Binge drinking among Brazilian students: a gradient of association with socioeconomic status in five geo-economics regions

Abstract

Aims

1) To describe the characteristics of binge drinking (BD) among high school students in Brazil and 2) the association of BD with students' socioeconomic status (SES) in the five different Brazilian macroregions.

Design

A national multistage probabilistic sample of high school students.

Setting

Students were drawn from 789 public and private schools in each of the 27 Brazilian state capitals.

Participants

17,297 high school students, aged 14 to 18 years.

Measurement

Self-report data about BD practices and SES were analyzed via weighted logistic regressions and a funnel plot.

Findings

Almost 32% of the students engaged in BD in the past-year. Being in the highest SES stratum doubled the risk of BD among students in all five Brazilian macroregions. There was a gradient in the association between past-year BD and socioeconomic status: as SES increased; the chance of having recently engaged in BD also increased. In the Brazilian capitals as a whole, boys versus girls (aOR = 1.40 [95% CI 1.26 to 1.58]), being older (aOR = 1.47 [95% CI 1.40 to 1.55] per each additional year of age) and those attending private schools versus public schools (aOR = 1.39 [95% CI 1.18 to 1.62]), were at greater risk for BD.

Conclusions

Contrary to what is observed in developed countries, students living in Brazilian capitals may be at an increased risk of BD when they belong to the highest socioeconomic status. Adolescents growing up in other emerging economies might have the same association between high SES and BD.

1. Introduction

Most youth experiment alcohol during their adolescence, but the patterns and frequency of alcohol use, particularly, binge drinking patterns (BD), are associated with health-related consequences. BD is defined as consuming at least four/five alcoholic drinks on one single occasion (Wechsler and Nelson 2001) and this behavior is sometimes referred to as “heavy episodic drinking” (Kuntsche et al. 2004). The practice of BD by adolescents is a risk behavior, due to the physiological effects of alcohol intoxication and possible death (Gunja 2011), and also for being associated with higher rates of sexual violence (Chersich et al. 2007), traffic accidents (Zhao et al. 2010), poor school performance, and involvement with other risk behaviors (Miller et al. 2007) and higher incidence of alcoholism in the future (Pitkänen et al. 2005).

In Brazil, a country in which laws prohibit the sale to and consumption of alcoholic beverages by adolescents under 18 years (Romano et al. 2007), almost 16% of adolescents aged 14-17 years-old practiced BD according to a 2007 national household survey (Pinsky et al. 2010). In a representative sample of private school students in the country's largest city (São Paulo), the prevalence of past-month BD was 33% (Sanchez et al. 2011), mainly among wealthier adolescents. This picture puts Brazilian students in a higher range of risk than the ones in U.S. (Nelson et al. 2008) and in several European countries (Hibell et al. 2009).

Brazil is a country of particular interest because it is one of the BRICS countries group (Brazil, Russia, India, China and South Africa), which have as common features very high economic growth in recent years and that apparently will be among the dominant countries of the global economy within a maximum of 50 years (O'Neill et al. 2005). The BRICS countries despite of having a similar pattern on economy are diverse on their alcohol use, according to the WHO estimates for per capita consumption of alcohol. Among the BRICS, Russia is the country with the highest rates of alcohol consumption (15.7 liter of pure alcohol per person/year –l/p) and Brazil (9.2 l/p) and South Africa (9.5 l/p) are in an intermediate position, since the consumption patterns in India (2.6 l/p) and China (5.9 l/p) are well below (WHO 2011).

Brazil's economy outweighs that of all other BRICS countries, and Brazil is expanding its presence in world markets. Traditionally it was a low income country that is now consolidating its upper middle-income rank according to the World Bank (WorldBank 2011) and has recently achieved the position of 6^th world economy (Inman 2011) (at http://www.guardian.co.uk/business/2011/dec/26/brazil-overtakes-uk-economy). However, although this large economy growth Brazil is the 13^rd most unequal country in the world (CIA 2011), with an average Gini coefficient of 0.54 in their geopolitical regions and extreme regional differences of involvement in gross domestic product (GDP) (IBGE 2010). According to the Brazilian government, in the past year, 50% of the Brazilian population belonged to the “medium-class” of socioeconomic status (SES), with an average family income of US$ 1500 and a high power of purchase due to the Brazilian policy of installment payment (at http://blog.planalto.gov.br/ao-vivo-seminario-politicas-publicas-para-uma-nova-classe-media/ last accessed in 01/04/2012).

The country is divided into 5 geo-economic macroregions with very different GDPs per capita, defining 5 smaller Brazils. GDPs in each region varies from US$ 4,679 (in the Northeast, the poorest region) to US$ 13,238 (in the Midwest, the richest region), considering a parity of 1.6 reais to 1 US dollar and data from the IBGE ((IBGE 2010)).

According to the literature, income inequality is as much a risk factor for alcohol use as poverty is. A world survey of adolescents in 34 countries revealed that young adolescents consumed more alcohol and reported more episodes of drunkenness in countries with high income inequality (Elgar et al. 2005). Also, the socioeconomic status of adolescents is not only associated to youth alcohol use (Tomcikova et al. 2011), but also to patterns of alcohol in early adulthood. According to Humensky (2010) white non-Hispanic adolescents in the US that have parents with higher education and higher income are more likely to binge drink on ages 18 to 27.

While most European and North American studies emphasize alcohol consumption as more prevalent among adolescents of lower socioeconomic status (SES) (Baumann et al. 2007; Guilamo-Ramos et al. 2005; Helasoja et al. 2007); according to recent Brazilian epidemiological studies, high SES is associated with alcohol consumption among Brazilian adolescents (Galduroz and Caetano 2004; Sanchez et al. 2011; Soldera et al. 2004). A review of the Brazilian literature on the subject pointed to a lack of information about the patterns of binge drinking in Brazil, in particular the patterns of underage binge drinking (Silveira et al. 2008). Similarly, there is a lack of information of adolescent binge drinking patterns in other emerging economies such as the BRICS countries. In a systematic review of worldwide studies that investigated the relationship between socioeconomic status and health behaviors in adolescence (Hanson and Chen 2007), the authors suggest that the existence of an association between socioeconomic status and alcohol use is not evident in the analyzed articles, but would require further analysis in different countries to improve the evidence.

The peculiarities of the economic and social conditions of Brazil raise the theory that perhaps the association between SES and BD may be different from that found in richer countries. The principle of the social and emotional impact of being rich in a rich country may be different than that of being rich in an emerging economy and highly unequal society, and this relationship can vary according to regional patterns of income and inequality.

Given the need for more studies to build consensus of scientific knowledge or deepening of scientific evidence about the influence of SES on alcohol consumption in emerging economies, this study aims to: 1) describe the characteristics of binge drinking among high school students in Brazil; 2) evaluate the association of the most dangerous pattern of Brazilian adolescent alcohol use - binge drinking - with adolescents' socioeconomic status, in different regions of Brazil.

2. Materials and Methods

2.1 Study design and sample selection

Data came from a cross sectional survey of school-attending yo uths in all the 27 Brazilian state capitals, with classroom survey data collected in 2010 from a sample of the cities' private and public schools. The study' s target population was designed as a representative sample of high school students (10^th to 12^th grade) in these schools, with a two steps random selection process. A total of 789 schools participated in this study, with a school response rate of 86%. The sample size considered a maximum relative error of 10% and a 95% confidence interval for a prevalence of 50%. The student response rate was 79.2% (20.5% were absence in the day of the survey and 0.3% refused to participate). Ninety-eight questionnaires were excluded of the analysis for having a positive answer for a fictitious drug. The present study were limited to high school students aged between 14 and 18 years-old (n=17297).

Some participants had missing or invalid responses to key study variables. For this reason, the effective sample size for the present investigation and the proportion of designated participants with useable data for logistic regressions is 14714 and 85.1%, respectively.

2.2 Assessment Plan

Anonymous standardized paper and pencil questionnaire data were gathered by a trained team of interviewers who worked in the classroom without a teacher present. A questionnaire, with closed form questions adapted from standardized World Health Organization items (Smart et al. 1980) and the European School Survey Project on Alcohol and Other Drugs (ESPAD) questionnaire (Hibell et al. 2009) was used. In average, it took one 40 minutes to the students to complete the questionnaire.

The protocol was reviewed and approved by the UNIFESP Research Ethics Committee (Protocol #0348/08), with provisions for participants to participate anonymously, to decline to participate, to leave questions unanswered, and that they could interrupt their participation at any time.

2.3 Measures

2.3.1 Key Response Variable

The key response variable in this study described binge drinking (BD) in the past year, defined as at least one episode of consumption of five or more servings of alcoholic beverages on the same occasion, as used in the ESPAD survey (Hibell et al. 2009). A serving was defined as a5-oz glass of wine, a 12-oz can of beer or a 1.5-oz shot of liquor and the equivalence examples were drawn on the questionnaire to facilitate students understand.

2.3.2 Covariates under study

The suspected covariate of central interest is social rank as indexed in relation to a highly standardized survey assessment of socio-economic status (SES) known as the ABEP index (Associação Brasileira de Empresas de Pesquisa – Brazilian Association of Research Agencies). The ABEP index (ABEP. 2008) is based upon the educational level of the head of the household, possession of various types of household goods (e.g., television sets), and number of housekeepers. This scale was used to sort participants into standardized subgroups labeled from A to E (A1, A2, B1, B2, C1, C2, D, E; where A1 was the highest economic strata).

Due to the inequality of distribution of Brazilian income, we defined 5 ranks of SES: high SES (A1+A2), medium-high SES (B1+B2), medium-low SES (C1), and low SES (C2+D+E). It was important to include the C2 stratum among the low SES group to have enough cases to run the regressions in all regions, otherwise it would be impossible to analyze data from the South region.

Differences according to Brazilian regions were analyzed considering the 5 geo-economic regions defined by the Brazilian federal government: South (3 states capitals), Southeast (4 states capitals), Midwest (3 states capitals and the Brazilian Capital, Brasilia) Northeast (9 states capitals) and North (7 states capitals) (IBGE, 2011) and that are widely used on the population census.

2.3.3 Descriptive measures

Besides past year binge drinking (BD) as the main outcome variable, we also describe prevalence of lifetime and past month binge drinking. Three questions about patterns of the last binge drinking episode were also analyzed. The students were asked about where they were and with whom and also about what kind of alcoholic beverage they drunk. The possible responses are presented on table 1.

Table 1. Description of patterns of binge drinking among public and private high school students in the 27 Brazilian State Capitals (N=14714).

	N	wgt%	95% CI
		min	max
lifetime binge drinking	5129	34.9	33.1	36.7
binge drinking in the past year	4579	31.6	29.8	33.4
binge drinking in the past month	2913	20.9	19.4	22.4
once	961	6.3	5.6	7.1
twice	785	5.5	4.9	6.2
3 - 5 episodes	734	5.0	4.5	5.7
6 - 9 episodes	219	1.7	1.4	2.1
10 + episodes	214	1.8	1.5	2.2
Characterization of the last binge episodea
Where did it occur?
bar, nighthouse, pub	2936	25.0	23.2	26.9
friends' house	2591	21.5	20.0	23.0
relatives' house	864	9.4	8.4	10.4
own house	742	8.0	7.0	8.9
another placeb	753	8.2	7.1	9.4
With whom?
alone	189	2.3	1.8	2.8
friends	4553	33.4	31.4	35.4
parents or uncles	1049	11.0	10.0	12.1
brothers or cousins	1790	16.9	15.5	18.3
otherc	248	2.7	2.2	3.2
What type of alcoholic beverage did you drink?d
beer	3545	28.5	26.7	30.2
vodca	2630	21.8	20.1	23.7
alcopope	2902	24.9	23.2	26.7
wine	1744	19.5	17.9	21.3
cocktail	2024	19.9	18.1	21.8
whisky	1295	12.2	11.1	13.5

^arespondents could give more than one answer

^bmainly parties, birthday parties, parties in beaches.

^cmanily boyfriend/girlfriend, husband or wife

^dliquors with less than 10% of prevalence were not included on the table

^epre-mixed flavored spirits

^*missing data for the binge measures (lifetime, year and month) varied from 1% to 3%.

2.5 Analysis approach

Analyses were conducted on data weighted to correct for unequal probabilities of selection into the sample. The complex survey design took into account the city and type of school, the school as primary sampling unit, the expansion weight and the final probability of drawing the student who answered the questionnaire. The outcome variable of interest was binge drinking in the past year. The independent variables included SES, type of school, age and sex. We described students' general characteristics and the characteristics of those reporting recent (past year) binge drinking by weighted proportions and crude odds ratio from logistic regression. To estimate the association between SES and recent BD we did three type of weighted logistic regression stratifying for Brazilian regions: 1) high SES × all other together; 2) high SES × each other (separate) 3) low SES × each other. The first analysis (using SES as a binary variable – high SES × all other together) derived a funnel plot that shows adjusted odds ratio in the 5 Brazilian regions. All the analyses were controlled for age, sex and type of school.

Analyzes were performed using Stata Version 11, with svyset procedures to address variance estimation under the complex sample design in these regression models and in estimation of all 95% confidence intervals (CI). Results are presented via weighted proportions (wgt%), crude Odds Ratios (cORs), adjusted Odds Ratios (aORs) and 95% confidence interval.

3. Results

BD is a very common behavior among high school students in Brazilian state capitals. Table 1 provides a description of this behavior among the study sample. Almost 35% of the students engaged in BD at least one in their lifetime. This behavior is also common in the 30 days prior to the study (20.9% and the frequency varies from one to five times (from 6.3% to 5.0%) in the past month. In general, the last episode of BD occurred on pubs and nightclubs (25.0%) or at friends' houses (25.0%), with friends and by drinking beer, alcopop (spirits mixed with soda or industrialized juice) and vodka.

Table 2 describes the study sample in relation to past–year BD and the covariates under study. Most of the respondents were girls (56.3% [95%CI 55.1%-57.6 %]), with average age of 15.9 years (SD=0.3), attending public schools (75.6% [95%CI 71.9%-78.9%]) and living in the Southeast region of Brazil (42.7% [95%CI 38.5%-46.9%]). According to the ABEP index the majority of the students belonged to families with medium socioeconomic status (38.8% [95%CI 36.7%-40.9%] on the C stratum and 43.9% [95%CI 41.8%-45.8%] on the B stratum). Only 50 cases were found on the lowest stratum (E= 0.4 % [95%CI 0.3%-0.5%]). Being an older boy attending private school and belonging to the high socioeconomic SES was associated with past-year BD on the crude analysis.

Table 2. Descriptive weighted analysis of factors associated with recent binge drinking (past-year) among students in Brazilian high schools (n = 14518).

	Recent Binge Yes				Recent Binge No				Total				Bivariate
	N	wgt%	95% CI		N	wgt%	95% CI		N	wgt%	95% CI		crude OR	95% CI
gender
female	2218	49.4	47.1	51.5	5878	59.6	58.1	61.5	8096	56.3	55.1	57.6	ref
male	2352	50.6	48.4	52.8	4041	40.4	39.9	41.9	6393	43.7	42.4	44.9	1.51	1.35	1.68
age group
14-15	1203	25.5	21.7	29.7	4193	41.5	38.1	44.9	5396	36.4	33.3	39.6	ref
16-18	3376	74.5	70.3	78.2	5746	58.5	55.1	61.9	9122	63.6	60.4	66.7	2.06	1.71	2.49
type of school
private	2297	29.4	24.9	34.9	4180	22.1	19.1	25.3	6477	24.4	21.1	28.0	1.47	1.26	1.71
public	2282	70.6	65.6	75.1	5759	77.9	74.6	80.8	8041	75.6	71.9	78.9	ref
SESa
A	1231	18.4	14.7	22.8	1662	10.3	8.6	12.3	2893	12.9	10.6	15.6	ref
B	2010	45.1	41.9	48.3	4451	43.3	41.2	45.3	6461	43.9	41.8	45.8	0.58	0.49	0.70
C	1202	32.3	30.1	35.9	3368	41.5	39.4	43.7	4570	38.8	36.7	40.9	0.44	0.37	0.54
D	112	2.9	2.3	3.6	425	4.6	3.9	5.3	537	4.0	3.5	4.6	0.36	0.27	0.48
E	24	0.6	0.4	1.0	33	0.3	0.2	0.4	50	0.4	0.3	0.5	1.13	0.59	2.19
Brazilian Region
North Region	798	7.7	6.3	9.5	2329	12.4	10.3	14.7	3127	10.9	9.1	12.9	ref
Notheast Region	1593	23.7	20.7	26.9	3805	28.7	25.4	32.2	5398	27.0	24.2	30.2	1.32	1.12	1.55
Midwest Region	790	12.0	9.7	14.6	1536	11.8	10.5	13.8	2326	11.9	10.1	13.3	1.87	1.52	2.29
Southeast Region	801	47.4	42.6	52.1	1525	40.5	35.9	45.2	2326	42.7	38.5	46.9	2.24	1.85	2.71
South Region	597	9.2	7.6	11.2	744	6.6	5.2	8.3	1341	7.5	6.1	9.1	1.62	1.33	1.97

^aSES according to ABEP scale (described in the methods section)

^*Missing data for the outcome measure was 1.3%

In Brazil as a whole, being part of the most favored SES group increased by 84% (aOR=1.84; 95% CI 1:52 to 2:24) the odds of a student having engaged in BD in the past year as compared to the pool of students of all other SES combined. Figure 1 presents regional differences on the odds ratio estimates with covariate adjustment for sex, age and type of school. This association was significant across all regions, but with more strength in the poorest regions of the country (North: aOR = 2.29 [95% CI: 1.57- 3.36] and the Northeast: aOR = 2.04 [95% CI: 1.64-2.54]). In the richest region (Southeast), it shows the extreme minimum 95 % CI closer to the invalidation of association (95% CI: 1.12-2.38).

Figure 1. Funnel-plot for the adjusted odds ratio of high SES and recent binge drinking (past-year) in five Brazilian regions.

Table 3 presents the odds ratios for the control variables. Older age was the only significant covariable in the five regions, increasing from 40% to 54% the odds of belonging to the group who practiced BD.

Table 3.

Results of the adjusted logistic regression for recent binge drinking (past-year) associated with the highest SES among high school students in the 27 Brazilian capitals and in the five geo-economic regions (n = 14489).

Binge drinking in the past year		aOR	95% CI		p value
Brazil
	High SESa	1.84	1.52	2.24	< 0.001
	male	1.40	1.26	1.57	< 0.001
	age	1.47	1.38	1.55	< 0.001
	private school	1.39	1.18	1.62	< 0.001
North Region
	High SESa	2.29	1.57	3.36	< 0.001
	male	1.57	1.09	2.26	0.017
	age	1.45	1.25	1.68	< 0.001
	private school	1.02	0.80	1.30	0.894
Northeast Region
	High SESa	2.04	1.64	2.54	< 0.001
	male	1.40	1.19	1.64	< 0.001
	age	1.49	1.40	1.58	< 0.001
	private school	1.22	0.98	1.51	0.075
South Region
	High SESa	1.70	1.23	2.33	0.002
	male	1.24	0.93	1.65	0.140
	age	1.40	1.19	1.63	< 0.001
	private school	1.47	1.07	2.01	0.019
Southeast Region
	High SESa	1.63	1.12	2.38	0.011
	male	1.44	1.18	1.76	< 0.001
	age	1.54	1.38	1.72	< 0.001
	private school	1.64	1.20	2.23	0.002
Midwest Region
	High SESa	1.61	1.25	2.06	< 0.001
	male	1.18	0.90	1.54	0.227
	age	1.37	1.24	1.51	< 0.001
	private school	1.38	1.04	1.83	0.026

^aHigh SES = “A” group according to the ABEP scale. The reference group was all other social classes compared to class A.

Table 4 compares the different SES categories, using the higher SES as a reference for the first columns and lower SES as a reference in the last columns. In the analysis in which the reference encompasses the students of class A (the higher class), there is a clear inverse gradient of odds ratio controlled for sex, age and type of school in Brazil as a whole and in the five geo-economic regions in separate. Despite all the gradients being well defined in all regions, showing that those who belong to the higher SES are at greater risk than those who belong to lower classes, in the Midwest there is the greatest difference between the adjusted odds among different strata. In this region, the drop from one stratum to another decreases in 20% the odds of BD, while the poorest (low SES) are up to 70% (aOR = 0.34, 95% CI: 0.24-0.49) less likely to have done binge in the year than the wealthiest (high SES).

Table 4.

Results of the adjusted logistic regression for recent binge drinking (past-year) associated with four levels of SES among high school students in the 27 Brazilian State Capitals and according to the 5 Brazilian geo-economic region, considering the two extreme classes as reference (N = 14489).

Binge drinking in the past year		aOR	95% CI		p value	aOR	95% CI		p value	N
Brazil
	high SES	ref				2.79	2.18	3.58	< 0.001	2934
	medium-high SES	0.59	0.48	0.71	< 0.001	1.65	1.40	1.94	< 0.001	6553
	medium-low SES	0.49	0.39	0.63	< 0.001	1.39	1.17	1.65	< 0.001	2920
	low SES	0.36	0.28	0,46	< 0.001	ref				2307
North Region
	high SES	ref				2.73	1.52	4.91	0.001	744
	medium-high SES	0.48	0.33	0.70	< 0.001	1.31	0.80	2.15	0.279	1384
	medium-low SES	0.39	0.25	0.62	< 0.001	1.07	0.77	1.50	0.684	637
	low SES	0.37	0.20	0.65	0.001	ref				407
Northeast Region
	high SES	ref				2.84	2.15	3.75	< 0.001	921
	medium-high SES	0.52	0.41	0.65	< 0.001	1.47	1.18	1.82	0.001	2036
	medium-low SES	0.51	0.38	0.66	< 0.001	1.44	1.14	1.83	0.003	1227
	low SES	0.35	0.27	0.46	< 0.001	ref				1291
Midwest Region
	high SES	ref				2.89	2.01	4.15	< 0.001	474
	medium-high SES	0.70	0.54	0.90	0.006	2.02	1.42	2.86	< 0.001	1207
	medium-low SES	0.50	0.36	0.70	< 0.001	1.46	1.00	2.12	0.050	384
	low SES	0.34	0.24	0.49	< 0.001	ref				283
Southeast Region
	high SES	ref				2.34	1.39	3.94	0.002	422
	medium-high SES	0.64	0.44	0.94	0.022	1.50	1.06	2.13	0.022	1174
	medium-low SES	0.58	0.36	0.92	0.022	1.35	0.93	1.96	0.112	497
	low SES	0.42	0.25	0.71	0.002	ref				260
South Region
	high SES	ref				2.08	0.97	4.46	0.059	373
	medium-high SES	0.60	0.43	0.82	0.002	1.25	0.63	2.48	0.511	752
	medium-low SES	0.55	0.33	0.91	0.023	1.16	0.57	2.38	0.679	175
	low SES	0.48	0.22	1.02	0.059	ref				66

^aanalysis controled for age, sex and type of school

^bhigh SES was defined as the “A” at ABEP scale, Medium-high SES as the “B” at ABEP scale, medium low SES as the “C1” at the ABEP scale and Low SES as the C2, D and E at the ABEP scale.

The analysis using the poorest group as a reference shows that there is not always a risk gradient between a lower to the next higher classes, although it always exists between the extremes SES (low and high). Although the gradient is clear from the perspective of the risk level in Brazil as a whole (where the wealthy are almost three times more likely than poor to have engaged in BD - aOR = 2.79, 95% CI: 2.18 - 3.58), when we take the regions in particular, we note that in the Southeast region there is no difference between the two lower classes (medium-low SES and low SES) and in the North the only significance lies in the extreme.

4. Discussion

4.1 Recap of main findings

Three substantive findings emerge from these analyses: 1) BD is a highly prevalent risk behavior among high school students in the 27 Brazilian state capitals, mainly among older boys, those studying in private schools versus public schools and those that belong to the highest SES (class A as categorized by ABEP index); 2) considering stratified analyses for each one of five the Brazilian rmacroegions these result are maintained and it is possible to note that the poorest the region the strongest the association (regions as Northeast and North, with smaller PIBs, presented higher OR of binge drinking in the highest socioeconomic classes); 3) there is a very clear and consistent inverse gradient of association between SES and BD in all regions, considering the highest SES as reference; however, this gradient is not so clear when using the lowest SES as reference.

4.2 Selected limitations and offsetting strengths

Before detailed discussion of these findings, a few limitations should be mentioned. Due to the fact that a self-report questionnaire was used, the questions were subject to interpretation by the participants and to a possible information bias. However, the anonymous nature of the survey and the absence of the teacher in the classroom should help promote response validity. Also the question about a fictitious drug allowed us to drop the questionnaires with proved bias information. Some degree of non-participation (especially because of absence on the day of the survey) and data missingness excluded some students from the analysis. However the levels of participation were larger than those obtained in the US Monitoring the Future study (Johnston et al. 2009), considering that almost all the students that were invited to participated agreed with. Another possible limitation is the fact that we did not differentiate the number of drinks by sex similar to what is done in the ESPAD (Hibell et al. 2009) study.

4.3 Research issues

It is important to note that as it is a cross sectional survey, association does not imply causation. However, as we are dealing with especially high SES adolescents and this status is based parents income and education, it is hard to believe that this status would change during the five years that understand this period of life (adolescence). One must consider that there may be exceptions, however, in general SES is a stable variable for adolescents (Spijkerman et al. 2008) and would not be dependent on the practice of BD. In studies among adults, thinking about causality related to SES is much more complex, since the drug itself could be acting as a causal factor to the decline of social class, for example, and maybe not the cause of the drug use.

Although causal inference always requires extreme caution, we suggest that the data in this paper allows us, cautiously, to suggest that belonging to the higher SES classes in the capitals of any Brazilian macroregion is a risk for the practice of BD. This risk is most evident in the poorest regions (North and Northeast, with GDP per capita of US$ 6,385 and US$ 4,679) than in the richest (Southeast and Midwest, whose GDP per capita is almost double: US$ 12,732 and US$ 13,238 respectively, considering a parity of 1.6 reais to 1 US dollar and data from the IBGE - Brazilian Institute of Geography and Statistics, 2008; IBGE, 2011). It is noteworthy here that all comparisons of macro-economics and SES should be taken with caution, due to inequality of income distribution by countries, as suggested by Ritter & Chalmers (Ritter and Chalmers 2011). These researchers have shown evidence of direct association of alcohol consumption and the state of country's economy, suggesting that the pattern of use and daily amount used by a population vary in waves according to macro-economic variations.

Certainly, it is hard to define the process of causal relation (if it truly exists). However, we speculate that three main pathways may exist: 1) financial: the richest adolescents have more pocket money to spend in nightclubs, pubs and parties for the purchase of drinks and/or 2) socio-cultural: there could be family factors that distinguish parental monitoring and rules in families with higher income versus families with lower income levels, and/or 3) behavioral: the feeling of omnipotence against the risk of intoxication and feeling of superiority derived from the emotional weight of being rich in a mostly low to middle income country.

Regarding the financial issues, what could explain the increased risk of BD among those with SES in the poorest regions would be the question of parity of purchasing power. In the poorest region, a person in a high class has greater purchasing power than a person with the same income (and also classified as high SES) in the richest region, for example, which indirectly makes them even more rich compared to the rest of the population living in that specific region. This is consistent to the findings of Bellis et al (2007), among British adolescents, which found that adolescents with more spending money were more likely to drink frequently, binge drink and to drink in public, as well as in a study of Martin et al (2009) in the US, that found that college students with lower levels of spending money had lower levels of drinking and getting drunk.

It is noticeable that the high SES group is the only one that always presents itself at a greater risk than any other SES group, suggesting that the increased pocket money is not the only thing behind this association (otherwise, when we take the lowest class as reference, the gradient should be kept in all regions of Brazil), but there should be other socio-cultural and behavioral aspects that are shared by the wealthy class such as weaker parental supervision or felling of omnipotence or intangibility.

In the same direction, two Finnish studies, not so recent, have discussed the issue of pocket money and suggested that the availability of money is at least to some degree related to parent's affluence and have a clear relationship with drunkenness among adolescents (Kouvonen and Lintonen 2002; Lintonen et al. 2000). However, the relationship between SES and drunkenness among adolescents is not so clear in other 27 countries analyzed by the Health Behavior in School-aged Children study (HBSC). In this comparative study, in less than 20% of European countries there was evidence of increased risk associated with the higher classes. According to the authors, the regional differences of the association between SES and alcohol drunkenness depend on the alcohol culture of each country (Richter et al. 2006), which points to the need for this analysis for each individual country and highlights the relevance of the present study. Goodman and Huang (Goodman and Huang 2002) offer a provocative hypothesis that a SES gradient in health behavior is well established, with higher SES linked to better health. To the extent that BD is a manifestation of unhealthy behavior, the evidence from Brazil seems to contradict this hypothesis and goes deeply against the assertion that the SES gradient and health behavior is very well established.

In this field of research that shows controversial results about association of higher SES or income inequality on adolescent drinking, the fact that this analysis fails to explain the process and why this phenomenon occurs do not invalidated the inverse gradient found, but points to the need of more probing research, with a mixed methods approach combining quantitative survey analysis with qualitative interviewing and social network analysis can help clarify the dynamic processes that produce the observed SES gradient (Tashakkori and Teddlie 2010).

4.5 Conclusions

In summary, we set out to discover whether there would be a gradient of association of SES and BD among high school students in Brazil. However, contrary to what has been observed in developed countries, students living in Brazilian capitals may be at an increased risk of BD when they belong to the highest socioeconomic status in all regions. This association is stronger in the poorest regions of the country and a clear inverse OR gradient was detected. Older boys in private schools are in more risk for this behavior. Aspects of the local alcohol culture and weight to be rich in a highly unequal country may be influencing this result. Adolescents in other emerging economies might have the same association between high SES and BD.