Applied causal inference methods for sequential mediators.

BACKGROUND: Mediation analysis aims at estimating to what extent the effect of an exposure on an outcome is explained by a set of mediators on the causal pathway between the exposure and the outcome. The total effect of the exposure on the outcome can be decomposed into an indirect effect, i.e. the effect explained by the mediators jointly, and a direct effect, i.e. the effect unexplained by the mediators. However finer decompositions are possible in presence of independent or sequential mediators. METHODS: We review four statistical methods to analyse multiple sequential mediators, the inverse odds ratio weighting approach, the inverse probability weighting approach, the imputation approach and the extended imputation approach. These approaches are compared and implemented using a case-study with the aim to investigate the mediating role of adverse reproductive outcomes and infant respiratory infections in the effect of maternal pregnancy mental health on infant wheezing in the Ninfea birth cohort. RESULTS: Using the inverse odds ratio weighting approach, the direct effect of maternal depression or anxiety in pregnancy is equal to a 59% (95% CI: 27%,94%) increased prevalence of infant wheezing and the mediated effect through adverse reproductive outcomes is equal to a 3% (95% CI: -6%,12%) increased prevalence of infant wheezing. When including infant lower respiratory infections in the mediation pathway, the direct effect decreases to 57% (95% CI: 25%,92%) and the indirect effect increases to 5% (95% CI: -5%,15%). The estimates of the effects obtained using the weighting and the imputation approaches are similar. The extended imputation approach suggests that the small joint indirect effect through adverse reproductive outcomes and lower respiratory infections is due entirely to the contribution of infant lower respiratory infections, and not to an increased prevalence of adverse reproductive outcomes. CONCLUSIONS: The four methods revealed similar results of small mediating role of adverse reproductive outcomes and early respiratory tract infections in the effect of maternal pregnancy mental health on infant wheezing. The choice of the method depends on what is the effect of main interest, the type of the variables involved in the analysis (binary, categorical, count or continuous) and the confidence in specifying the models for the exposure, the mediators and the outcome.

Alcohol and cigarette consumption predict mortality in patients with head and neck cancer: a pooled analysis within the International Head and Neck Cancer Epidemiology (INHANCE) Consortium.

BACKGROUND: This study evaluated whether demographics, pre-diagnosis lifestyle habits and clinical data are associated with the overall survival (OS) and head and neck cancer (HNC)-specific survival in patients with HNC. PATIENTS AND METHODS: We conducted a pooled analysis, including 4759 HNC patients from five studies within the International Head and Neck Cancer Epidemiology (INHANCE) Consortium. Cox proportional hazard ratios (HRs) and the corresponding 95% confidence intervals (CIs) were estimated including terms reported significantly associated with the survival in the univariate analysis. RESULTS: Five-year OS was 51.4% for all HNC sites combined: 50.3% for oral cavity, 41.1% for oropharynx, 35.0% for hypopharynx and 63.9% for larynx. When we considered HNC-specific survival, 5-year survival rates were 57.4% for all HNC combined: 54.6% for oral cavity, 45.4% for oropharynx, 37.1% for hypopharynx and 72.3% for larynx. Older ages at diagnosis and advanced tumour staging were unfavourable predictors of OS and HNC-specific survival. In laryngeal cancer, low educational level was an unfavourable prognostic factor for OS (HR = 2.54, 95% CI 1.01-6.38, for high school or lower versus college graduate), and status and intensity of alcohol drinking were prognostic factors both of the OS (current drinkers HR = 1.73, 95% CI 1.16-2.58) and HNC-specific survival (current drinkers HR = 2.11, 95% CI 1.22-3.66). In oropharyngeal cancer, smoking status was an independent prognostic factors for OS. Smoking intensity (>20 cigarettes/day HR = 1.41, 95% CI 1.03-1.92) was also an independent prognostic factor for OS in patients with cancer of the oral cavity. CONCLUSIONS: OS and HNC-specific survival differ among HNC sites. Pre-diagnosis cigarette smoking is a prognostic factor of the OS for patients with cancer of the oral cavity and oropharynx, whereas pre-diagnosis alcohol drinking is a prognostic factor of OS and HNC-specific survival for patients with cancer of the larynx. Low educational level is an unfavourable prognostic factor for OS in laryngeal cancer patients.

Role of medical history and medication use in the aetiology of upper aerodigestive tract cancers in Europe: the ARCAGE study.

BACKGROUND: The study aimed to investigate the role of medical history (skin warts, Candida albicans, herpetic lesions, heartburn, regurgitation) and medication use (for heartburn; for regurgitation; aspirin) in the aetiology of upper aerodigestive tract (UADT) cancer. METHODS: A multicentre (10 European countries) case-control study [Alcohol-Related CAncers and GEnetic susceptibility (ARCAGE) project]. RESULTS: There were 1779 cases of UADT cancer and 1993 controls. History of warts or C. albicans infection was associated with a reduced risk [odds ratio (OR) 0.80, 95% confidence interval (CI) 0.68-0.94 and OR 0.73, 95% CI 0.60-0.89, respectively] but there was no association with herpetic lesions, heartburn, regurgitation or medication for related symptoms. Regurgitation was associated with an increased risk for cancer of the oesophagus (OR 1.47, 95% CI 0.98-2.21). Regular aspirin use was not associated with risk of UADT cancer overall but was associated with a reduced risk for cancer of oesophagus (OR 0.51, 95% CI 0.28-0.96), hypopharynx (OR 0.53, 95% CI 0.28-1.02) and larynx (OR 0.74, 95% CI 0.54-1.01). CONCLUSIONS: A history of some infections appears to be a marker for decreased risk of UADT cancer. The role of medical history and medication use varied by UADT subsites with aspirin use associated with a decreased risk of oesophageal cancer and suggestive of a decreased risk of hypopharyngeal and laryngeal cancers.

Age at puberty and risk of testicular cancer: a meta-analysis.

Testicular cancer is one of the most rapidly increasing tumour types but its aetiology is still largely unexplained. Cryptorchidism and familial testicular cancer, established risk factors, explain less than 10% of all cases. Among investigated post-natal factors, early puberty was suggested as a potential risk factor but the topic has been poorly investigated. We undertook a meta-analysis of the effect of age at puberty on testicular cancer risk, attempting at enhancing the homogeneity in the definition of the exposure among studies to obtain valid pooled estimates. Search strategies were conducted in PubMed on December 2011. All markers of puberty onset (age at voice change, age when started shaving and reported age at onset) were considered. We re-categorized age at puberty from all studies into a common three-level variable: younger than peers, same age as peers, older than peers. A total of 391 references were retrieved, of which 12 met the inclusion criteria. Later puberty appeared to be protective. In particular late vs. same age at start shaving gave an OR of 0.84 (95% CI: 0.75-0.95, five studies); late vs. same age at voice change gave an OR of 0.87 (95% CI: 0.75-1.01, five studies); and later age than peers at reported onset of puberty gave an OR of 0.81 (95% CI: 0.73-0.89, eight studies). Early puberty showed no effect on testicular cancer risk. This meta-analysis has found consistent evidence of a decreased risk of testicular cancer in association with later puberty, suggesting that post-natal factors may contribute to testicular cancer risk.

Life course social mobility and risk of upper aerodigestive tract cancer in men.

The aim of this study was to explore associations between social mobility and tumours of the upper aero-digestive tract (UADT), focussing on life-course transitions in social prestige (SP) based on occupational history. 1,796 cases diagnosed between 1993 and 2005 in ten European countries were compared with 1585 controls. SP was classified by the Standard International Occupational Prestige Scale (SIOPS) based on job histories. SIOPS was categorised in high (H), medium (M) and low (L). Time weighted average achieved and transitions between SP with nine trajectories: H --> H, H --> M, H --> L, M --> H, M --> M, M --> L, L --> H, L --> M and L --> L were analysed. Odds ratios (ORs) and 95%-confidence intervals [95%-CIs] were estimated with logistic regression models including age, consumption of fruits/vegetables, study centre, smoking and alcohol consumption. The adjusted OR for the lowest versus the highest of three categories (time weighted average of SP) was 1.28 [1.04-1.56]. The distance of SP widened between cases and controls during working life. The downward trajectory H --> L gave an OR of 1.71 [0.75-3.87] as compared to H --> H. Subjects with M --> M and L --> L trajectories ORs were also elevated relative to subjects with H --> H trajectories. The association between SP and UADT is not fully explained by confounding factors. Downward social trajectory during the life course may be an independent risk factor for UADT cancers.

Integrating environmental, entomological, animal, and human data to model the Leishmania infantum transmission risk in a newly endemic area in Northern Italy.

INTRODUCTION: Historically, leishmaniasis in Italy was constrained to areas with Mediterranean climate. In the last 20 years, sand fly vectors (Phlebotomus perniciosus), cases of canine leishmaniasis (CanL) and cases of human visceral leishmaniasis (VL) have been observed in Northern Italian regions, traditionally classified as cold areas unsuitable for sand fly survival. AIM: We aim to evaluate through a One-Health approach the risk of endemic transmission of Leishmania infantum in the Piedmont Region, Northern Italy. METHODS: We collected environmental, entomological, animal, and human data. We applied a geostatistical binomial model to map the probability of P. perniciosus presence in the study area, using selected environmental parameters as predictors. We evaluated the spatial relationship between the probability of P. perniciosus presence and the geographical distribution of CanL and VL cases observed between 1999 and 2013. RESULTS: Between 1999 and 2003, 142 sampling sets (17%) out of 839 resulted positive for P. perniciosus. Elevation, degree of slope, normalized difference vegetation index (NDVI) and summer temperatures were associated with positive sampling sets. During the study period, 164 (13.6%) of Piedmont municipalities reported at least one autochthonous case of CanL, while 89 VL cases were observed in 54 municipalities (4.5%). We observed an association between municipalities affected by autochthonous CanL cases and the estimated probability of P. perniciosus presence (Odds Ratio for 10% increase of probability: 2.66; 95% confidence intervals (CI): 2.16-3.37). We found that human VL incident cases were positively associated with the probability of the municipality of residence of being endemic for CanL (Incidence Rate Ratio for 10% increase of probability: 1.49; 95% CI 1.02-2.16). CONCLUSIONS: Using a One-Health approach, we quantified the spatial association between the distribution of P. perniciosus, municipalities endemic for CanL and incident cases of human VL, suggesting that the disease has become endemic in the Piedmont region.

Rates of orchiopexies in Sweden: 1977-1991.

Cryptorchidism is a known risk factor for testicular cancer and the secular increase in testicular cancer incidence might have been paralleled by a similar increase in cryptorchidism. Data on trends in prevalence of cryptorchidism are however conflicting and decreases have recently been reported. To analyse Swedish trends in rates of orchiopexy, we used the Swedish Hospital Discharge Register to identify all cases of orchiopexy carried out for cryptorchidism between 1977 and 1991, that is, before the era of outpatient orchiopexies in Sweden. Observed trends were analysed in 5-year age groups. The estimated average annual per cent changes (EAPCs) and the years in which the EAPC significantly changed were estimated using Joinpoint Regression. Finally, we estimated the cumulative incidence of orchiopexy by birth cohort. Among boys aged less than 10, the orchiopexy rate started to decrease in the early 1980s. EAPCs were -2.88 (95% confidence interval (CI): -5.48, -0.21) among boys aged 5-9 and -6.63 among those aged 0-4 (95% CI: -8.84, -4.37). Among subjects aged at least 10, the rate decreased over the whole study period. Although the use of orchiopexy rates to measure prevalence of cryptorchidism has limitations, our findings may suggest that cryptorchidism prevalence decreased in Sweden starting from the early 1980s.

Sources of confounding in life course epidemiology.

In epidemiologic analytical studies, the primary goal is to obtain a valid and precise estimate of the effect of the exposure of interest on a given outcome in the population under study. A crucial source of violation of the internal validity of a study involves bias arising from confounding, which is always a challenge in observational research, including life course epidemiology. The increasingly popular approach of meta-analyzing individual participant data from several observational studies also brings up to discussion the problem of confounding when combining data from different populations. In this study, we review and discuss the most common sources of confounding in life course epidemiology: (i) confounding by indication, (ii) impact of baseline selection on confounding, (iii) time-varying confounding and (iv) mediator-outcome confounding. We also discuss the issue of addressing confounding in the context of an individual participant data meta-analysis.

Maternal anxiety, depression and sleep disorders before and during pregnancy, and preschool ADHD symptoms in the NINFEA birth cohort study.

AIMS.: Maternal mental disorders have been associated with the risk of attention-deficit/hyperactivity disorder (ADHD) in children. Within the context of a mother-child cohort, we examined whether maternal anxiety, depression and sleep disorders are associated with pre-school ADHD symptoms. METHODS.: The study included 3634 singletons from the Italian NINFEA (Nascita e INFanzia: gli Effetti dell'Ambiente') cohort. Maternal doctor-diagnosed anxiety, depression and sleep disorders before and during pregnancy were assessed from the questionnaires completed during pregnancy and 6 months after delivery. Mothers rated child ADHD symptoms at 4 years of age, according to the Diagnostic and Statistical Manual of Mental Disorders. Hyperactive-impulsive (ADHD-H), inattentive (ADHD-I) and total ADHD scores were analysed in the models adjusted for child's gender, first-born status, maternal age, education, alcohol consumption and smoking during pregnancy. RESULTS.: The total ADHD score at age 4 was associated with maternal lifetime anxiety (17.1% percentage difference in score compared with never; 95% CI 7.3-27.9%), sleep disorders (35.7%; 95% CI 10.7-66.5%) and depression (17.5%; 95% CI 3.2-33.8%). Similar positive associations were observed also for ADHD-H and ADHD-I traits, with slightly attenuated associations between maternal sleep disorders and child ADHD-I score, and maternal depression and both ADHD scores. All the estimates were enhanced when the disorders were active during pregnancy and attenuated for disorders active only during the pre-pregnancy period. CONCLUSIONS.: Maternal anxiety, depression and sleep disorders are associated with a relative increase in the number of ADHD-H, ADHD-I and total ADHD symptoms in preschoolers.

Baldness and testicular cancer: the EPSAM case-control study.

The etiology of testicular cancer is largely unexplained. Research has mainly focused on prenatal exposures, especially to sex hormones, while less attention has been paid to exposures that may act also postnatally. As baldness has been previously associated with testicular cancer risk we focused on baldness and body hairiness, which are both associated with androgen activity. We used data of the Postnatal Exposures and Male Health (EPSAM) study, a case-control study on testicular cancer conducted in the Province of Turin, Italy, involving cases diagnosed between 1997 and 2008. Information was collected using mailed questionnaires. Analyses included 255 cases and 459 controls. We calculated ORs and 95% CIs to estimate testicular cancer risk among those who developed baldness and among those with body hairiness. We found an inverse association between testicular cancer and baldness (OR: 0.67, 95% CI: 0.46-0.98) and body hairiness (OR: 0.78, 95% CI: 0.53-1.16), although the latter had wider CIs. The inverse association between baldness and testicular cancer is consistent with the results from previous studies. These results suggest that androgens activity may influence testicular cancer risk.

Effect of different approaches to treatment of smoking as a potential confounder in a case-control study on occupational exposures.

AIM: To evaluate the effect of different approaches to treatment of smoking as a potential confounder in an occupational study of lung cancer. METHODS: Data were used from a case-control study on 956 men with lung cancer and 1253 population controls recruited in two northern Italian areas during 1990-1992. The risk of lung cancer associated with 11 selected job titles and eight selected industrial activities was estimated using seven different methods to treat smoking history. To evaluate the confounding effect of smoking, odds ratios obtained using the first six models were compared with estimates from the seventh and most complex model, in which cumulative tobacco consumption and time since cessation were considered. RESULTS: Although crude odds ratios for some of the occupational categories were biased by up to 25%, such bias decreased to less than 10% when a simple model including smoking status (never, ex-, current) was used. CONCLUSIONS: In occupational studies on lung cancer risk, information on smoking status may allow satisfactory control of the potential confounding effect of the habit.

Global DNA hypomethylation in prostate cancer development and progression: a systematic review.

BACKGROUND: The role of global DNA methylation in prostate cancer (PCa) remains largely unknown. Our aim was to summarize evidence on the role of global DNA hypomethylation in PCa development and progression. METHODS: We searched PubMed through December 2013 for all studies containing information on global methylation levels in PCa tissue and at least one non-tumor comparison tissue and/or studies reporting association between global methylation levels in PCa tissue and survival, disease recurrence or at least one clinicopathological prognostic factor. We summarized results using non-parametric comparisons and P-value summary methods. RESULTS: We included 15 studies in the review: 6 studies with both diagnostic and prognostic information, 5 studies with only diagnostic information and 4 studies with only prognostic information. Quantitative meta-analysis was not possible because of the large heterogeneity in molecular techniques, types of tissues analyzed, aims and study designs. Summary statistical tests showed association of DNA hypomethylation with PCa diagnosis (P<0.006) and prognosis (P<0.001). Restriction to studies assessing 5-methylcytosine or long interspersed nucleotide element-1 revealed results in the same direction. Analyses restricted to specific clinicopathological features showed association with the presence of metastasis and tumor stage in all tests with P<0.03, and no association with Gleason score (all tests P>0.1 except for the weighted Z-test, P=0.05). CONCLUSION: DNA hypomethylation was associated with PCa development and progression. However, due to the heterogeneity and small sample sizes of the included studies, along with the possibility of publication bias, this association requires additional assessment.

Risk of oral squamous cell carcinoma in 402 patients with oral lichen planus: a follow-up study in an Italian population.

The most important complication of oral lichen planus (OLP) is the development of oral squamous cell carcinoma (OSCC) but this is a very controversial matter. The aim of the study was to estimate in a Northern Italian cohort of OLP patients the risk for OSCC. Four hundred and two patients with histologically confirmed OLP diagnosed from January 1988 to July 1999, were followed-up to the end of February 2001. The standardized incidence ratio (SIR) of OSCC was calculated for the entire cohort and specific for gender, type of OLP, therapy for OLP and hepatitis C virus (HCV) infection. The relative risk (RR) of OSCC according to HCV infection was also estimated in the cohort. During the follow-up period, two men (1.3%) and seven women (2.9%) developed an OSCC. The SIR was 44.9 (95% CI: 20.5-85.2), being higher among women, but statistically significant in both genders. The RR of OSCC for patients with HCV as compared with those without HCV infection was 3.16 (0.8-12.5). Patients with OLP had a significantly increased risk of OSCC, irrespective of the clinical type of OLP and therapy. HCV infection apparently increased the risk for OSCC although this result could reflect the role of confounders, such as liver cirrhosis.

[Proportion of lung tumors attributable to occupation]. / La proporzione di tumori polmonari attribuibili all'occupazione.

This paper reviews estimates of the proportion of lung cancer attributable to occupational exposures and presents estimates for the City of Turin where a population-based case-control study on lung cancer has been conducted in 1991-1992 among the residents. An analysis of six hundred and five cases (505 males and 100 females) and 859 controls (684 males and 175 females) with histological/cytological confirmed diagnosis of lung cancer and occupational and smoking complete histories has been performed to estimate the proportion of lung cancers attributable to occupation in the city of Turin. The occupational exposure has been assessed with two different classifications: a) classification of a checklist of occupations as obtained from the study questionnaire according to industries and occupations known (List A) and suspected (List B) to be associated with lung cancer; b) translation of ISCO-ISIC codes for industries and occupations according to the same List A and List B. The population attributable risks are comparable when estimated with the two different classifications. In the city of Turin, at the beginning of the Nineties, about 10%-15% of lung cancers in males and 2%-5% in females could be attributed to occupational exposure.

[Database of occupations and industrial activities that involve the risk of pulmonary tumors]. / Database di occupazioni e attività industriali che comportano rischio di tumore del polmone.

Lists of occupations and industries associated with lung cancer are a useful tool to estimate attributable risks for occupational exposures. They were first published in 1982, based on IARC Monographs. List A included industries, processes and occupations definitely entailing carcinogenic risk, list B those probably/possibly carcinogenic. In 1995 the lists had been updated. We carried out a further update, reviewing IARC Monographs Volumes 62-75, focusing on Group 1, 2 A, and 2 B agents, and coded the lists according to widely used classifications: ISIC Rev. 2, ISIC Rev. 3, NACE Rev. 1, and ISTAT ATECO 1991 (for economic activities); ILO-ISCO 1968 and ISTAT-Classificazione delle professioni 1991 (for occupations). In order to evaluate temporal and geographical variations in attributable risk, exposure assessment must be performed consistently across different studies and standard tools to identify exposures, such as the one we propose here, are needed. The lists can also help to develop maps of industrial activities entailing carcinogenic risk at local, regional, and national level, and to identify economic activities that deserve priority action to control occupational exposures to carcinogens. The classifications were originally developed for economic and demographic purposes, and some exposure circumstances cannot be coded with sufficient specificity. This applies to productions or processes (i.e.: PVC production) that could be classified only by codes corresponding to wide groups of economic activities and/or occupations: in these instances we associated no code, so as not to inflate the estimates of exposed workers. As a consequence, however, certain exposures are not represented in the coded version of the lists. Even keeping in mind such limitations, coding makes the lists easier to apply, and increases the comparability of studies on lung cancer and occupation, as well as of surveys on exposure prevalence.

[Health effects of passive smoking]. / Effetti per la salute del fumo passivo.

Exposure to environmental tobacco smoke (ETS) is an established cause of lung cancer, ischaemic heart disease and chronic respiratory disease in adults and of asthma in children. A meta-analysis of 48 risk estimates from 39 published studies in non smokers gave a relative risk of 1.24 (95% Confidence Interval: 1.15-1.34), 1.14 (95% CI: 0.98-1.33) and 1.15 (95% CI: 0.92-1.43) for spousal, working environment and childhood ETS exposures. Misclassification of active smokers, exposure to ETS in reference groups, dietary confounding and publication bias have little effect on the association with spousal exposure. Estimates of the number of lung cancer deaths attributable to ETS from spousal exposure gave 1146 and 200 yearly deaths in the European Union (EU) and Italy respectively. Review of the available evidence confirms that exposure to ETS increases the risk for lung cancer and other diseases and is an important public health problem in the EU and Italy.

Socioeconomic factors associated with risk of upper aerodigestive tract cancer in Europe.

INTRODUCTION: In the European Union, there are 180,000 new cases of upper aerodigestive tract (UADT) cancer cases per year--more than half of whom will die of the disease. Socioeconomic inequalities in UADT cancer incidence are recognised across Europe. We aimed to assess the components of socioeconomic risk both independently and through their influence on the known behavioural risk factors of smoking, alcohol consumption and diet. PATIENTS AND METHODS: A multicentre case-control study with 2198 cases of UADT cancer and 2141 controls from hospital and population sources was undertaken involving 14 centres from 10 countries. Personal interviews collected information on demographics, lifetime occupation history, smoking, alcohol consumption and diet. Socioeconomic status was measured by education, occupational social class and unemployment. Odds ratios (ORs) and 95% confidence intervals (CIs) were computed using unconditional logistic regression. RESULTS: When controlling for age, sex and centre significantly increased risks for UADT cancer were observed for those with low versus high educational attainment OR=1.98 (95% CI 1.67, 2.36). Similarly, for occupational socioeconomic indicators--comparing the lowest versus highest International Socio-Economic Index (ISEI) quartile for the longest occupation gave OR=1.60 (1.28, 2.00); and for unemployment OR=1.64 (1.24, 2.17). Statistical significance remained for low education when adjusting for smoking, alcohol and diet behaviours OR=1.29 (1.06, 1.57) in the multivariate analysis. Inequalities were observed only among men but not among women and were greater among those in the British Isles and Eastern European countries than in Southern and Central/Northern European countries. Associations were broadly consistent for subsite and source of controls (hospital and community). CONCLUSION: Socioeconomic inequalities for UADT cancers are only observed among men and are not totally explained by smoking, alcohol drinking and diet.

Using directed acyclic graphs to consider adjustment for socioeconomic status in occupational cancer studies.

There is an ongoing debate on whether analyses of occupational studies should be adjusted for socioeconomic status (SES). In this paper directed acyclic graphs (DAGs) were used to evaluate common scenarios in occupational cancer studies with the aim of clarifying this issue. It was assumed that the occupational exposure of interest is associated with SES and different scenarios were evaluated in which (a) SES is not a cause of the cancer under study, (b) SES is not a cause of the cancer under study, but is associated with other occupational factors that are causes of the cancer, (c) SES causes the cancer under study and is associated with other causal occupational factors. These examples illustrate that a unique answer to the issue of adjustment for SES in occupational cancer studies is not possible, as in some circumstances the adjustment introduces bias, in some it is appropriate and in others both the adjusted and the crude estimates are biased. These examples also illustrate the benefits of using DAGs in discussions of whether or not to adjust for SES and other potential confounders.