Article Text

Download PDFPDF

Temporal changes of under-reporting of cigarette consumption in population-based studies
  1. Silvano Gallus1,
  2. Irene Tramacere1,
  3. Paolo Boffetta2,3,
  4. Esteve Fernandez4,5,
  5. Silvia Rossi6,
  6. Piergiorgio Zuccaro6,
  7. Paolo Colombo7,
  8. Carlo La Vecchia1,8
  1. 1Istituto di Ricerche Farmacologiche ‘Mario Negri’, Milan, Italy
  2. 2International Prevention Research Institute, Lyon, France
  3. 3The Tisch Cancer Institute, Mount Sinai School of Medicine, New York, USA
  4. 4Cancer Prevention & Control Unit, Institut Català d'Oncologia-IDIBELL, L'Hospitalet, Barcelona, Spain
  5. 5Department of Clinical Sciences, School of Medicine, Universitat de Barcelona, L'Hospitalet, Barcelona, Spain
  6. 6Dipartimento del Farmaco, Istituto Superiore di Sanità, Rome, Italy
  7. 7Istituto DOXA, Gallup International Association, Milan, Italy
  8. 8Istituto di Statistica Medica e Biometria ‘G.A. Maccacaro’, Università degli Studi di Milano, Milan, Italy
  1. Correspondence to Dr Silvano Gallus, Istituto di Ricerche Farmacologiche ‘Mario Negri’, Via Giuseppe La Masa 19, 20156 Milan, Italy; gallus{at}marionegri.it

Abstract

Objectives To monitor trends in under-reporting of smoking in Italy over the last two decades.

Methods A total of 9 representative population-based surveys on smoking conducted in Italy in 1990 and annually between 2001 and 2008, covering 26 397 individuals, were studied. The number of cigarettes per day per person aged 15 years or over, obtained from these interview-based surveys, was compared with official data from legal sales.

Results Over the last two decades, self-reported smoking prevalence progressively decreased from 32% in 1990 to 22% in 2008. Self-reported daily per capita consumption of cigarettes also showed a reduction between 1990 and 2008, notably so over the last few years (from 5.2 in 1990 to 4.0 in 2004 to 3.2 in 2008). According to data from legal sales, number of cigarettes per day per person decreased from 5.3 in 1990 to 5.0 in 1992, levelled off from 1992 to 1997, subsequently increased from 5.0 in 1997 to 5.8 in 2002 (likely due to control of smuggling), and decreased over the last 6 years (to 4.9 in 2008). These figures correspond to an under-reporting of approximately 1% in 1990, 25% in 2001 and up to 35% in 2008.

Conclusions The difference in cigarette consumption between legal sale and self-reported data has substantially increased over the last two decades in Italy, reflecting increasing under-reporting of cigarette consumption mainly due to a decreasing social acceptability of smoking. Comparisons between interview-based and legal sale data are complicated by factors such as smuggling control and changes in the population (eg, increased proportion of immigrants); however these are able to justify only a small proportion of the gap found in Italy.

  • Epidemiology
  • Italy
  • population survey
  • smuggling
  • tobacco smoking
  • prevalence
  • smuggling
  • surveillance and monitoring

Statistics from Altmetric.com

Introduction

Smoking prevalence and consumption for the general population are usually derived from surveys based on self-reported data.1 2 However, misreporting of smoking is a well known phenomenon; its magnitude varies depending upon the historical period, the geographic area, the population studied and the study design,1 3 4 as well as the stage in the cigarette epidemic model.5

In addition to validating self-reported smoking consumption with biomarkers,1 2 4 misreporting of smoking in the general population can be quantified by comparing self-reported data on cigarette consumption (eg, in population surveys) with those derived from national cigarette sales.6–9 In 1978, Warner compared data from four national surveys conducted between 1964 and 1975 in the USA, with US tobacco sales.6 He showed that data from surveys justified 73% of official tobacco trades in 1964 and the proportion declined to 64% in 1975. The explanations provided by the author for the increased gap between official sale and self-reported data included in particular problems in the survey methodologies and increases in the under-reporting of cigarette consumption. A subsequent study showed that data from seven US surveys conducted between 1974 and 1985 corresponded to 72% of cigarette consumption from legal sales (range 69% to 78%), in the absence of a specific temporal pattern.9 Using the same methodology, Todd7 compared misreporting in various countries in different periods. In Australia (2 surveys, period 1974–1976), Canada (3 surveys, period 1965–1974), Denmark (2 surveys, period 1952–1954) and the USA (2 surveys, period 1955–1966) the percentage difference between cigarette consumption based on legal sales and that obtained from surveys was estimated around 20% or over, in Sweden (1 survey, period 1963) it was around 10%, and in Ireland (4 surveys, period 1961–1975) and Japan (5 surveys, period 1970–1974) close to null, in the absence of a specific temporal pattern. A substantial growth in the gap between legal sale and reported data was evident only for Canada, where percentage differences increased from 21% in 1965 to 24% in 1970 and increased to 31% in 1974.7 Van Reek8 showed data from three different surveys conducted in The Netherlands: differences between survey estimates and national figures increased from 0% in 1958 to 23% in 1972 and up to 32% in 1982. Jackson and Beaglehole,10 taking advantage of self-reported census data from New Zealand, found differences of 28% in 1976 and 30% in 1981.

In Italy, La Vecchia,11 comparing data from legal sales of manufactured cigarettes with data from six surveys on smoking conducted with different study designs and sample sizes between 1949 and 1983, showed that interview-based estimates tended to be substantially higher than sale data in earlier years. This has been attributed to the fact that smuggled cigarettes, not taken into account in official legal sales, represented a considerable proportion of the total tobacco trade in the 1950s. Moreover, in 1949, over 30% of smokers also smoked hand-rolled cigarettes.11 From the mid-1970s onwards interview-based estimates tended to be lower than legal data. In fact, in 1983 interview-based estimates explained approximately 75% of legal sales.11 This percentage decreased to 65% in 1986–1987.12 13

Given that the gap between self-reported cigarette consumption and official sales data has rapidly changed over time in several countries, it is important to monitor trends in misreporting of smoking to understand the validity of findings provided by population-based surveys in various populations. In this study, we aimed to consider the issue in Italy over the last two decades, taking advantage of nine population-based surveys on smoking conducted in 1990 and annually between 2001 and 2008.13–20

Methods

We considered data from nine companion surveys on smoking, conducted in 1990 and annually from 2001 to 2008 in Italy by DOXA, the Italian branch of the Gallup International Association.13–20 The participants of the nine surveys were subjects aged 15 years or over and representative of the general Italian population in terms of age, sex, geographic area and socioeconomic characteristics. In 1990, the survey was conducted on approximately 1000 subjects. The sample size for the surveys conducted between 2001 and 2008 was similar (more than 3000 participants each). Overall, the present study is based on 26 397 individuals.

All 9 surveys used the same sampling criteria: the samples were defined through a representative multistage sampling of adults from municipalities (the smallest Italian administrative division) in all 20 Italian regions (the largest Italian administrative division) identified to be representative of the geographic areas sampled. In the municipalities considered, individuals were randomly selected from electoral rolls within strata of sex and age group to be representative of the demographic structure of the population. Wherever the selected subjects were unavailable, they were replaced by selecting among neighbouring (living in the same floor/building/street) subjects with the same gender and age group. Subjects aged between 15 and 17 years, whose names do not appear on the electoral lists, were chosen by means of the ‘quota’ method (by gender and exact age) using the same approach. For each survey, statistical weights were applied in the analysis in order to obtain population estimates.

Ad hoc trained interviewers collected data using a structured questionnaire in the context of a computer-assisted personal in-house interview (CAPI). Besides general information on sociodemographic characteristics, data were collected on smoking behaviours, including smoking status (never/former/current smoker) and number of cigarettes smoked per day. Ever smokers were participants who had smoked 100 or more cigarettes in their lifetime. In the last three surveys (2006–2008), in order to estimate a proxy of the ‘social acceptability’ of smoking, a question was formulated as follows: ‘Are your guests free to smoke in your house?’.18–20

For comparative purposes, in figure 1 we also show cigarette consumption per person from another set of surveys: the Multipurpose Household Surveys, an integrated system of cross-sectional social surveys annually conducted by the National Institute of Statistics (ISTAT) between 1993 and 2007 (2004 excluded).21 These investigations included large samples of families (approximately 20 000 households, corresponding to 50 000 individuals each year), representative of the Italian population aged 15 years or more.22 23

Figure 1

Average number of cigarettes per day per person aged 15 years or over from legal sales data, and from 9 DOXA (the Italian branch of the Gallup International Association) and 14 National Institute of Statistics (ISTAT) surveys. Italy, 1990–2008.

From ISTAT, we also collected data on total legal tobacco sales and the Italian population aged 15 years or over for the period 1990–2008.22 From these data, we estimated the number of cigarettes consumed in 1 year per person aged 15 years or over, assuming 1 cigarette=1 g of tobacco (in Italy, tobacco sales are measured in ‘conventional kg’, based on the assumption that 1 ‘conventional kg’ corresponds to 1000 cigarettes. Therefore, grams of tobacco represent a direct measure of cigarettes). These estimates were compared with the number of cigarettes consumed in 1 year per person aged 15 years or over derived from the survey data.

In order to obtain an estimate of the annual percentage change (APC) of the average number of cigarettes per day per person derived from legal sales data between 1990 and 2008, we applied a linear regression model with a logarithmic transformation.

Results

Table 1 and figure 1 show the estimates of cigarette consumption per person aged ≥15 years per day from DOXA survey data (in 1990 and between 2001 and 2008) and data from official legal sales (between 1990 and 2008). Figure 1 also shows estimates of cigarette consumption per person aged ≥15 years from ISTAT data (between 1993 and 2003 and between 2005 and 2007).

Table 1

Cigarette consumption of the population aged 15 years or over from interview-based and legal sales data, and the corresponding under-reporting of smoking. Italy, 1990 and 2001–2008

Taking into account DOXA surveys, over the last two decades, self-reported smoking prevalence progressively decreased by 10 percentage points (from 32% in 1990 to 22% in 2008), with a slight decline during 1990s and a greater drop over the last 7 years (from 29% in 2001 to 22% in 2008). Self-reported data on number of cigarettes smoked per day per smoker showed a slight decrease from 1990 to 2001 (from 16 to 15 cigarettes per smoker per day), followed by stable values in cigarette consumption (with a mean around 14.5 cigarettes per day over the last decade). Consequently, there was a reduction of cigarettes per day per person aged ≥15 years between 1990 and 2008, to a great extent over the last decade (from 5.2 in 1990 to 4.3 in 2001 to 3.2 in 2008). ISTAT surveys showed a similar decreasing trend of cigarettes per day per person between 1993 and 2007.

According to data from legal sales, the number of cigarettes per day per person aged ≥15 years decreased in the early 1990s (from 5.3 in 1990 to 5.0 in 1992, APC −2.30), levelled off from 1992 to 1997 (APC −0.02), and subsequently increased until 2002 (from 5.0 in 1997 to 5.8 in 2002, APC 3.06). A progressive decrease was evident over the last 6 years (to 4.9 in 2008, APC −2.78). A greater decline was evident between 2004 and 2005.

These figures correspond to a discrepancy on cigarette consumption between legal sale and self-reported data of approximately 1% in 1990, of 25% in 2001 and a progressive increase up to 35% in 2008.

We observed a rise of the percentage prevalence of subjects reporting that their guests could not smoke inside their house from 54.6% in 2006 to 58.4% in 2007 to 62.0% in 2008 (p for trend<0.001) (table 2). Corresponding estimates for non-smokers (similar for never and ex-smokers) were 62.8% in 2006, 67.2% in 2007 and 70.7% in 2008, and those for current smokers were 28.8% in 2006, 30.0% in 2007 and 31.1% in 2008.

Table 2

Percentage distribution of the Italian population according to the question ‘Are your guests free to smoke in your home?’

Discussion

Comparing data from a series of nine population-based surveys and data from official legal sales, we found a substantial increase in the under-reporting of cigarette smoking in Italy since 1990. These trends should however be interpreted with caution, because of three possible sources of bias: smuggling, social acceptability and immigration.

Smuggling

In 1990, legal sales figures well compared with those derived from interview-based surveys. In Italy, cigarette smuggling had increased during the late 1980s and accounted for 10% to 30% of cigarette sales in the early 1990s.13 24–27 Thus, the apparent consistency in per capita cigarette consumption between interview-based and legal sale data in 1990 could be partly or largely explained by smuggled cigarettes.13 Considering a proportion of smuggling on total tobacco trades of 20% in 1990, the gap between self-reported and sales data would grow to 20.9%, an estimate lower but similar to those observed in the early 2000s.

Smuggling had substantially decreased after the NATO intervention in Kosovo in the late 1990s and the corresponding stricter control of the Italian coast.25 27 Moreover, the control of the supply chain of cigarettes and the enforcement action by Custom Authorities in the early 2000s has contributed to substantially reduce large-scale organised smuggling in Italy.26 27 Thus, the increase in legal sales shown between 1997 and 2002 (figure 1) should be largely attributable to the drop of the proportion of cigarette smuggling on total tobacco trades. This is also confirmed by data on smoking prevalence and cigarette consumption provided by population-based surveys conducted by ISTAT, showing a decrease in per capita cigarette consumption also over the period 1997–2002.21 Using data from the surveys described in the present study, we found that over the last few years in Italy smuggling contributed only to a negligible proportion of total tobacco trades (less than 2%).25 27 Moreover, the internet is not used as an instrument for cigarette sales in Italy.25 27

Thus, smuggling may explain most of the apparent consistency observed between legal sales and self-reported data in 1990. In contrast, between 2001 and 2008 illicit trades had a negligible impact on sales data, and therefore on observed trends of the under-reporting of cigarette smoking.

Social acceptability

Socially unacceptable behaviours are likely under-reported.4 28–30 Thus, variation of under-reporting of smoking could be explained in terms of change in the social acceptability of smoking.1 6 31 In particular, the more social pressure that exists against smoking in a society, the larger a smoker's propensity to deny smoking may be.2 This is confirmed by the widely differing rates of deceivers in various subpopulations with more or less pressure against smoking.1 3 4

A household smoking ban is associated with a negative attitude about the social acceptability of smoking.32 33 Thus, over the last three surveys we used a question on household smoking bans as a proxy of social acceptability of smoking.18–20 Among Italians, the prevalence of those who did not allow their guests to smoke inside their home increased from 55% in 2006 to 58% in 2007 and increased to 62% in 2008. The prevalence in 2008 was similar to that found in Finland,34 lower than in the USA,32 35–39 but higher than in Korea40 and Mexico.39 Several studies consistently showed that policies implementing smoke-free public places facilitate the introduction of voluntary smoke-free homes41–44 and decrease the social acceptability of smoking.45–47 This could be due to the fact that the implementation of policies banning smoking in public places increases the knowledge of the personal threat associated with smoking and the harmful effects of secondhand smoke on health, probably conditioned by anti-smoking campaigns aimed at amassing community support for such a ban before its implementation.7 8 46 However, a US study on 5511 students showed that anti-tobacco campaigns had a marginal role on social desirability responses in surveys.48 The last finding notwithstanding, increased social unacceptability of smoking and the corresponding increased under-reporting of cigarettes in Italy over the last few years could be attributed, at least in part, to anti-smoking policies, including in particular comprehensive legislation that came into force in January 2005, banning smoking in all public places including cafes and restaurants.18 49 50

Immigration

Another issue that may have affected temporal changes between self-reported and actual estimates of cigarette consumption could be due to the exclusion from the surveys of an expanding group of subjects whose smoking rates are different from those of the Italians.6 DOXA surveys were representative of the Italian population aged 15 years or over, but they did not include immigrants as they were not included in the electoral rolls. In Italy, according to data from the National Institute of Statistics, regular immigrants resident in Italy numbered 1 335 000 (2.3% of the Italian population) in 2001 and increased to 2 402 000 (4.1%) in 2005 and up to 3 443 000 (5.8%) in 2008.51 Other estimates show that immigrants (not only residents) in Italy numbered almost 4 million (6.7%) in 2008.52 The most prevalent countries of origin were Romania (21.5% of all immigrants), Albania (10.9%), Morocco (10.0%), China (4.2%), Ukraine (3.5%) and Philippines (2.9%).52 As immigrants may have a different distribution by age, gender, socioeconomic level 51 and different smoking prevalence and cigarette consumption related to place of origin,21 they may also have different smoking behaviours compared with the general Italian population. For example, a study conducted in Dublin, Ireland, on 1545 immigrants from Poland (the country of origin of most of the Irish immigrants) and 484 Dubliners, showed that Polish immigrants' smoking prevalence was much higher than that of their Irish counterparts.53 Data from a representative survey conducted in Italy in 2005 on approximately 130 000 individuals, including 3509 immigrants aged 14–64 years and resident in Italy, showed that smoking prevalence in regular immigrants was 21.4% (27.9% in men and 14.6% in women), lower than the prevalence of the Italian population (25.5% overall, 30.9% in men and 20.0% in women).54 However, the smoking prevalence for regular Romanian immigrants was the highest reported (33.0% overall, 38.4% in men and 28.3% in women).54 Moreover, Romanian immigrants resident in Italy numbered 75 000 (5.6% of all Italian immigrants) in 2001, 249 000 (10.4%) in 2005 and 625 000 (18.4%) in 2008, after Romania entered the European Union in 2007.51 52 Given the lack of studies on smoking characteristics of immigrants in Italy, we are not able to accurately quantify the role of immigrants on smoking misreporting. However, also assuming that all new immigrants from 2001 had a smoking prevalence of 40% in men and 30% in women, with smokers consuming an average of 20 cigarettes per day, the level of under-reporting would have been 23.3% in 2001 and 29.9% in 2008. Therefore, although the latter estimates have been derived from a ‘worst case’ analysis, using extremely overestimated data on smoking prevalence and consumption for immigrants, the figures are not able to justify the increased under-reporting observed over the period 2001–2008.

It is also possible that immigrants from countries where cigarette price is lower, including Romania, purchase cigarettes from their country of origin. Accordingly, in the Irish study, the majority of Polish immigrant smokers (57%) purchased cigarettes solely from Poland.53 These considerations also highlight the need for an exploratory study surrounding the economics of tobacco use among immigrant populations in Italy and in other Western European countries with high migrant rates over the last decade.

Alternative explanations

Other alternative explanations suggested by Warner6 seem unlikely to have had a major role on temporal changes of under-reporting of smoking in Italy. These include the reasons discussed below.

Improvement in data on legal sales over time

On this point, we obtained consistent and accurate data on official trades from the same source,22 and we considered sales on cigarettes only (excluding other tobacco products). Moreover, the estimates are not biased by the use of hand-rolled cigarettes, since this habit has gradually decreased since the 1960s11 and over the last two decades represented only a negligible portion of total cigarette sales. In fact, although the use of hand-rolled cigarettes has been increasing during the last few years, it still accounts for less than 2% of total tobacco trades.22 55

Deficiencies in the survey sampling procedures or other methodological problems

The present surveys have some limitations, including the relatively small sample size. Moreover, we are not able to quantify the role of quota sampling using replacements on the representativeness of the sample of smokers. However, all nine surveys considered were conducted by DOXA with the same study design (in house CAPI), the same smoking definition, the same sample size for the period 2001–2008 and approximately the same response rate.13–20 Thus, reproducibility of the findings is satisfactory. Moreover, among various survey designs, interviewer-administered questionnaires provide the most reliable data.1

In Italy, another set of surveys is available on smoking.22 Given the household design of their data collection, these large surveys provide less precise and reliable estimates of smoking prevalence and in particular of cigarette consumption.56 57 A study from the USA comparing smoking status assessed through self-respondents and household (proxy) respondents (2930 matched pairs) showed a relatively limited discrepancy between proxy reports and self-reports for smoking status (4.3%), but the discrepancy substantially increased when considering occasional smokers (35%).56 This suggests that proxy reports may be acceptable to assess smoking prevalence but not cigarette consumption. Accordingly, we found a large discrepancy between per adult cigarette consumption from legal sales and from ISTAT surveys (systematically higher than discrepancies obtained using DOXA estimates over the study period). The gap consistently increased from 23% in 1993, to 36.6% in 2000, to 41.6% in 2005 and did not change thereafter. Compared to DOXA surveys, ISTAT surveys are less susceptible to changes in smoking acceptability (and therefore in under-reporting), being interviews conducted through proxy respondents.

Inappropriate interpretation of the comparison between interview-based and legal sales data

An increase in, for example, foreign tourism during the relevant years could bias the estimates of under-reporting. However, this or other factors, including changes in the proportion of Italian residents purchasing cigarettes abroad, are unlikely to play a role since the absolute magnitude of such factors is small,6 and tourism has not been increasing in Italy since 2000.58

The surveys were representative of Italians aged 15 years or over. Therefore, cigarettes smoked by adolescents aged <15 years are not allowed for in the estimate of the total number of cigarettes smoked from individual-level data. However, average age of starting smoking is relatively high in Italy.59 Moreover, using data from the DOXA surveys conducted between 2001 and 2008, we found, if anything, a decreasing cigarette consumption among teenagers, the number of cigarettes per day per person aged 15–19 years being consistently around two between 2001 and 2006, and one thereafter.

Differences between ‘consumption’ and ‘smoking’ or a decrease in the consumption of ‘cigarette equivalents’

These also seem to be unlikely explanations, although from 1 January 2004 legislation came into force to further limit the maximum content of nicotine (1.0 mg per cigarette), tar yield (10 mg per cigarette) and carbon monoxide (10 mg per cigarette).60 Thus, it is possible that smokers perceive themselves to be smoking fewer cigarettes (ie, ‘cigarette equivalents’).6

Conclusions

In conclusion, we found that levels of under-reporting of smoking sensibly increased over the last two decades in Italy, and today survey-based self-reported cigarette consumption significantly and substantially understates actual sales. The trend of under-reporting of cigarettes is mainly attributable to the increase in social unacceptability of smoking. Changes in smuggling and immigration may have biased the trends, but the latter factors alone are not able to justify the increased gap between self-reported and legal sale data.

What this paper adds

  • Under-reporting of cigarette consumption in population-based surveys has substantially increased over the last two decades in Italy.

  • This mainly reflects increasing social unacceptability of smoking, though other factors may have played an additional role.

References

Footnotes

  • Funding This work was conducted with contributions from the Italian Ministry of Health, the Italian League Against Cancer and the Italian Association for Cancer Research. EF is supported by Instituto de Salud Carlos III, Government of Spain (RTICC RD06/0020/0089) and Ministry of Universities and Research, Government of Catalonia (SGR200500646).

  • Competing interests None.

  • Patient consent Obtained.

  • Provenance and peer review Not commissioned; externally peer reviewed.

Request Permissions

If you wish to reuse any or all of this article please use the link below which will take you to the Copyright Clearance Center’s RightsLink service. You will be able to get a quick price and instant permission to reuse the content in many different ways.