Methods The study included all cases of Patau, Edwards and Down syndrome, delivered during 1985–2003 and resident in the former Northern Region of England, including terminations of pregnancy for fetal anomaly. We applied the K-function test for space-time clustering with fixed thresholds of close in space and time using residential addresses at time of delivery. The Knox test was used to indicate the range over which the clustering effect occurred. Tests were repeated using nearest neighbour (NN) thresholds to adjust for variable population density.

Results The study analysed 116 cases of Patau syndrome, 240 cases of Edwards syndrome and 1084 cases of Down syndrome. There was evidence of space-time clustering for Down syndrome (fixed threshold of close in space: P = 0.01, NN threshold: P = 0.02), but little or no clustering for Patau (P = 0.57, P = 0.19) or Edwards (P = 0.37, P = 0.06) syndromes. Clustering of Down syndrome was associated with cases from more densely populated areas and evidence of clustering persisted when cases were restricted to maternal age <40 years.

Conclusions The highly novel space-time clustering for Down syndrome suggests an aetiological role for transient environmental factors, such as infections.

Methods Three estimators termed direct, standard IV and adjusted IV, of the phenotype-disease log odds ratio are compared through a simulation study which incorporates unmeasured confounding. The simulations are verified using formulae relating marginal and conditional estimates given in the Appendix.

Results The simulations show that the direct estimator is biased by unmeasured confounding factors and the standard IV estimator is attenuated towards the null. Under most circumstances the adjusted IV estimator has the smallest bias, although it has inflated type I error when the unmeasured confounders have a large effect.

Conclusions In a Mendelian randomization study with a binary disease outcome the bias associated with estimating the phenotype-disease log odds ratio may be of practical importance and so estimates should be subject to a sensitivity analysis against different amounts of hypothesized confounding.

Methods We used a simulated population with a known marginal treatment effect and applied a propensity score method and logistic regression analysis to adjust for confounding.

Results The adjusted treatment effect in logistic regression is in general further away from the true marginal treatment effect than the adjusted effect in propensity score methods. The difference is systematic and dependent on the incidence proportion, the number of prognostic factors and the magnitude of the treatment effect. For instance, a substantial difference of 20% is found when the treatment effect is 2.0, the incidence proportion is 0.20 and there are more than 11 prognostic factors.

Conclusions Propensity score methods give in general treatment effect estimates that are closer to the true marginal treatment effect than a logistic regression model in which all confounders are modelled.

Methods We use repeated cross-sectional data from the U.S. National Health and Nutrition Examination Surveys (NHANES), a nationally representative sample of over 14 000 participants ages 25 and older. We calculate pre-fortification (1991–94) and post-fortification (1999–2002) absolute differences and relative prevalence ratios of low red blood cell (RBC) folate status (<362.6 nmol), by race/ethnicity and income quartile. We also estimate kernel density plots and relative and absolute concentration curves pre- and post-fortification.

Results The excess prevalence of low RBC folate status associated with the lowest income quartile and black race declined by 67% and 48%, respectively, following fortification. Despite these absolute gains, the relative ratio of low folate status increased after fortification for the lowest compared with the highest income groups (from 1.27 to 2.08) and among whites compared with blacks (from 1.64 to 3.75).

Conclusions The effects of the fortification policy highlight the importance of distinguishing absolute from relative differences when evaluating interventions to reduce health disparities. Targeting of high risk populations is likely needed to eliminate remaining folate disparities.

Methods We developed and implemented sequential and combinatorial algorithms that evaluate the change in between-study heterogeneity as one or more studies are excluded from the calculations. The algorithms exclude studies aiming to achieve either the maximum or the minimum final I² below a desired pre-set threshold. We applied these algorithms in databases of meta-analyses of binary outcome and ≥4 studies from Cochrane Database of Systematic Reviews (Issue 4, 2005, n = 1011) and meta-analyses of genetic associations (n = 50). Two I² thresholds were used (50% and 25%).

Results Both algorithms have succeeded in achieving the pre-specified final I² thresholds. Differences in the number of excluded studies varied from 0% to 6% depending on the database and the heterogeneity threshold, while it was common to exclude different specific studies. Among meta-analyses with initial I² > 50%, in the large majority [19 (90.5%) and 208 (85.9%) in genetic and Cochrane meta-analyses, respectively] exclusion of one or two studies sufficed to decrease I² < 50%. Similarly, among meta-analyses with initial I² > 25%, in most cases [16 (57.1%) and 382 (81.3%), respectively) exclusion of one or two studies sufficed to decrease heterogeneity even <25%. The number of excluded studies correlated modestly with initial estimated I² (correlation coefficients 0.52–0.68 depending on algorithm used).

Conclusions The proposed algorithms can be routinely applied in meta-analyses as standardized sensitivity analyses for heterogeneity. Caution is needed evaluating post hoc which specific studies are responsible for the heterogeneity.

Methods A cross-sectional study was done in Porto Alegre, Brazil, in 2003, with a sample representative of families covered by the Family Health Program. Health expenses were recorded with reference to 30 days prior to the interview and income data were collected with reference to the previous calendar month. Health expenditure was explored in terms of total household health expenditure >5, 10 and 20% of household income and >40% households capacity to pay.

Results The final study sample included 869 households. Medicines were responsible for 47% of household expenditure with health; second came private health plans which accounted for 22%. The richest spent, on average, 70 times more them the poorest with health plans, 26 times more with dental treatment and six times more with medicines. About 16% households committed 20% or more of their income with health, independent of economic position. Similarly, 12% of the households had health expenditure in excess of 40% of their capacity to pay.

Conclusion The proportion of income spent on health was similar across economic groups, but this ‘equality’ is achieved at an unacceptably high level. Specific strategies to reduce such vulnerability are needed.

Methods To examine if pre-natal exposure to paracetamol is associated with the risk of asthma or wheezing in early childhood, we selected 66 445 women from the Danish National Birth Cohort for whom we had information on paracetamol use during pregnancy and who participated in an interview when their children were 18-months-old and 12 733 women whose children had reached the age of 7 and estimated the prevalence of physician-diagnosed asthma and wheezing at the ages of 18 months and 7 years. We also linked our population to the Danish National Hospital Registry to record all hospitalizations due to asthma up to age of 18 months.

Results Paracetamol use during any time of pregnancy was associated with a small but statistically significant increased risk of physician-diagnosed asthma or bronchitis among children at 18 months [relative risk (RR) = 1.17, 1.13–1.23)], hospitalizations due to asthma up to 18 months (hazard ratio = 1.24, 1.11–1.38) and physician-diagnosed asthma at 7 years (RR = 1.15, 1.02–1.29). The highest risks were observed for paracetamol use during the first trimester of pregnancy and persistent wheezing (wheezing at both 18 months and 7 years) (RR = 1.45, 1.13–1.85).

Conclusion Paracetamol use during pregnancy was associated with an increased risk of asthma and wheezing in childhood. If this association is causal, we may need to revisit the clinical practice on use of paracetamol during pregnancy.

Methods In this analysis of the Copenhagen City Heart Study, in which alcohol intake was measured four times, 9318 Danish women with no previous diagnosis of cancer were followed for breast cancer for 27 years, from 1976 to 2002. During follow-up, breast cancer was diagnosed in 476 women.

Results The association between alcohol intake at first measurement (baseline alcohol intake) and breast cancer was positive and approximately linear. When alcohol intake was updated during follow-up, no association was observed between breast cancer and alcohol intake. It is suggested that this difference in results may be attributable to long latency time between alcohol intake and breast cancer occurrence, because a markedly increased risk was estimated on the basis of direct lagging of risk time.

Conclusions Our results support the hypothesis that baseline alcohol intake is more strongly associated with breast cancer risk than updated intake, and we suggest that this is due to the long latency between alcohol intake and breast cancer.

Methods A cohort of 75 868 subjects aged 35 years and above, participants of the Trivandrum Oral Cancer Study—a cluster-randomized controlled trial originally implemented to evaluate the efficacy of visual inspection on oral cancer, in Kerala State, South India—were followed up from 1995 to 2004. Weight and height were measured both at baseline and in 3.5-year follow-up surveys. Early years of follow-up were excluded from the analyses. Relative risks of overall death and cause-specific death were estimated according to the body mass index (BMI) category of the WHO Asian population definitions, and to weight changes between two surveys.

Results Low BMI was a predictor of mortality, while high BMI was not. Mortality risks in men adjusted for age, smoking habits and other potential confounders, as compared with a BMI 18.5–22.9 kg/m², were 1.26 (95% CI 1.03–1.55) for BMI < 16 kg/m²; 1.16 (1.03–1.32) for BMI = 16–18.4 kg/m²; 0.95 (0.81–1.12) for BMI = 23–24.9 kg/m²; 0.85 (0.69–1.05) for BMI = 25–27.4 kg/m²; and 0.89 (0.65–1.21) for BMI ≥ 27.5 kg/m². Similar findings were observed in women. BMI was not associated with deaths from cancer, cardiovascular and cerebrovascular diseases, and diabetes. A low BMI (<16 kg/m²) was associated with increased deaths from chronic respiratory diseases. Smoking and socio-economical status did modify the association. A moderate weight gain of 4–10% between the two surveys was associated with decreased risk of death, while moderate and severe weight loss were predictive factors of death. Similar results were observed in both men and women.

Conclusions Among this Indian rural population, mild to severe leanness (BMI < 16 kg/m²) and weight loss were important determinants of mortality, especially from chronic respiratory diseases, while overweight and above (BMI > 23 kg/m²) did not show any detrimental effect.

Methods Literature search for review and research articles on temporal trends in US breast cancer incidence during the past 25 years, starting in the mid-1980s, when extant epidemiologic evidence had already indicated that HT increased risk of breast cancer.

Results Among the 21 articles identified, spanning from 1987 to 2007, nine included no mention of HT as a possible factor contributing to the steep rise in breast cancer incidence in the 1980s, seven included a minor mention and only five (one published in 2003, the others in 2006 and 2007) provided any substantive discussion of this issue—but only in relation to current trends and not the 1980 rise in breast cancer incidence.

Conclusion A critical appraisal of the epidemiologic literature highlights important gaps in explanations for breast cancer incidence trends and also how current and changing population patterns of disease distribution are ultimately what put our aetiologic explanations to the test.

Methods In 1996, 2338 adult offspring of participants of the 1970s Renfrew/Paisley study took part in a screening examination. They provided information on sociological, behavioural and clinical risk factors, as their parents had done 20 years previously. Social class and father's social class were available, enabling their influence on risk factors to be investigated.

Results Generally risk factors improved for offspring compared with parents, except for Body mass index and obesity, which worsened. Risk factors were less favourable in manual compared with non-manual offspring, and were more closely related to own than father's social class. There was a large amount of upward social mobility involving 35% of sons and 50% of daughters. Risk factors for the upwardly mobile tended to be more favourable than the class they left behind but less favourable than the class they joined.

Conclusions The concomitants of social mobility may reflect behavioural choices, such as smoking, and adverse factors, which are more difficult to leave behind. The relatively fast changes in social class profile may not be reflected in as quick changes in population health, as the upwardly mobile bring their earlier life adversities with them.

Method Three-year follow-up cross-sectional survey was carried out in 40 rural Ethiopian communities to evaluate the programme. Households were randomly selected and all children were invited for eye examination for active trachoma. In 2005, eye swabs were taken for Polymerase Chain Reaction (PCR) detection of ocular C. trachomatis DNA. Adult knowledge and behaviour related to trachoma were assessed.

Results Community summarized mean prevalence, overall, was 35.6% (SD = 17.6) for active trachoma, 34.0% (18.7) for trachomatous inflammation, follicular (TF) alone and 4.3% (5.3) for PCR positivity for C. trachomatis. After adjustment, odds of active trachoma were reduced in communities receiving antibiotics and one or two educational intervention components (OR = 0.35, 95% CI: 0.13–0.89 or OR = 0.31, 0.11–0.89, respectively). The odds of being PCR positive were lower in these intervention arms, compared with control (OR = 0.20, 0.06–0.62 and OR = 0.07, 0.02–0.30, respectively). Knowledge of treatment and preventative methods were reported with much higher frequency, compared with baseline.

Conclusions Trachoma remains a public health problem in Ethiopia. Antibiotic administration remains the most effective intervention but community-based health education programmes can impact, to additionally reduce prevalence of C. trachomatis.

Methods Cross-sectional design, including data of the Survey of Health, Ageing and Retirement in Europe (Sweden, Denmark, Austria, France, Germany, The Netherlands, Spain, Italy and Greece). The population aged from 50 to 74 years was included. Absolute and relative social class dimension inequalities in poor self-reported health and long-term illness were determined for each sex and political tradition. Relative inequalities were assessed by fitting Poisson regression models with robust variance estimators.

Results Absolute and relative health inequalities by social class dimensions are found in the three political traditions, but these differences are more marked in Late democracies and mainly among women. For example the prevalence ratio of poor self-perceived health comparing poorly educated women with highly educated women, was 1.75 (95% CI: 1.39–2.21) in Late democracies and 1.36 (95% CI: 1.21–1.52) in Social democracies. The prevalence differences were 24.2 and 13.7%, respectively.

Conclusion This study is one of the first to show the impact of different political traditions on social class inequalities in health. These results emphasize the need to evaluate the impact of the implementation of public policies.

Methods The International Study of Asthma and Allergies in Childhood (ISAAC) used a standard questionnaire to measure the 12-month period prevalence of asthma symptoms by parental report in 6–7 year olds in 40 countries, and by self-report in 13–14 year olds in 60 countries. The initial survey was in the mid 1990s (Phase One) and this was repeated in the early 2000s (Phase Three). We correlated the prevalence values of any wheeze and severe wheeze with national data on mortality and hospital admissions for asthma in 5–14 year olds.

Results All correlations with prevalence were positive. In 13–14 year olds, the correlations between severe wheeze in Phase One and contemporaneous mortality and hospital admission rates were r = 0.32 (P = 0.047) and r = 0.73 (P = 0.003), respectively. In 6–7 year olds in Phase One, the correlation with severe wheeze and mortality was r = 0.42 (P = 0.024). In 14 countries the correlation between admission and mortality rates in the 5–14 year age group was r = 0.53 (P = 0.054).

Conclusions There are consistently positive associations between asthma symptom prevalence, admissions and mortality. The prevalence of asthma symptoms in children obtained from local questionnaire studies may provide a guide to estimate the incidence of severe episodes of asthma in countries with incomplete data on hospital admissions or mortality, or vice versa.

Methods The cohort comprised 64 937 individuals ever employed at the study sites between 1946 and 2002, followed up to 2005; radiation exposures as measured by personal dosimeters (‘film badges’) were available for 42 426 individuals classified as ‘radiation workers’. Poisson regression models were used to investigate the relationship between excess mortality rates and cumulative radiation exposure, using both relative and additive risk models.

Results The cohort shows a pronounced ‘healthy worker’ effect. Overall, socio-economic status as indicated by employment status has a greater influence on mortality than does radiation exposure status. For male radiation workers, there is an apparent dose response for mortality from circulatory system disease [P < 0.001, ERR = 0.65 (90% CI 0.36–0.98) Sv^–1]. However there is evidence for inhomogeneity in the apparent dose response (P = 0.016), arising principally at cumulative doses in excess of 300 mSv, when the four categories of employment and radiation exposure status are examined separately.

Conclusions We have found evidence for an association between mortality from non-cancer causes of death, particularly circulatory system disease, and external exposure to ionizing radiation in this cohort. However, the tentative nature of biological mechanisms that might explain such an effect at low chronic doses and the above inhomogeneities in apparent dose–response, mean that the results of our analysis are not consistent with any simple causal interpretation. Further work is required to explain these inhomogeneities, and on the possible role of factors associated with socio-economic status and shift working, before any further conclusions can be drawn.

Methods A mathematical model simulated observed levels of HIV prevalence under the complete range of current levels of circumcision. Increased MC from 2007 was incorporated in this model and used to simulate HIV prevalence in 2020.

Results Complete coverage by MC could reduce HIV prevalence from 12 to 6% for an average population country in sub-Saharan Africa in 2020. This reduction is scaled proportionally when lower circumcision levels are achieved. These benefits are achieved mostly by circumcising men between 20 and 30 years of age (adult prevalence reduced from 12 to 10%), and those with riskier behaviour (8 to 6.9%). Complete negation of these benefits requires at least 40% of circumcised males to significantly increase risky behaviour.

Conclusions MC provides an effective intervention in sub-Saharan Africa to reduce HIV prevalence. It is most effective when applied to 20–30 year old risky males with diminishing returns with application to the wider male population.

Methods Information on smoking was collected from 10 areas in four continents among population-based, age-stratified random samples of women aged 15 years or older. HPV testing was performed using PCR-based enzyme immunoassay. Unconditional logistic regression was used to estimate odds ratios (OR) and corresponding 95% confidence intervals (CI) of being HPV-positive by smoking habits, adjusted for age and lifetime number of sexual partners.

Results Ten thousand five hundred and seventy-seven women (mean age 41.4 years) were included. Among current smokers, the risk of being HPV-positive increased with smoking intensity, after allowing for lifetime number of sexual partners: ORs for <5, 5–14 and ≥15 cigarettes per day were 1.21 (95% CI 0.95–1.54), 1.39 (95% CI 1.04–1.87) and 2.01 (95% CI 1.32–3.08), respectively, as compared with never-smokers. The risk among former smokers (OR = 0.95, 95% CI 0.73–1.23) was similar to that among never-smokers. Analyses stratified by lifetime number of sexual partners showed a significant trend in risk only for women with one lifetime sexual partner.

Conclusions Our study suggests that current, though not former, smoking is associated with an increased prevalence of HPV, after allowance for sexual covariates. Among current smokers, HPV prevalence increased with smoking intensity, but a clear dose–response relationship was exclusively seen among women who declared one lifetime sexual partner.

Methods The source population has children aged 4–6 commencing school in the Australian Capital Territory from 2000 to 2005 inclusive. Over 80% of these children and their families completed a health questionnaire on asthma, other atopic disease and respiratory symptoms using some questions from the International Study of Asthma and Allergies in Childhood (n = 22 882). Current asthma has been previously validated against physician assessment in this setting.

Results The prevalence of current asthma declined (P < 0.001) but eczema ever increased (P < 0.001) from 2000 to 2005. The asthma decline was predominantly linear in form, and accompanied by a reduction in night cough and shortness of breath but not recent wheeze. Compared with Australian-born children, children from New Zealand and the United Kingdom had a similar prevalence of asthma, hay fever and eczema history. However, children born in other countries, such as Asia, generally had a lower prevalence of these disorders. The temporal trends for atopic disorders or respiratory symptoms did not differ for overseas-born compared with Australian-born children.

Conclusion The decline in current asthma prevalence from 2000 to 2005 was linear in form and appeared uncoupled from trends in child eczema. Country of birth was associated with marked variation in atopic disorder prevalence. The similar temporal trends for Australian vs overseas-born children indicate that the factors underlying the asthma prevalence decline are unlikely to be only in the pre-natal period.

Methods Of the GAZEL cohort, 14 445 participants aged 39–54 years in 1993 and followed-up over 12.7 years, completed the Bortner-Type-A-scale, the Buss-Durkee-Hostility-Inventory and the Grossarth-Maticek and Eysenck-Personality-Stress-Inventory. Indicators of SEP, such as father's social class, education, occupational grade and income, were assessed at baseline. Relative indices of inequality in Cox regression models were used to estimate associations.

Results In age-adjusted analyses, risk of death was inversely associated with SEP among men and women. Among men, the attenuation in this association depended on the measures of SEP and was 28–29% for ‘neurotic-hostility’, 13–22% for ‘anti-social’ and 13–16% for ‘CHD-prone’ personality. In women, the attenuation was evident only for type-A-behaviour, by 11%. After controlling simultaneously for all personality factors that predicted mortality, associations between SEP and mortality were attenuated in men: by 34% for education, 29% for occupational position and 28% for income; but were only attenuated by 11% for income in women. For cardiovascular mortality, the corresponding percentages of reduction were 42, 31 and 44% after adjustment for ‘CHD-prone’ personality in men.

Conclusions Personality measures explained some of the mortality gradients observed for measures of adult socioeconomic position in men, but had little explanatory power in women. Whether personality represents a predictor or an outcome of social circumstances needs further research.

Methods A systematic literature review identified nine diseases with sufficient evidence of a causal relationship with UVR exposure and for which the population attributable fraction (PAF) for UVR could be estimated. For cutaneous malignant melanoma and cataract, the PAF was directly applied to disease burdens already calculated by WHO. For seven other diseases, we developed population-level exposure–disease relationships and used these to calculate disease incidence and mortality, and thence disease burden. We also estimated the disease burden from rickets, osteomalacia and osteoporosis that might result if global UVR exposure was reduced to very low levels.

Results UVR exposure is a minor contributor to the world's disease burden, causing an estimated annual loss of 1.6 million DALYs; i.e. 0.1% of the total global disease burden. A markedly larger annual disease burden, 3.3 billion DALYs, might result from reduction in global UVR exposure to very low levels.

Conclusions Sun protection messages are important to prevent diseases of UVR exposure. However, without high dietary (or supplemental) intake of vitamin D, some sun exposure is essential to avoid diseases of vitamin D insufficiency.

Methods Using the voters’ list as the selection frame, 148 173 men and women aged ≥35 years were recruited during 1991–97 in Mumbai city and were followed-up during 1997–2003.

Results During 774 129 person-years of follow-up, 13 261 deaths were observed. After adjusting for the potential confounders, increased mortality was observed in all under-weight categories [relative risk (RR) = 1.94 for BMI < 16.0 kg/m², 1.38 for BMI 16.0 to <17.0 and 1.24 for BMI 17.0 to <18.5 among women; the corresponding values for men were 2.24, 1.45 and 1.27, respectively] when compared with the rate in the normal weight category (BMI 18.5 to <25.0). Extremely thin (BMI < 16.0 kg/m²) cohort members were at highest risk for death due to tuberculosis (TB) (RR = 7.20 and 14.94 in women and men, respectively), cancer (RR = 1.87 and 2.44, respectively) and respiratory diseases (RR = 3.46 and 4.35, respectively). Subjects with above normal BMI had lower mortality risk than those with normal BMI values. Over-weight (BMI 25.0 to <30.0) women (RR = 0.89) and men (RR = 0.87) were at a lower risk; however, obese (BMI ≥ 30.0) men <60 years of age had an increased mortality risk (RR = 1.22).

Conclusion In Mumbai, under-weight was associated with an increased risk of pre-mature death. Despite the growing concerns regarding the gradual transition toward increasing rates of obesity, under-nutrition remains a major health problem in India. This study suggests the need for public health policies focusing on reducing under-nutrition.

Methods The P-value is by far the most commonly used measure, but requires careful calibration when the a priori probability of an association is small, and discards information by not considering the power associated with each test. The q-value is a frequentist method by which the false discovery rate (FDR) may be controlled.

Results We advocate the use of the Bayes factor as a summary of the information in the data with respect to the comparison of the null and alternative hypotheses, and describe a recently-proposed approach to the calculation of the Bayes factor that is easily implemented. The combination of data across studies is straightforward using the Bayes factor approach, as are power calculations.

Conclusions The Bayes factor and the q-value provide complementary information and when used in addition to the P-value may be used to reduce the number of reported findings that are subsequently not reproduced.

Methods A total of 20 936 women with singleton pregnancies were identified within three population-based pregnancy cohorts in Northern Finland (1985–1986) and in Denmark (1984–1987 and 1989–1991). We collected self-reported data on their pre-pregnancy and pregnancy smoking habits and followed the children to school age where teachers and parents rated hyperactivity and inattention symptoms.

Results Children, whose mothers smoked during pregnancy, had an increased prevalence of a high hyperactivity-inattention score compared with children of nonsmokers in each of the cohorts after adjustment for confounders but we found no statistical significant difference between the associations across the cohorts.

Conclusion The estimated association was not strongest in the population with the fewest smokers which does not support the hypothesis that the association is entirely due to genetic confounding.

Methods We investigated the association between education and lifetime smoking patterns in a birth cohort established in 1959 and followed through adulthood (n = 1311). We controlled for a wide range of potential confounders that were measured prior to school entry, and also estimated sibling fixed effects models to control for unmeasured familial vulnerability to smoking.

Results In the full sample of participants, regression analyses adjusting for multiple childhood factors (including socioeconomic status, IQ, behavioural problems, and medical conditions) indicated that the number of pack-years smoked was higher among individuals with less than high school education [rate ratio (RR) = 1.58, confidence interval (CI) = 1.31, 1.91]. However, in the sibling fixed effects analysis the RR was 1.23 (CI = 0.80, 1.93). Similarly, adjusted models estimated in the full sample showed that individuals with less than high school education had fewer short-term (RR = 0.40; CI = 0.23, 0.69) and long-term (RR = 0.59; CI = 0.42, 0.83) quit attempts, and were less likely to quit smoking (odds ratio = 0.34; CI = 0.19, 0.62). The effects of education on quitting smoking were attenuated in the sibling fixed effects models that controlled for familial vulnerability to smoking.

Conclusions A substantial portion of the education differential in smoking that has been repeatedly observed is attributable to factors shared by siblings that contribute to shortened educational careers and to lifetime smoking trajectories. Reducing disparities in cigarette smoking, including educational disparities, may therefore require approaches that focus on factors early in life that influence smoking risk over the adult life span.

Background A better understanding of behavioural factors associated with the interruption of human immunodeficiency virus (HIV) dynamics in Uganda are fundamental for guiding HIV prevention strategies. A comparative analysis of population level HIV surveillance and behavioural data in Uganda and countries with similar epidemic dynamics offers a framework to identify distinctive elements of the Ugandan success.

Methods We analysed HIV prevalence among pregnant women attending antenatal surveillance sites between 1990-1998 in Uganda, Kenya, Zambia, and Malawi, and AIDS behavioural data collected in Uganda (1989, 1995), Kenya (1998), Zambia, and Malawi (1996).

Results In Uganda between 1991 and 1998, HIV prevalence rates fell 54% overall, with 75% and 57% declines among 15-19 and 20-24 age groups in urban areas, respectively. HIV declines were associated with a 60% reduction in multiple sexual partners between 1989 and 1995, as well as increased condom use and sexual abstinence. The lack of comparable HIV declines in comparison countries with similar rates of abstinence and condom use, but with multiple sexual partnership rates similar to Uganda in 1989, suggests that partner reduction is paramount in interrupting sexual HIV dynamics. Comparatively, Ugandans were more likely to acknowledge people with AIDS and communicate about HIV through personal networks. Modelling suggests an 80% reduction in incidence among youth early in the 1990s, suggesting a preceding behavioural process during the late 1980s, which reduced HIV risk.

Conclusion The distinctive element explaining declines in HIV prevalence in Uganda is sexual partner reduction. The Ugandan response to AIDS involved a primary HIV intervention whose key outcome was risk avoidance through partner reduction. This focus for HIV prevention could prove more effective than other widely advocated strategies in sub-Saharan Africa.