Compared to the richest wealth group those belonging to the poor wealth group showed the highest odds of reporting hardcore smoking whereas in GATS 2 the odds of reporting hardcore smoking was highest among poorest wealth group. The accumulation of hardcore smokers was gradually shifting to the poorest group in GATS 2. While there was an overall decrease in hardcore smoking some population groups such as those belonging to the poorest section of the society seem to continue hardcore smoking. It may be important to initiate tailor made tobacco cessation efforts for smokers in the low SES groups.
While majority of the Indian states reported reduction in hardcore smokers a few states such as Jammu and Kashmir, MP, Assam and Gujarat reported an increase in the prevalence of hardcore smokers. Respondents from the state of Goa did not report hardcore smoking in both the surveys, and there was a reduction in the current daily smoking prevalence. This highlights the importance of implementation of Cigarettes and Other Tobacco Products Act (COTPA) in all states particularly those where there was increase in hardcore  smoking. In the states where there was an increase in hardcore smoking there was reduction in current daily smoking during the same period. In Odisha interesting findings were observed. Both the states had an overall reduction in the current smokers and hardcore smokers as a percentage among total sample ≥ 25 years. However, the percentage of hardcore smokers within current smokers from GATS 1 to GATS 2 has grown in Odisha, Jammu and Kashmir, Uttarakhand, Uttar Pradesh, Chhattisgarh, Madhya Pradesh, Odisha, Bihar, Assam, Gujarat and Puducherry. This could be due to the accumulation of hardcore smokers when the overall prevalence of smoking reduced as reported in the Australian study 8 .
Only three states reported an increase in current daily smoking: Punjab, Jharkhand and Tamil Nadu. In all these three states there was reduction in hardcore smoking.
Since smoking in India and most south Asian countries is predominantly a male behavior it is important to see the trends in hardcore smoking among males. Among the states where there was increase in hardcore smoking in the total population, Jammu & Kashmir, Uttarakhand, Madhya Pradesh, Manipur, Assam, Gujarat and Bihar etc., also reported an increase in hardcore smoking among males, although the increase was small. In Puducherry, Tamil Nadu, Goa, Sikkim, Jharkhand there was no change in hardcore smoking among males during this period.
Hardcore smoking was not reported by females in several states in GATS 2. However, a few states namely Jammu and Kashmir, Punjab, Uttarakhand, Haryana, Delhi, UP, Assam, Karnataka, Kerala and Gujarat reported an increase in hardcore smoking indicating an unhealthy hardcore smoking trend in women. The increasing trend in smoking among women in India was reported earlier also 17 . This could be due to the targeting of women and people belonging to low socioeconomic group by the tobacco companies circumventing the tobacco control laws in India as reported earlier 18 .
In GATS 1 the major factors associated with hardcore smoking were male gender, urban residents, working adults, older adults, those who initiated smoking after 20 years and those with no schooling or low education. Significance of urban rural difference of hardcore smoking disappeared in GATS 2. This was because of the reduced prevalence of hardcore smoking in both rural and urban areas in GATS 2 compared to the GATS 1. Those who initiated smoking above 20 years were more likely to report hardcore smoking in both GATS 1 and GATS 2 19 .
Regarding smoking practices inside home, in both GATS 1 and GATS 2 hardcore smoking was more likely to be reported where smoking was allowed inside home, where there were no rules for smoking inside home or there were exemptions for the rule compared to those households where smoking was not allowed inside the home. This is a very positive finding that can be replicated in several Indian and other similar settings. It was already reported that such smoke free home initiatives of not allowing smoking inside home in Kerala State of India 20 and Indonesia 21 were successfully implemented.
Although the use of the term hardcore smoking and hardening has been challenged recently 22 researchers continue to report hardcore smoking and the groups in which this continues to be a problem. The strength of this study was the large representative data of the entire country. A limitation was that tobacco use was self-reported and there was no validation by bio-chemical measures such as cotinine estimation.

Conclusions
This study based on GATS 1 and GATS 2 data found that there was reduction in current daily smoking and hardcore smoking in India during the period between the two surveys.
Although there was reduction in current daily smoking and hardcore smoking in the country as a whole a few states namely Punjab, Jharkhand and Tamil Nadu reported an increase in current daily smoking and four states namely Jammu and Kashmir, MP, Assam and Gujarat reported an increase in hardcore smoking. In all these four states where there was increase in hardcore smoking, current daily smoking prevalence decreased probably due to the accumulation of hardcore smokers as a result of reduced prevalence of current daily smokers. It was also seen that only in the poorest wealth group there was an increase in the hardcore smoking during the period between GATS 1 and GATS 2. In all other wealth groups, there was reduction in hardcore smoking. This is similar to the finding of the study from Australia that when current smoking prevalence decreases there is likely to be an accumulation of hardcore smokers particularly in the poorest socioeconomic groups. Tobacco control and cessation measures need to be implemented focusing on people belonging to the poorest socioeconomic groups in India.

Open Peer Review
My main comment is related to the definition of 'Hardcore Smoking' employed in the manuscript. I don't understand why lack of knowledge of harmful effects of smoking is considered an attribute of HCS. It could be considered more 'hardcore' to know that smoking is harmful but to continue regardless than through lack of awareness. I don't think that including lack of awareness of health impacts of smoking is a useful measure of HCS. It would be good to demonstrate what the results would be if this part of the definition were omitted.

Minor issues:
"Smoking was attributed to…" -I believe this should be phrased the other way around, i.e., the deaths were attributed to smoking not smoking attributed to the deaths.
○ I was surprised to see no mention of smokeless tobacco in the discussion, given high rates of smokeless tobacco use in India compared to other countries mentioned in the manuscript (e.g. Australia).

If applicable, is the statistical analysis and its interpretation appropriate? Yes
Are all the source data underlying the results available to ensure full reproducibility? Yes Are the conclusions drawn adequately supported by the results? Partly © 2022 Thekkumkara Surendran Nair A. This is an open access peer review report distributed under the terms