Introduction

Cigarette smoking prevalence among PLHIV varies greatly between studies, although it is consistently higher than the prevalence of smoking in the general population [1,2,3,4,5,6,7,8,9]. A nationally representative US study reported that 42% of PLHIV were current cigarette smokers, which was twice the prevalence of smoking among the general US population [10]. Living with HIV has also been associated with not quitting smoking [11]. Smoking tobacco is a major contributor to morbidity and mortality among PLHIV [12] with more years of life lost from smoking than HIV itself [13]. This significant health burden highlights the need for action.

The need for targeted, or tailored smoking cessation interventions for PLHIV has been previously acknowledged [9, 14,15,16,17]. It is argued that further understanding of the characteristics of PLHIV who smoke, may assist the development of interventions, inform tobacco control policy [18] and increase cessation rates by better addressing their unique needs. Current international literature suggests that gender [10], sexual orientation, employment status [19], education level, standard of living [10] or socio-economic status [20] as well as other substance use [10, 21,22,23], such as alcohol and cannabis [6, 9, 16], is associated with cigarette smoking. Uncontrolled HIV infection and a greater time since diagnosis have also been associated with current smoking among PLHIV in France [24].

Many studies that report factors associated with smoking are however convenience samples or sub-population [3, 6, 25] or from smoking cessation trials [19, 26] and mixed findings are common. Evidence suggests contextual societal characteristics of neighbourhoods and countries also shape smoking behaviour [27]. Continuing to develop this body of literature and in a variety of contexts is therefore necessary and lacking within the Australian setting. Quitting smoking is associated with increased treatment engagement and better HIV related outcomes [28], but few studies have investigated the socio-demographic characteristics related to quitting. Further evidence on the determinants of smoking and successful quitting will elucidate the facilitators and barriers to smoking cessation among PLHIV.

Major health events (such as illness onset or diagnosis) can prompt health promoting behaviours and therefore offer ‘teachable moments’ for smoking cessation [29]. Although, the exact timing of behaviour change was not reported, a previous study found that many PLHIV engage in health promoting behaviours, such as diet improvement and quitting or reducing smoking cigarettes, following their HIV diagnosis [30]. While others have reported a significant rise in intention to quit in the first 3 months following HIV treatment initiation [31]. An increase in smoking cessation within the first 12 months post HIV treatment initiation was also reported but was not significant [31].

Using a large cross-sectional sample of PLHIV in Australia this study provides one of the first prevalence estimates of daily smoking and explores the factors associated with smoking and quitting and the relationship between quitting and HIV diagnosis and treatment.

Methods

Procedure

For this study we utilised data that were collected as part of the ‘HIV Futures 7’ study, a national survey of PLHIV living in Australia [32]. HIV Futures is a repeated cross-sectional study that has been conducted periodically since 1997 [32]. Data for the seventh survey in the series (HIV Futures 7), were collected between October 2011 and April 2012. The HIV Futures 7 study report and full methods have been previously reported [32].

Participants in the HIV Futures 7 study were required to be aged 18 years or older, residing in Australia and living with HIV. A self-report survey that could be completed via a hardcopy booklet or online was advertised through networks of PLHIV, relevant websites, HIV clinics, PLHIV community organisations and events.

Ethics approval for this survey and its secondary analysis was acquired through the La Trobe University College of Science, Health and Engineering Human Research Ethics Committee. Ethical approval for secondary analysis was also approved by the School of Public Health, University of Queensland Human Research Ethics Committee.

Measures

The survey contained approximately 170 items relating to demographics, financial and employment security, physical and mental health, use of antiretroviral treatment, relationships and sexuality, social support, drug and alcohol use and use of HIV and other clinical services. The current study utilised measures of:

  • Demographics: age, sex, sexuality, relationship status, Aboriginal or Torres Strait Islander status, location, employment, education and income source

  • HIV diagnosis and treatment: length of time since HIV diagnosis, year of HIV diagnosis, source of infection and treatment status

  • Drug and alcohol use: frequency of alcohol and cannabis use, purpose of cannabis use, other illicit drug use (in past 12 months), relevance of drug use to identity

  • Cigarette smoking: smoking status, average cigarettes per day (CPD), length of time smoking, products or programs used in a quit attempt

Please see the first supplementary file for survey items used.

Analysis

Data analysis was conducted using IBM Statistical Package for the Social Sciences 23 [33]. In the first instance self-reported smoking status and CPD data was used to create new smoking status categories. For the purpose of the analysis, current daily smokers included those who self-identified as a currently smoking cigarettes and smoked on average one or more CPD, past regular smokers included those who identified as smoking in the past and when they smoked, smoked one or more CPD and never smokers were those who identified as having never smoked cigarettes. Participants who had missing data for either smoking status or CPD were excluded from the analysis.

To provide a prevalence estimate of smoking, data were weighted (adjusting for age and sex) to reflect the Australian PLHIV population. Estimates of the age and sex distribution, from national surveillance data (obtained from Gray, R, The Kirby Institute, University of New South Wales on May 19, 2016) of the PLHIV population in Australia were used to create a weighting variable. Weighted data were used for reporting all descriptive statistics. Fifteen participants could not be weighted due to missing age or sex data and were therefore removed from the data set. The data set also contained missing data for other socio-demographic items (e.g. education level), contained in the table of the second supplementary file. A logistic regression analysis was completed to investigate associations between socio-demographic variables and current smoking. For this analysis participants who had ever regularly smoked (current regular and past regular smokers) were compared with never smokers. A second logistic regression investigated associations between socio-demographic variables and successful quitting by comparing past regular smokers with current regular smokers. Only cisgender males and females were included in the regression analysis due to the small numbers of transgender and other respondents (total n = 4). Variables were removed from the regression analysis where there were concerns of multicollinearity and where small cells existed, categories of variables were also collapsed to increase statistical power. Tables 1 and 2 report the results of these analyses and are written to two decimal places to assist with the interpretation of small effects. Finally, among past smokers, year of quitting, HIV diagnosis and ART initiation were compared to explore the relationship between these variables (see Fig. 1). Exact dates (beyond year), were not available and it is therefore not possible to report whether quitting occurred before or after diagnosis/treatment initiation for participants who quit within the same years as their diagnosis and/or treatment initiation.

Table 1 Logistic regression of factors associated with smoking
Full size table
Table 2 Logistic regression of factors associated with past smoking
Full size table
Fig. 1
figure 1

Years to quitting in relation to diagnosis and treatment years

Full size image

Results

Smoking Status

After participants with missing data relevant to smoking status and weighting were removed, a total of 1011 participants remained. After weighting participants (on age and sex of the Australian PLHIV population) 31.6% were current regular smokers, while 30.8% previously smoked regularly but had successfully quit and the remaining 38.6% had never regularly smoked cigarettes. Past smokers smoked a greater number of CPD, smoking on average 22 CPD (SD = 13.7) when they did smoke, compared to the current smokers, who smoked 17 CPD (SD = 9.0), t(535.46) = 4.7, p = < 0.001. Many of the past smokers (38.1%) had quit ten or more years ago.

Characteristics of the weighted sample by smoking status can be seen in the second supplementary file. The median age of current smokers was 46 years and slightly younger than the never and past smokers (49 and 52 years respectively). The median age of the entire sample was 49 years. Most respondents were men (92.9%). There was a greater proportion of females in the never smoker category (9.5%) than the current or past smoker categories (5.2% and 4.8% respectively). The proportion of respondents not in a relationship was greatest among those currently smoking, with 61.6% of current smokers not in a relationship compared to 56.5% of past smokers and 53.8% of never smokers.

Among respondents who worked full-time, 43.4% were never smokers, 29.0% were past smokers and 27.6% were current smokers. While among those who were unemployed, 28.4% were never smokers and 48.0% were current smokers. The most frequently reported highest level of education among those who currently smoked was a tertiary diploma or trade certificate (26.1%) while among those who had never smoked, the highest educational attainment reported most frequently was a postgraduate university degree or diploma (29.8%); 55.2% of those with a postgraduate university qualification had never smoked while only 20.0% currently smoked. Of the current smokers, salary was the main income source for 44.9% of respondents and social benefits for 44.3%. A smaller proportion of never smokers reported social benefits as a primary income source (25.8%). Among current smokers, 28.3% smoked cannabis daily or weekly and 32.1% had smoked cannabis in the past 12 months. In contrast, among those who had never smoked cigarettes, only 3.9% smoked cannabis daily or weekly and 14.1% had smoked cannabis in the past 12 months.

Among both current and past smokers, the product most commonly used in a quit attempt was nicotine replacement therapy (50.3% and 34.1% respectively) followed by prescription medication (38.4% and 27.0% respectively). 65% of current smokers had used at least one of the listed products or programs in a quit attempt, while only 51.8% of past smokers had done so.

Ever Smoking

A logistic regression was performed to assess the relationship between socio-demographic variables and whether a participant had ever been a regular smoker. The full model containing all predictors was able to distinguish between respondents who were identified as having ever been or never been a regular smoker [x2 (12, n = 927) = 168.5, p < 0.01]. The model as a whole explained between 16.6% (Cox and Snell R2) and 22.6% (Nagelkerke R2) of the variance in smoking status and correctly classified 70.1% of cases.

As shown in Table 1, in the unadjusted model: sex, sexuality, education level, treatment status, alcohol and cannabis use and income source were all significantly associated with having ever smoked cigarettes regularly. However, after adjustment for co-variates, only four of the independent variables (level of education, alcohol use, cannabis use and income source) were statistically significant. The strongest predictor of having ever been a regular smoker (past or present) was regular cannabis use; individuals who reported regular use (weekly or daily) had 6.2 times higher odds of having ever smoked cigarettes regularly than those who reported no cannabis use.

Quitting Profile

Of the 311 past smokers, 76 (24.4%) quit prior to their HIV diagnosis while 23 (7.4%) quit in the same year as their diagnosis and 205 (65.9%) quit one or more years after diagnosis. Data on year of quitting was unavailable for eight (2.6%) past smokers. An increase in quitting occurred around the time of diagnosis (see Fig. 1.) Specifically, an increase in quitting occurred 1–2 years prior to diagnosis, this trend increased the year of diagnosis and remained high until 10 years post diagnosis when a decline occurred. The median number of years to quitting after diagnosis was eight (IQR = 4–16).

Of those who quit smoking, approximately one-third (32.9%) quit prior to the year they commenced ART, 5.4% (n = 17) quit in the same year of commencing ART, while 50.6% (n = 157) quit smoking one or more years after commencing ART. The remaining 35 respondents (11.2%) had missing data (either year of treatment initiation or year quit) or had never taken medication for their HIV diagnosis. The number of individuals who quit, rose in the year of, and year or two prior to treatment initiation and peaked in the 2 years after treatment initiation.

Table 2 shows the results for the unadjusted and adjusted models which explored the relationship between the same socio-demographic variables and having successfully quit smoking cigarettes. In unadjusted models, age, treatment status, alcohol and cannabis use, income source, and CPD were significantly associated with having quit smoking. The full model containing all predictors was able to distinguish between respondents who were identified as having successfully quit smoking, or continued to smoke regularly [x2 (14, n = 575) = 180.0, p < 0.01]. The model explained between 26.9% (Cox and Snell R2) and 35.9% (Nagelkerke R2) of the variance in smoking status and correctly classified 75.7% of cases. As shown in Table 2, a number of variables made a statistically significant contribution to the adjusted model, including: age, receiving ART, not using cannabis, number of CPD and, length of time smoking. The strongest predictor of quitting was receiving ART with an odds ratio of 2.4 (95% CI 1.2–4.7).

Discussion

In this study, we estimated that the prevalence of daily smoking among PLHIV in Australia in 2011/12 was 30.6%, compared to 16.3% in the broader Australian population [34]. Smoking among PLHIV has been scarcely studied in the Australian context, however our findings are similar to international studies that have reported a smoking rate among PLHIV of 2–3 times that of the general population [10, 35,36,37]. A recent large-scale study of PLHIV from Europe, USA and Australia reported a similar percentage of never smokers (30%) although, the authors reported a higher ratio of current:former smokers (49% and 21% respectively) than in the present study [38]. Consistently higher rates of cigarette use among PLHIV reinforces the need to address smoking among PLHIV.

The sample of PLHIV in the current study smoked a higher number of CPD (17 CPD) than the general Australian population, who smoke on average 12 CPD [39], consistent with previous international findings [40,41,42,43,44]. Past smokers also smoked significantly more CPD than current smokers (22 compared to 17 CPD respectively). This finding may be explained by the fact that a large proportion of past smokers in the current sample quit 10 or more years ago and CPD consumption has steadily declined with increasing taxation and widespread public smoking bans [45]. In 1989 Australian males smoked on average 24 CPD (women 22) [45], compared to 12 CPD by current daily Australian smokers [39]. Alternatively, this finding may reflect a relationship between greater nicotine dependency and increased readiness to quit [46] or a greater perceived need to stop smoking.

We identified that lower income and education, and greater alcohol and cannabis use, were associated with ever having smoked cigarettes regularly. These features have previously been correlated with current smoking among PLHIV [6, 46, 47]. Those smoking cigarettes therefore reflect a further disadvantaged group of the PLHIV population. Although others have reported such disadvantage may impact quitting [48], our study only found these variables to be associated with ever smoking, rather than with quitting.

Regular cannabis use was the strongest predictor of ever being a regular smoker. Those who used cannabis regularly (weekly or daily) were six times more likely to smoke cigarettes than those who reported no cannabis use in the past 12 months, a rate similar to other PLHIV samples [9]. Consistent with other research, cannabis was also the most commonly used illicit drug [19, 49,50,51]. Illicit drug use is not only associated with smoking status [52], but with reduced interest in quitting cigarettes [3] and reduced likelihood of making a quit attempt [53]. The act of smoking cannabis, particularly if mixed with tobacco, presents a unique challenge to quitting smoking tobacco cigarettes and may be need to be considered in tailored smoking cessation programs that have been advocated for PLHIV [15, 54].

Past smoking (i.e. successful quitting) among ever smokers was associated with increased odds of being older, receiving ART, not regularly using cannabis, smoking a greater number of CPD and slightly decreased odds of smoking for longer. Although other studies have assessed factors associated with cigarette use [6, 24], few have investigated factors associated with successful quitting. Our results suggest those who quit may have been more aware of their need to quit (smoking more than others) and who are engaged in other health promoting behaviours, such as not using cannabis and on ART, and as a result, have lived longer. Longer duration of smoking was however associated with a slightly decreased odds of quitting and others have reported similar results [55]. This finding may reflect increased difficulty in quitting as smoking behaviour becomes further ingrained over time.

Our study also adds to early research on the potential important relationship between HIV diagnosis and treatment with smoking and quitting behaviour. Of those who quit, most (73%) did so in the year of, or subsequent year to their HIV diagnosis. Burkhalter and colleagues reported that 77% of their sample of past smokers quit after diagnosis [3] while Benard et al. reported 67% of their sample quit after diagnosis [1].

There was also an observed rise in quitting around the time of, and soon after both HIV diagnosis and ART initiation. Those receiving ART were also more likely to have quit. Previous research has indicated a relationship between treatment initiation and an increased intention to quit smoking [31], being more likely to quit after initiating HIV care [55] and that current smokers are less likely to adhere to ART (compared to past or never smokers) [56]. HIV diagnosis or treatment uptake might therefore stimulate quitting, possibly due to fears of future ill health which strongly contributes to perceived quality of life among PLHIV [57]. HIV diagnosis and ART initiation may therefore present opportune times for addressing smoking and providing cessation intervention options at these times could increase quit rates.

This study is not without limitations however and due to the use of a convenience sample, generalisability cannot be assumed. Data were however weighted by age and sex (using national surveillance data of PLHIV) in an effort to assist in accurately reflecting the Australian PLHIV population. Second, as exact date of quitting, ART initiation and HIV diagnosis was unavailable it was not possible to know the order of events for participants who quit in the same year as their diagnosis or treatment initiation (Fig. 1). Furthermore, although Fig. 1 does not show statistical association, it does illustrate a trend that warrants further research into the timing of smoking cessation interventions for PLHIV.

The current study provides some of the first data on smoking, including its prevalence among PLHIV in Australia using one of the largest data sets of Australian PLHIV available. This research also adds to the body of literature on factors associated with smoking and the limited understanding of factors associated with quitting. Understanding the characteristics of PLHIV who smoke and quit will assist the development of smoking cessation programs [58]. For example, the findings regarding cannabis may suggest the need for concurrent treatment of cannabis and tobacco smoking [59], or the need to address dual tobacco and cannabis use during smoking cessation interventions. Finally, our study also suggested there is potential for HIV diagnosis and ART initiation to stimulate quit attempts. Further research of this topic is needed however, in particular on the influence of offering smoking cessation interventions at these significant times. Addressing smoking at these times may also result in improved HIV related outcomes, with quitting associated with increased HIV treatment engagement, ART adherence and being more likely to sustain HIV RNA suppression [28].