The Influence of Family Income Trajectories From Birth to Adulthood on Adult Oral Health: Findings From the 1982 Pelotas Birth Cohort

Abstract

Objectives. We assessed whether 3 models of life course socioeconomic status (critical period, accumulation of risk, and social mobility) predicted unsound teeth in adulthood among a Brazilian cohort.

Methods. Life course data were collected on the 5914 live-born infants in the 1982 Pelotas Birth Cohort study. Participants' oral health was assessed at 15 (n = 888) and 24 (n = 720) years of age. We assessed family income trajectories and number of episodes of poverty in the life course through Poisson regressions, yielding unadjusted and adjusted prevalence ratios for number of unsound teeth at age 24 years.

Results. The adjusted prevalence ratio for participants born into poverty was 30% higher than for those who were not. Participants who were always poor had the highest prevalence of unsound teeth; those who were downwardly or upwardly mobile also had more unsound teeth than did other participants, after adjustment for confounders. More episodes of poverty were associated with greater prevalence of unsound teeth in adulthood.

Conclusions. Poverty at birth and during the life course was correlated with the number of unsound teeth at 24 years of age.

The relationship between adults' socioeconomic position and their health is well known. However, the majority of studies addressing this issue have used measurements of adulthood socioeconomic position or relied on adults' retrospective reports about their childhood.¹

Adult health may be affected by socioeconomic position during different periods in the life course, and at least 3 major theories have been proposed to explain how and when life course socioeconomic factors influence adult health. One theory proposes that during a critical period of development in early life, exposures to deprivation have long-term effects on adult health, independent of adult circumstances.² Galobardes et al. updated a systematic review of the association between childhood socioeconomic conditions and cause-specific mortality; they confirmed that mortality risk for all causes was higher among those who experienced poorer socioeconomic status (SES) during childhood, although not all causes of death were equally related to childhood socioeconomic circumstances.³

Others theorize that the intensity and duration of exposure to unfavorable or favorable physical and social environments throughout life affect health status in a dose–response relationship; this has been termed the accumulation-of-risk hypothesis.⁴ For example, the number of episodes of being in the manual social class (a cumulative harmful exposure) measured at 3 life stages was strongly and positively associated with mortality from cardiovascular disease among Scottish men.⁵

A third theory, the social mobility hypothesis, postulates that the importance of the early life environment lies in its effect on the socioeconomic trajectories of individuals. Circumstances in early life are identified as the first step in the pathway to adult health, but with an indirect effect, influencing adult health through social mechanisms such as restricting educational opportunities, thus shaping socioeconomic circumstances and health in later life.⁶ In a New Zealander birth cohort, Poulton et al. investigated the association between socioeconomic trajectories during the life course and aspects of health in adulthood; they found that upward mobility did not mitigate or reverse the adverse effects of low childhood SES on adult health.¹

Globally, the burden of common oral conditions is high: one of the most common chronic diseases worldwide is dental caries, severe periodontitis affects between 5% and 15% of most populations, and oral cancer is the eighth most common cancer worldwide.⁷ This evidence led the World Health Assembly to call for oral health to be integrated into chronic disease prevention programs.⁸

Despite substantial evidence showing that SES is strongly associated with oral health,^9,10 the dynamics of how SES over time affects adults' oral health remain unclear. We assessed whether 3 hypotheses about life course SES (critical period, accumulation of risk, and social mobility) predicted an important oral health outcome in early adulthood.

METHODS

In Pelotas, Brazil, a medium-sized city in the south of Brazil, a water fluoridation scheme had been implemented in 1962 that covered almost all the urban population. In 1982, a population-based birth cohort study was initiated in Pelotas. The study began as a perinatal health survey that enrolled all live infants born at 3 maternity hospitals. The 5914 infants were weighed and the mothers were weighed and measured; the mothers also completed a questionnaire shortly after their infant's birth. A wide range of data were collected, including socioeconomic, demographic, and health-related information.¹¹

The cohort was followed up several times thereafter, and further information on those assessments is available elsewhere.¹¹ In 1997, when the mean age of the cohort was 15 years, a systematic sample of 27% of the city's census tracts was selected. Every household in these tracts was visited, and 1076 cohort members were interviewed. Of these, 900 were randomly selected for the Oral Health Study; 888 (98.7%) of those selected took part.¹² In 2006, when they were aged 24 years, the 888 cohort members assessed in the first Oral Health Study were invited to have dental examinations and to be interviewed again. Of these participants, 720 (81.1%) took part.

Sociodemographic Variables

Income.

Information on family income was collected at the time of the cohort participants' birth and again when they were aged 15 and 23 years. Monthly family income was collected in a face-to-face questionnaire and then categorized according to Brazilian minimum wage units into 5 categories: less than 1 minimum wage unit (21.9% of the entire cohort), 1.1 to 3 units (47.4%), 3.1 to 6 (18.5%), 6.1 to 10 units (6.5%), and more than 10 units (5.7%). One minimum wage unit at that time was equivalent to US$ 50 per month.

Unfortunately, information on the continuous level of income was unavailable, because income information was collected into 5 precodified categories. We intended to classify family income into tertiles to allow the study of change in income levels since childhood. Because of the unequal numbers of participants in each category, we carried out a principal components analysis with 4 variables: health services payment mode (out of pocket, public free, or private health insurance), maternal education, height, and skin color. We derived a score from the first component that we used to rank individuals within family income groups. We then found cutoff points within each category to form 3 nearly equal-sized groups. To build the tertile-equivalent groups, we added the 1288 individuals in the lowest family income category to the 675 poorest individuals in the second category. The next 1979 individuals in this second category formed the second tertile, and all the remaining individuals formed the last tertile.

The monthly income of each member of the family was also collected in 1997 and 2005, with income as a continuous variable. We generated tertiles from the sum of the reported values.^12,13 We categorized the combined middle and higher tertiles of family income as nonpoor and the lower tertile as poor. We compared the poorest tertile with the other 2 tertiles because health inequalities in Brazil follow a bottom inequity pattern¹⁴; that is, the middle and upper classes are reasonably similar, but the poor lag well behind.

Family income trajectory.

We performed group-based trajectory analysis to estimate family income trajectory groups with the PROC TRAJ macro in SAS version 9.1 (SAS Institute Inc, Cary, NC).^15,16 We fitted the trajectory model with the logit distribution, because of the nature of the family income data available (whether respondents were poor or not poor at a given age). We determined the parameters for the trajectory model by maximum likelihood, and we used the Bayesian information criterion to help identify the correct number of groups for the family income trajectory model.

Cumulative poverty.

We calculated experience of poverty in the life course through a simple count of the number of episodes of poverty (yes or no) in each of the 3 waves of the cohort study (birth, age 15 years, and age 23 years); the resultant variable ranged from 0 to 3 episodes.

Education.

Maternal education level (number of years of schooling) was collected at the time of each participant's birth. Cohort members' educational attainment was assessed in the same way at age 23 years.

Self-reported skin color.

When they were aged 15 years, cohort members were required to classify their color or race according to the following precoded categories: White, lighter-skinned Black (pardo), dark-skinned Black (pretos), yellow (Asiatic ancestry), or indigenous.

Mediators and confounders.

We also analyzed information on some potential mediators and confounders that were collected, such as low (< 2500 g) or adequate (≥ 2500 g) birth weight, current smoking status at age 24 years (yes or no), use of dental services in the previous year at age 15 and 24 years (yes or no); receipt of dental hygiene advice in the most recent dental visit at age 15 years (yes or no); check-up as reason for the most recent dental visit at age 24 years (yes or no); and percentage of teeth with dental calculus at age 24 years, used here as a proxy for toothbrushing effectiveness.

Adult Dental Assessment

The dental examination of participants aged 24 years included collecting information on dental caries, gingival bleeding, dental calculus, periodontal pockets, the use of and need for a dental prosthesis, the quality of dental restorations, and soft tissue lesions (in that order). For our analysis, we used the number of sound teeth at age 24 years as the outcome, defined as the total number of sound untreated teeth and filled teeth without dental caries. A filled tooth with dental caries or a decayed or missing tooth was considered an unsound tooth.

Dental caries assessments followed diagnostic criteria promulgated by the World Health Organization in 1997.¹⁷ A full mouth examination was performed to assess dental calculus. Each tooth was assessed at 6 sites (mesiobuccal, midbuccal, distobuccal, distolingual, midlingual, and mesiolingual). Dental examinations were performed in the homes of the participants under artificial illumination (headlamp), with plane dental mirrors, Community Periodontal Index probes (Trinity, Campo Mourão, Brazil), and wooden spatulas to assist in retracting the tongue and cheeks for better visualization. Dental calculus was considered to be present if it was detected in at least 1 site examined.

Dental examinations were performed by 10 dentists, and questionnaires were administered by 7 interviewers. All examiners and interviewers were trained and standardized by methods described elsewhere.¹⁸ Examiner reliability statistics were calculated with a weighted and simple κ test and the intraclass correlation coefficient when appropriate. The lowest value was 0.60 for use of and need for a dental prosthesis; most of the other values were in the range of 0.80 to 1.00. For data quality control, 10% of the interviews were repeated with a shorter version of the questionnaire to assess the degree of concordance (reliability).

Statistical Analysis

After we computed univariate statistics, we cross-tabulated the 4 family income trajectories from birth to adulthood against birth weight and the demographic and educational characteristics of the cohort participants. We then examined the association between the 4 family income trajectories and the oral health variables. We then conducted similar cross-tabulations after substituting the number of poverty episodes in the life course for the socioeconomic trajectories. For comparisons among groups, we used the Pearson χ² test (as well as linear trends when appropriate). We considered P values of less than .05 to be statistically significant.

We then tested the 3 hypotheses about life course socioeconomic influences on adult health (critical period, social mobility, and accumulation of risk), with the categorized number of unsound teeth at age 24 years as the outcome of interest. The number of unsound teeth (originally defined as the proportion of unsound teeth among all erupted teeth present in the mouth) was transformed into a binary variable through a median split.

To test the critical period hypothesis, we grouped the always poor and upwardly mobile cohort members into a new category, poor at birth, and then compared it with those who were not poor at birth (the combined never poor and downwardly mobile groups). We used multivariable Poisson regression models with robust variance to obtain prevalence ratios and 95% confidence intervals, with participants who were not poor at birth as the reference group. All explanatory variables with a P value of .25 or less were selected to be entered into the multivariable models.

We tested the social mobility hypothesis by comparing prevalence ratios for the downwardly mobile, upwardly mobile, and always poor groups; the never poor were the reference group. Finally, we examined the accumulation of risk hypothesis by comparing groups according to their number of episodes of poverty in the life course (1, 2, or 3); respondents with no episodes of poverty were the reference group. We used multivariable Poisson regression models with robust variance. For all analyses, we used Stata version 9.0 (StataCorp LP, College Station, TX).

RESULTS

We identified 4 family income trajectories among the 720 participants: never poor (46.1%), downwardly mobile (18.2%), upwardly mobile (13.1%), and always poor (22.6%). Almost half of the sample had never experienced poverty; nearly a quarter had lived all of their lives in poverty; and the proportions experiencing social mobility (downward or upward) were reasonably similar, at 18% and 13%, respectively (Table 1). Family income trajectories were not associated with cohort members' birth weight; they were weakly associated with gender and strongly associated with skin color, maternal education, and participant education. Women, Blacks and those with low educational attainment predominated in the always poor group. Men, Whites, Asiatic descendants, and those with high levels of educational attainment were more likely to be in the never poor group.

TABLE 1.

Sample Birthweight, Demographic, and Educational Characteristics, by Family Socioeconomic Trajectories From Birth To Adulthood: 1982 Pelotas Birth Cohort, Brazil

	All (n = 720), %	Never Poor (n = 332), %	Downwardly Mobile (n = 131), %	Upwardly Mobile (n = 94), %	Always Poor (n = 163), %	Pa
Total	100.0	46.1	18.2	13.1	22.6
Gender (n = 720)						.05
Male	52.6	49.6	18.0	13.7	18.7
Female	47.4	42.2	18.5	12.3	27.0
Birthweight (n = 720)						.438
Lowb	5.1	40.6	27.0	8.1	24.3
Adequatec	94.9	46.4	17.7	13.3	22.6
Self-reported skin color (n = 719)						< .001
White and Asiatic	77.7	53.0	19.1	9.6	18.3
Lighter-skinned Black	7.7	30.9	20.0	20.0	29.1
Dark-skinned Black	14.6	17.1	12.4	27.6	42.9
Maternal education (n = 718)						< .001
≥ 12 y	12.3	89.8	5.7	0.0	4.5
9–11 y	10.4	67.7	17.3	1.3	14.7
5–8 y	44.7	49.8	23.1	10.9	16.2
≤ 4 y	32.6	18.4	16.2	24.8	40.6
Participant education (n = 696)						< .001
≥ 12 y	14.5	85.1	8.9	4.0	2.0
9–11 y	52.8	50.9	19.6	14.2	15.3
5–8 y	26.7	22.0	22.6	15.6	39.8
≤ 4 y	6.0	9.5	11.9	14.3	64.3

^aχ² test or χ² for trend.

^bLess than 2500 grams.

^c2500 grams or more.

Those in the never poor group had better indicators of dental attendance and were more likely to have received dental hygiene advice during the dental visit in the year before the assessment at age 15 years. These participants were most likely to have had preventive dental care, had the most sound teeth, and were least likely to smoke and to have teeth with calculus. The worst profile for these indicators was observed in the always poor group. The downwardly mobile and upwardly mobile groups showed an intermediate pattern (Table 2).

TABLE 2.

Selected Oral Health–Related Variables of the Cohort Sample at Age 24 Years, by Family Socioeconomic Trajectories From Birth to Adulthood: 1982 Pelotas Birth Cohort, Brazil

	All (n = 720), % or Mean (Median)	Never Poor (n = 332), % or Mean (Median)	Downwardly Mobile (n = 131), % or Mean (Median)	Upwardly Mobile (n = 94), % or Mean (Median)	Always Poor (n = 163), % or Mean (Median)	Pa
Total	100.0	46.1	18.2	13.1	22.6
Visited dentist in past year at age 15 y (n = 710)	53.0	66.2	51.9	45.7	31.3	< .01
Received dental hygiene advice in the most recent dental visit at age 15 y (n = 710)	90.9	94.8	88.6	90.4	85.0	.04
Visited dentist in past year at age 24 y (n = 691)	55.6	60.7	61.2	51.7	41.9	.01
Checkup as reason for most recent dental visit at age 24 y (n = 720)	22.1	29.5	19.1	24.5	8.0	< .01
Current smoker at age 24 y (n = 519)	26.6	20.1	29.4	26.9	35.7	.01
Sound teeth (n = 720)	25.0 (26)	25.3 (27)	25.1 (26)	24.9 (26)	24.3 (25)	< .01
Sound teeth below median (n = 720)	40.6	32.2	45.0	43.6	52.2	< .01
Teeth with dental calculus (n = 720)	28.6	28.7	27.8	25.5	31.0	.49

^aχ² test or χ² for trend.

Across the entire cohort, 45.3% had had no episodes of poverty during the life course, 25.5% had 1 episode, 17.6% had 2 episodes, and 11.8% had 3. Data on birth weight and demographic and educational characteristics of the cohort are presented in Table 3 by the number of episodes of poverty. We found very strong associations between the number of episodes of poverty and cohort members' skin color, education level, and maternal education level. The group that had never experienced poverty contained the highest proportion of Whites and Asiatic descendants, persons with high levels of educational attainment, and persons whose mothers were highly educated.

TABLE 3.

Sample Birthweight, Demographic, and Educational Characteristics, by Number of Episodes of Poverty From Birth to Adulthood: 1982 Pelotas Birth Cohort, Brazil

		Episodes of Poverty
	All (n = 686), %	None (n = 311), %	1 (n = 173), %	2 (n = 121), %	3 (n = 81), %	Pa
Total	100.0	45.3	25.2	17.6	11.8
Gender (n = 686)						.086
Male	51.6	49.2	24.0	17.5	9.3
Female	48.4	41.3	26.5	17.8	14.4
Birthweight (n = 686)						.363
Lowb	5.3	38.9	36.1	11.1	13.9
Adequatec	94.7	45.7	24.6	18.0	11.7
Self-reported skin color (n = 686)						< .001
White and Asiatic	78.0	52.3	25.4	13.3	9.0
Lighter-skinned Black	7.3	28.0	30.0	20.0	22.0
Dark-skinned Black	14.7	16.8	21.8	39.6	21.8
Maternal education (n = 685)						< .001
≥ 12 y	11.8	88.9	6.2	4.9	0.0
9–11 y	10.8	66.2	17.6	16.2	0.0
5–8 y	44.8	49.2	30.0	15.3	5.5
≤ 4 y	32.6	17.5	28.2	26.0	28.3
Participant education (n = 686)						< .001
≥ 12 y	14.6	85.0	13.0	2.0	0.0
9–11 y	52.5	50.3	28.1	15.0	6.6
5–8 y	27.0	22.2	28.1	28.6	21.1
≤ 4 y	5.9	9.7	17.1	29.3	43.9

^aχ² test or χ² for trend.

^bLess than 2500 grams.

^c2500 grams or more.

We also found a clear gradient in oral health across the number of episodes of poverty (Table 4), with higher proportions of smokers and adults with fewer sound teeth among the groups with greater experience of poverty in the life course. On all indicators, respondents who had experienced 3 episodes of poverty had the worst oral health, and those who had not experienced poverty had the best oral health and most favorable associated behaviors.

TABLE 4.

Selected Oral Health–Related Variables of the Cohort Sample at Age 24 Years, by Number of Episodes of Poverty From Birth to Adulthood: 1982 Pelotas Birth Cohort, Brazil

		Episodes of Poverty
	All (n = 686), % or Mean (Median)	None (n = 311), % or Mean (Median)	1 (n = 173), % or Mean (Median)	2 (n = 121), % or Mean (Median)	3 (n = 81), % or Mean (Median)	Pa
Total	100.0	45.3	25.2	17.6	11.8
Visited dentist in past year at age 15 y (n = 686)	53.1	65.9	53.2	38.8	24.7	< .01
Received dental hygiene advice in the most recent dental visit at age 15 y (n = 686)	90.8	94.9	89.0	89.3	81.5	< .01
Visited dentist in past year at age 24 y (n = 659)	56.2	60.7	59.1	50.9	38.0	< .01
Checkup as reason for most recent dental visit at age 24 y (n = 686)	23.0	31.2	22.5	15.7	3.7	< .01
Current smoker at age 24 y (n = 496)	27.0	20.3	28.2	36.2	33.9	.02
Sound teeth (n = 686)	25.0 (26)	25.4 (27)	25.1 (26)	24.8 (26)	23.9 (25)	< .01
Sound teeth below median (n = 686)	40.7	32.5	43.9	45.5	58.0	< .01
Teeth with dental calculus (n = 686)	28.6	28.6	26.8	27.8	31.9	.43

^aχ² test or χ² for trend.

The outcomes of the multivariable models are presented in Table 5, which shows both unadjusted and adjusted estimates of associations with the categorized number of unsound teeth (median split, higher number = 1; lower number = 0) for the critical period, social mobility, and the accumulation-of-risk hypotheses.

TABLE 5.

Poisson Regression Analysis of Associations Between Unsound Teeth at Age 24 Years and Poverty at Birth, Social Mobility, and Number of Episodes of Poverty From Birth to Adulthood: 1982 Pelotas Birth Cohort, Brazil

		Social Mobility Hypothesisb			Accumulation-of-Risk Hypothesisc
	Critical Period Hypothesisa	Downward Mobility, PR (95% CI)	Upward Mobility, PR (95% CI)	Always Poor, PR (95% CI)	1 Episode of Poverty, PR (95% CI)	2 Episodes of Poverty, PR (95% CI)	3 Episodes of Poverty, PR (95% CI)
Model 1	1.4** (1.1, 1.6)	1.4** (1.1, 1.8)	1.4** (1.0, 1.8)	1.6** (1.3, 2.0)	1.4** (1.1, 1.7)	1.4** (1.1, 1.8)	1.8** (1.4, 2.3)
Model 2	1.4** (1.2, 1.6)	1.4** (1.1, 1.8)	1.4** (1.0, 1.8)	1.7** (1.3, 2.1)	1.4** (1.1, 1.7)	1.4** (1.1, 1.8)	1.8** (1.4, 2.4)
Model 3	1.4** (1.1, 1.6)	1.4** (1.1, 1.8)	1.4** (1.0, 1.8)	1.6** (1.3, 2.1)	1.4** (1.1, 1.7)	1.4** (1.1, 1.8)	1.8** (1.4, 2.3)
Model 4	1.4** (1.1, 1.6)	1.4** (1.1, 1.8)	1.4** (1.0, 1.8)	1.6** (1.3, 2.1)	1.4** (1.1, 1.7)	1.4** (1.1, 1.8)	1.8** (1.4, 2.3)
Model 5	1.3* (1.1, 1.6)	1.4** (1.1, 1.8)	1.3** (1.0, 1.8)	1.6** (1.3, 2.0)	1.3** (1.1, 1.7)	1.4** (1.1, 1.8)	1.8** (1.4, 2.3)

Note. PR = prevalence ratio; CI = confidence interval.

^aModel 1, poor at birth; model 2, model 1 plus gender; model 3, model 2 plus checkups; model 4, model 3 plus dental calculus; model 5, model 4 plus skin color. The reference group was not poor at birth.

^bModel 1, mobile groups; model 2, model 1 plus gender; model 3, model 2 plus checkups; model 4, model 3 plus dental calculus; model 5, model 4 plus skin color. The reference group was never poor.

^cModel 1, no episodes of poverty; model 2, model 1 plus gender; model 3, model 2 plus checkups; model 4, model 3 plus dental calculus; model 5, model 4 plus skin color. The reference group was never poor.

*P = .02; **P < .01.

For the critical period hypothesis, the unadjusted prevalence ratio for cohort participants who were born in poverty was 40% higher than for those who did not experience poverty at birth; after controlling for confounders and mediators (such as gender, dental visiting pattern, dental calculus, and skin color), the difference was 30% (Table 5). For the critical period hypothesis, not being in the never poor group was associated with the number of unsound teeth. The always poor group had the highest prevalence ratio, but the downwardly mobile and upwardly mobile groups also had higher numbers of unsound teeth, regardless of gender, pattern of dental attendance, dental calculus, or skin color. We observed a similar pattern for the accumulation-of-risk hypothesis. More episodes of poverty correlated with higher prevalence ratios for unsound teeth.

DISCUSSION

Our findings support the critical period, social mobility, and accumulation-of-risk hypotheses in explaining the prevalence of unsound teeth among young adults with different life course experiences of poverty. Individuals who experienced poverty around the time of their birth had high proportions of unsound teeth as adults, regardless of their family income in adolescence and adulthood. Social mobility from birth to adulthood influenced the proportion of unsound teeth in adulthood. The always poor group had the highest proportion of unsound teeth. The groups who experienced income mobility in either direction had outcomes similar to each other's but worse than those of the never poor group, who had the lowest proportion of unsound teeth. Cumulative exposure to poverty in the life course was strongly and positively associated with the number of unsound teeth in a dose–response relationship, showing the adverse effect for dental health of cumulative episodes of poverty.

Our findings on social origins and oral health are similar to those reported by Poulton et al.¹ and Thomson et al.¹⁹ on the oral health of the Dunedin birth cohort study. Those studies' results for dental caries, tooth loss, and periodontal disease experience in participants aged 26 years also supported the social origins and social mobility hypotheses (albeit with some minor differences—e.g., the upwardly mobile group had better oral health than did the downwardly mobile group at 26 years of age), although they did not examine the critical period hypothesis.

The explanations for these findings are related to patterns of both dental care and preventive measures. The number of sound teeth is influenced by access to dental services, fluoride exposure, and the rational use of sugar in diet.⁷ Pelotas has a fluoridated water supply system (implemented in 1962) benefiting almost the entire population²⁰; consequently, differences in fluoridated water consumption cannot explain our findings. On the other hand, it is well established that the diets of people living in unfavorable socioeconomic conditions are less healthy than are those of their better-off counterparts. The consumption of carbohydrates, sugar, and soft drinks is greater in populations with low SES and educational attainment.^21,22

Individuals experiencing events of poverty in early life probably have less access to (and use of) a variety of oral hygiene items and may be more likely to develop harmful oral health behaviors later in life. These experiences might predispose individuals to dental caries.²³ The pattern of dental hygiene is poor among groups with low SES, as demonstrated in the dental assessment of the 1982 Pelotas birth cohort at age 15 years¹²; poor hygiene may indicate less frequent toothbrushing and consequently less exposure to fluoride in toothpaste.

Adolescents who were born and grew up in poverty—the always poor group—had poorer dental health (on average) and a worse profile of toothbrushing habits than did their better-off peers. In the dental assessment at age 15 years, the most important difference we found when we compared behavioral factors among the 4 groups was that the always poor group had a lower frequency of toothbrushing, especially among girls. Lower frequency of toothbrushing implies lower exposure to topical fluoride in toothpaste and poorer plaque removal, which may explain the higher levels of dental caries in the always poor group. In addition, this group had less frequent dental care visits than did their better-off counterparts, which may explain their large number of untreated dental cavitied caries (which persisted until adulthood).

Strengths and Limitations

Few population-based birth cohort studies collect data on oral health. The Pelotas study assembled a random representative sample of births in the city in 1982. The study had a very high participation rate after 9 years from the previous dental assessment, all dental indices used followed international standards, and the observers showed high reliability in diagnosis and were blinded for the hypotheses under study. Consequently, selection, recall, or classification bias is unlikely to have occurred, which reinforces the internal validity of the study. However, generalizing from these findings must be undertaken with caution, because the study was conducted in a relatively affluent area of Brazil.

By contrast to most research focused on SES, we used a household SES measure based on income rather than on occupation or education. A variety of data sets have established that household income is a powerful correlate of mortality, that the beneficial effect of income on mortality appears larger at lower levels of income than at higher levels, and that the strength of the association between income and mortality has increased over the past 30 years. Recent longitudinal studies of the linkage between household income and health and development show very strong associations. These findings suggest that family income is perhaps the most powerful component of SES in its linkage with health.²⁴

Unfortunately, sample size constraints precluded a full examination of possible gender differences in how the various models operated. Previous studies have shown consistent differences between boys and girls in their dental behaviors and patterns of dental attendance, with girls generally having more favorable behaviors than do boys.^25,26 On the other hand, in Brazilian society at least, women tend to earn less money than do men and therefore have less purchasing power for acquiring goods and services.

In addition to assessing oral hygiene (for which dental calculus was the indicator) and patterns of dental attendance, we intended to control for infant health (with infant birth weight as a marker¹) and for adolescent health behavior (with smoking as the marker), but we did not enter these variables into the models because they were not associated with the outcome in the bivariate analysis.

From a materialist point of view—little emphasized in the dental literature—income determines behavior, psychosocial states, and lifestyle, including access to and use of dental services, self-care (such as toothbrushing), and diet. Although increasing income is likely to produce better health status, it is important to remember that differences in health and material conditions exist across all levels of income.²⁷

Conclusions

Poverty at birth and the number of episodes of poverty during the life course were associated with the number of unsound teeth at 24 years of age. The rationale for social epidemiology is monitoring inequalities in health, explaining the etiology of such inequalities, and providing relevant evidence to mobilize society to eliminate such inequalities.²⁸ This goal is related to general policy, but specific public health strategies must be implemented to minimize the effect of social inequalities on dental health. Because adult oral health is influenced by childhood social conditions—and because oral health and general health share some common risk factors—it is reasonable to conclude that efforts aimed at ameliorating common risk factors at the population level may be a more effective approach to promoting oral health than are oral disease–oriented programs.²⁹