Non-causal Explanations Research Articles

Cannabis use and risk of psychotic or affective mental health outcomes: a systematic review

Whether cannabis can cause psychotic or affective symptoms that persist beyond transient intoxication is unclear. We systematically reviewed the evidence pertaining to cannabis use and occurrence of psychotic or affective mental health outcomes. We searched Medline, Embase, CINAHL, PsycINFO, ISI Web of Knowledge, ISI Proceedings, ZETOC, BIOSIS, LILACS, and MEDCARIB from their inception to September, 2006, searched reference lists of studies selected for inclusion, and contacted experts. Studies were included if longitudinal and population based. 35 studies from 4804 references were included. Data extraction and quality assessment were done independently and in duplicate. There was an increased risk of any psychotic outcome in individuals who had ever used cannabis (pooled adjusted odds ratio=1.41, 95% CI 1.20-1.65). Findings were consistent with a dose-response effect, with greater risk in people who used cannabis most frequently (2.09, 1.54-2.84). Results of analyses restricted to studies of more clinically relevant psychotic disorders were similar. Depression, suicidal thoughts, and anxiety outcomes were examined separately. Findings for these outcomes were less consistent, and fewer attempts were made to address non-causal explanations, than for psychosis. A substantial confounding effect was present for both psychotic and affective outcomes. The evidence is consistent with the view that cannabis increases risk of psychotic outcomes independently of confounding and transient intoxication effects, although evidence for affective outcomes is less strong. The uncertainty about whether cannabis causes psychosis is unlikely to be resolved by further longitudinal studies such as those reviewed here. However, we conclude that there is now sufficient evidence to warn young people that using cannabis could increase their risk of developing a psychotic illness later in life.

Did intense adverse media publicity impact on prescribing of paroxetine and the notification of suspected adverse drug reactions? Analysis of routine databases, 2001–2004

To document the impact on clinical practice in England of media attention around possible adverse effects of paroxetine. Analysis of national selective serotonin reuptake inhibitor (SSRI) prescribing trends and yellow-card adverse drug reaction reports, 2001-2004. From a steady state in 2001, paroxetine prescribing declined sharply from April 2002, coinciding with a USA regulatory action; the subsequent decline in paroxetine prescribing was 1.87% per month (95% confidence interval - 2.06, -1.68). Other SSRI prescribing increased by 1% per month until a major UK review of SSRIs in children in December 2003, after which prescribing plateaued. Media publicity was associated with short-term peaks in yellow-card reports related to paroxetine. Falls in paroxetine and other SSRI prescribing in the UK coincided, respectively, with regulatory communications from the USA and the UK, but associations may have noncausal or other explanations. Reports of adverse reactions to paroxetine appeared to increase after adverse media publicity about the drug.

The Bradford Hill considerations on causality: a counterfactual perspective

Bradford Hill's considerations published in 1965 had an enormous influence on attempts to separate causal from non-causal explanations of observed associations. These considerations were often applied as a checklist of criteria, although they were by no means intended to be used in this way by Hill himself. Hill, however, avoided defining explicitly what he meant by "causal effect".This paper provides a fresh point of view on Hill's considerations from the perspective of counterfactual causality. I argue that counterfactual arguments strongly contribute to the question of when to apply the Hill considerations. Some of the considerations, however, involve many counterfactuals in a broader causal system, and their heuristic value decreases as the complexity of a system increases; the danger of misapplying them can be high. The impacts of these insights for study design and data analysis are discussed. The key analysis tool to assess the applicability of Hill's considerations is multiple bias modelling (Bayesian methods and Monte Carlo sensitivity analysis); these methods should be used much more frequently.

Breast‐feeding and childhood cancer: A systematic review with metaanalysis

It has been suggested that breast milk may play a role in the prevention of certain childhood cancers. We undertook a systematic review of published studies investigating the association between breast-feeding and childhood cancers using Medline (1966 to June 2004), supplemented with auto alerts and manual searches. Analyses are based on odds ratios for specific cancers among those ever breast-fed compared with those never breast-fed, pooled using random-effects models. Forty-nine publications were potentially relevant; of these, 26 provided odds ratio estimates for at least one childhood cancer outcome and were included in metaanalyses. Overall, 92% of the studies were case-control studies, 85% relied on long-term recall of feeding history, only 8% examined breast-feeding exclusivity and control response rates were under 80% in over half. Metaanalyses suggested lower risks associated with having been breast-fed of 9% (95% CI = 2-16%) for acute lymphoblastic leukemia, 24% (3-40%) for Hodgkin's disease and 41% (22-56%) for neuroblastoma, with little between-study heterogeneity. The estimates for Hodgkin's disease and neuroblastoma, however, were driven by single studies. There was little evidence that breast-feeding was associated with acute nonlymphoblastic leukemia, non-Hodgkin's lymphoma, central nervous system cancers, malignant germ cell tumors, juvenile bone tumors, or other solid cancers. In conclusion, ever having been breast-fed is inversely associated with acute lymphoblastic leukemia, Hodgkin's disease and neuroblastoma in childhood, but noncausal explanations are possible. Even if causal, the public health importance of these associations may be small. Our estimates suggest that increasing breast-feeding from 50% to 100% would prevent at most 5% of cases of childhood acute leukemia or lymphoma. (c) 2005 Wiley-Liss, Inc.

Interrelation of erectile dysfunction and lower urinary tract symptoms

The association between erectile dysfunction (ED) and lower urinary tract symptoms (LUTS) has garnered attention as investigators have hypothesized a common pathophysiology to explain the assertion that they are causally linked. This relationship between LUTS and ED has received increased attention because both diseases are highly prevalent, frequently co-associate in the same aging male group, and significantly influence the overall quality of life. A causal association between LUTS and ED cannot be established on the basis of the ever-increasing number of epidemiological studies. Attempting to explain a causal relationship between ED and LUTS requires the use of Hill's criterion, which is used by many epidemiologists to separate causal from non-causal explanations. A review of the epidemiological evidence reveals a strong degree of association, internal consistency and dose response effects between ED and LUTS. The temporal relationships between the two remain unknown because of the strong cross-sectional flavor of the epidemiological studies. The issue of an "alternate explanation" to describe the LUTS-ED association appears to be accounted for in that several large studies have provided convincing multiple regression analyses in which the ED-LUTS relationship remains significant. Biologic plausibility is an important issue if the link between ED and LUTS is to have credence. There are four leading theories of how these diseases interrelate. These explanations have a variable amount of supporting data. These include: 1) autonomic hyperactivity effects on LUTS, prostate growth and ED; 2) nitric oxide synthase/nitric oxide levels decreased or altered in the prostate and penile smooth muscle; 3) prostate and penile ischemia; and 4) increased Rho-kinase activation/ endothelin activity. LUTS and sexual dysfunction are highly prevalent in aging men. Both conditions are also significant contributors to overall quality of life. New data has emerged to indicate potential links in epidemiological, physiologic, pathophysiologic and treatment aspects of these two diseases. Using Hill's causality method to separate causal from non-causal explanations linking ED with LUTS, it appears that most epidemiological components are fulfilled.

RELATION OF INDIVIDUAL TRIHALOMETHANE CONCENTRATIONS IN PUBLIC WATER SUPPLIES TO STILLBIRTH AND BIRTH WEIGHT PREVALENCE IN THREE WATER REGIONS

ISEE-242 Introduction: This study, in three water regions, is the largest yet conducted of the association between disinfection by-products in the public water supply, as measured by individual trihalomethanes (THMs), and stillbirth and birth weight prevalence. Methods: Modelled estimates of quarterly individual THM concentrations in water zones by year, categorised as low (for chloroform, >20 μg/l; for bromodichloromethane (BDCM) and total brominated THMs (TBTHM), >6 μg/l), medium (for chloroform, 20–40 μg/l; for BDCM and TBTHM, 6–12 μg/l) or high (for chloroform, <40 μg/l; for BDCM and TBTHM, <12 μg/l), were linked to 1,015,298 routine birth and still birth records using maternal residence at time of birth through the use of Geographical Information Systems. Results: In one region, a trend of increasing prevalence of stillbirth and low and very low birth weight from low to medium to high chloroform exposure areas was observed — for example, with adjustment for potential confounders including maternal age and social deprivation, risks of stillbirth and low birth weight in high chloroform compared with low chloroform exposure areas were 1.15 (95% CI 1.01 to 1.32) and 1.17 (95% CI 1.13 to 1.22) respectively. There was no significant relation between chloroform and excess adverse birth outcomes in the other regions. In a meta-analysis across the three regions, small, but significant, excess risks were found in areas with high chloroform concentrations for both stillbirths (OR 1.12; 95% CI 1.02 to 1.23) and low birth weight (OR 1.10; 95% CI 1.07 to 1.13); no associations were found with very low birth weight. Concentrations of BDCM and TBTHMs did not show any significant association with risk of stillbirths, low and very low birth weight in all three regions separately and combined. Analysis of stillbirth risk associated with specific THMs was also conducted by cause-of-death sub-category. Discussion: These findings suggest a need to focus on specific individual (as opposed to total) THMs and/or other disinfection by-products in relation to stillbirth and low birth weight risk. This type of focused research, which is very under-developed at present, informs cross-study comparisons and helps to refine risk estimates in order to differentiate between possible causal and alternative (non-causal) explanations.

Family history of colon cancer: What does it mean and how is it useful?

Background Family history of colon cancer can be deconstructed into causal and noncausal explanations, which include genetic factors, environmental factors, gene–environment interactions, misclassification, and differences in screening. Methods We investigated some of these causal and noncausal explanations by using data from a case–control study of colon cancer conducted among African Americans and whites in North Carolina. We examined the relationship between family history and polymorphisms in four genes (N-acetyltransferase 1 and 2 [NAT1, NAT2], methylenetetrahydrofolate reductase, and peroxisome proliferator-activated receptor gamma [PPARG]), environmental risk factors, the joint distributions of these genes and environmental risk factors, and the prevalence of colon cancer screening. Results Participants with one or more first-degree relatives with colon cancer showed a slightly higher prevalence of at-risk genotypes for each locus, but results were statistically significant only for NAT2. Participants with a family history showed a higher prevalence of at-risk combinations of genotypes and environmental risk factors ( NAT2 and well-done red meat consumption; PPARG and nonsteroidal anti-inflammatory medication use). The sensitivity and predictive value of family history for identifying persons with at-risk genotypes or environmental risk factors was low. History of cancer screening was similar in those with and without a family history. Conclusions Our results suggest that family history of colon cancer may represent aggregation of some genetic polymorphisms and environmental risk factors. Although it is premature to use family history as a screening tool when testing for genetic polymorphisms, further research is needed to identify additional genes and environmental factors that may be associated with family history.

Lung cancer risk in a population-based cohort of patients hospitalized for asthma in Sweden.

It has been suggested that asthma increases the risk of lung cancer in males but not in females. However, previous studies may suffer from report bias and are based on a small numbers of cases. The objective of the present study was to assess the incidence of lung cancer in males and females using a nationwide Swedish cohort of asthma patients. Patients (n=92,986) aged > or = 20 yrs with a hospital-discharge diagnosis of asthma and who were alive and free from malignancy I yr after first hospitalization were followed-up, for incidence of lung cancer during the period 1965-1994 (average duration of follow-up, 8.5 yrs). Their incidence of lung cancer was compared with that of the national population. The authors observed 713 lung cancers (standardized incidence ratio (SIR) 1.58, 95% confidence interval (CI) 1.47-1.70). The SIR was 1.51 in males (95% CI 1.38-1.65, 492 cases) and 1.78 in females (95% CI 1.55-2.03, 221 cases). The SIR decreased with duration of follow-up and increased with calendar period and age at first hospitalization. The risk of lung cancer was higher for squamous cell and small cell carcinoma than for adenocarcinoma, and it was higher in patients with other diseases as the main diagnosis and in patients hospitalized in departments other than internal and respiratory medicine. It was confirmed that asthma patients are at increased risk of lung cancer, but there is no heterogeneity in risk between the sexes. Several indirect arguments point towards a noncausal explanation of these findings; in particular, confounding by tobacco smoking is a plausible explanation.

Age at First Drink and Risk for Alcoholism: A Noncausal Association

Prior research indicates risk for alcoholism is increased among individuals who begin to drink at an early age. We replicate and extend these findings, addressing causal and noncausal explanations for this association. Structured psychiatric interviews, including assessment of lifetime DSM-IV alcohol abuse and alcohol dependence (AD), were conducted with 8746 adult twins ascertained through a population-based twin registry. We found strong evidence for an association between early drinking onset and risk for AD, but less evidence for an association with alcohol abuse. The results of twin-pair analyses suggest that all of the association between early drinking and later AD is due to familial sources, which probably reflect both shared environmental and genetic factors. These results suggest the association between drinking onset and diagnosis is noncausal, and attempts to prevent the development of AD by delaying drinking onset are unlikely to be successful.

Maternal smoking and twinning

In order to investigate a possible association between maternal smoking during pregnancy and twinning, information on 1,096,330 single births and 12,342 twin births in 1983-95 was obtained from the Swedish Medical Birth Registry (MBR). All odds ratios (OR) were estimated after stratification for year of birth and maternal age, parity, and educational level. Smoking women, compared with non-smoking women, were at increased risk of having dizygotic (DZ) twins, but the risk increase was only evident among multiparas. A strong association between previous involuntary childlessness and dizygotic (DZ) twinning (especially in primiparas) was found. The strongest association between maternal smoking and DZ twinning was found among multiparas without any history of involuntary childlessness (OR: 1.35, 95%CI:1.22-1.49), whereas among women who had experienced involuntary childlessness, the opposite was seen (OR: 0.82, 95%CI:0.66-1.00, no difference between parity strata). Weinberg's differential method was used to estimate the number of monozygotic (MZ) twins, and a method of estimating stratified ORs among mothers of MZ twins was presented. No association was found between MZ twinning and maternal smoking (OR: 0.96, 95%CI:0.86-1.07), and no confounding by parity or previous involuntary childlessness was indicated. Several non-causal explanations to the positive association between DZ twinning and maternal smoking among multiparas were discussed, but homogeneity over strata indicated that maternal smoking may be a true risk factor for double ovulation.

Maternal smoking and twinning

AbstractIn order to investigate a possible association between maternal smoking during pregnancy and twinning, information on 1 096 330 single births and 12 342 twin births in 1983–95 was obtained from the Swedish Medical Birth Registry (MBR). All odds ratios (OR) were estimated after stratification for year of birth and maternal age, parity, and educational level. Smoking women, compared with non-smoking women, were at increased risk of having dizygotic (DZ) twins, but the risk increase was only evident among multiparas. A strong association between previous involuntary childlessness and dizygotic (DZ) twinning (especially in primiparas) was found. The strongest association between maternal smoking and DZ twinning was found among multiparas without any history of involuntary childlessness (OR: 1.35, 95%CI:1.22–1.49), whereas among women who had experienced involuntary childlessness, the opposite was seen (OR: 0.82, 95%CI:0.66–1.00, no difference between parity strata). Weinberg's differential method was used to estimate the number of monozygotic (MZ) twins, and a method of estimating stratified ORs among mothers of MZ twins was presented. No association was found between MZ twinning and maternal smoking (OR: 0.96, 95%CI:0.86–1.07), and no confounding by parity or previous involuntary childlessness was indicated. Several non-causal explanations to the positive association between DZ twinning and maternal smoking among multiparas were discussed, but homogeneity over strata indicated that maternal smoking may be a true risk factor for double ovulation.

Does diesel exhaust cause human lung cancer?

Recent reviews of epidemiological evidence on the relation between exposure to diesel exhaust (DE) and lung cancer risk have reached conflicting conclusions, ranging from belief that there is sufficient evidence to conclude that DE is a human lung carcinogen (California EPA, 1994) to conclusions that there is inadequate evidence to support a causal association between DE and human lung cancer (Muscat and Wynder, 1995). Individual studies also conflict, with both increases and decreases in relative risks of lung cancer mortality being cited with 95% statistical confidence. On balance, reports of elevated risk outnumber reports of reduced risk. This paper reexamines the evidence linking DE exposures to lung cancer risk. After briefly reviewing animal data and biological mechanisms, it surveys the relevant epidemiological literature and examines possible explanations for the discrepancies. These explanations emphasize the distinction between statistical associations, which have been found in many studies, and causal associations, which appear not to have been established. Methodological threats to valid causal inference are identified and new approaches for controlling them are proposed using recent techniques from artificial intelligence (AI) and computational statistics. These threats have not been adequately controlled for in previous epidemiological studies. They provide plausible noncausal explanations for the reported increases in relative risks, making it impossible to infer causality between DE exposure and lung cancer risk from these studies. A key contribution is to show how recent techniques developed in the AI-and-statistics literature can help clarify the causal interpretation of complex multivariate data sets used in epidemiological risk assessments. Applied to the key study of Garshick et al. (1988), these methods show that DE concentration has no positive causal association with occupational lung cancer mortality risk.

Mortality Associated With Low Plasma Concentration of Beta Carotene and the Effect of Oral Supplementation

To examine the relationship between beta carotene plasma concentration and beta carotene supplementation and risk of death from major disease causes. Cohort study of plasma concentrations; randomized, controlled clinical trial of supplementation. Medical school-affiliated dermatology practices. A total of 1188 men and 532 women with mean age of 63.2 years, who had enrolled in a randomized clinical trial of beta carotene supplementation to prevent nonmelanoma skin cancer. Oral beta carotene, 50 mg per day for a median of 4.3 years. All-cause mortality and mortality from cardiovascular disease and cancer. During a median follow-up period of 8.2 years, there were 285 deaths. Persons whose initial plasma beta carotene concentrations were in the highest quartile (>0.52 micromol/L [27.7 microg/dL]) had a lower risk of death from all causes (adjusted relative rate [RR], 0.52; 95% confidence interval [CI] 0.44 to 0.87) and from cardiovascular diseases (adjusted RR, 0.57; 95% CI, 0.34 to 0.95) compared with persons with initial concentrations in the lowest quartile (<0.21 micromol/L [11.2 microg/dL]). Patients randomly assigned to beta carotene supplementation showed no reduction in relative mortality rates from all causes (adjusted RR, 1.03; 95% CI, 0.82 to 1.30) or from cardiovascular disease (adjusted RR, 1.16; 95% CI, 0.82 to 1.64). There was no evidence of lower mortality following supplementation among patients with initial beta carotene concentrations below the median for the study group. These analyses provide no support for a strong effect of supplemental beta carotene in reducing mortality from cardiovascular disease or other causes. Although the possibility exists that beta carotene supplementation produces benefits that are too small or too delayed to have been detected in this study, noncausal explanations should be sought for the association between plasma concentrations of beta carotene and diminished risk of death.

Meta-analysis of the role of oral contraceptive agents in inflammatory bowel disease.

Numerous epidemiological studies have been performed to determine factors that might contribute to the development of inflammatory bowel disease. Although the role of oral contraceptive agents in Crohn's disease (CD) and ulcerative colitis (UC) have been assessed, most studies were of small sample size and characterised by low statistical precision. A meta-analysis was performed to increase the statistical power and to investigate the association between the use of oral contraceptives and the development of CD and UC. The study was based on a search of a Medline database from 1975 to October 1993 and a review of reference lists from published articles, reviews, symposia proceedings, and abstracts from major gastrointestinal meetings. All studies specifically designed to evaluate this association were selected. The combined results of nine studies--two cohort studies (30,379 unexposed and 30,673 exposed patients) and seven case-control studies (482 CD, 237 UC, and 3198 controls)--which satisfied our selection criteria were evaluated. The pooled relative risk (adjusted for smoking) associated with oral contraceptive use was 1.44 (1.12, 1.86) for CD and 1.29 (0.94, 1.77) for UC. These results suggest modest associations between the use of oral contraceptives and the development of CD and UC. As these associations are weak, non-causal explanations for the findings cannot be eliminated.

The Role of Drugs in Workplace Injuries: Is Drug Testing Appropriate?

In the past decade, many employers have adopted drug-testing programs to reduce workplace injuries. However, little scientific evidence shows that drug use is a significant and substantial cause of total workplace injuries. The purpose of this study was to empirically assess the role of drugs and alcohol in causing workplace injuries. Questionnaires were received from 882 Ontario employees in a household survey. Results showed that many variables were significantly related to job injuries. The variables were ranked according to the relative importance of each variable's contribution to total job injuries. Alcohol problems, licit drug use, and illicit drug use ranked 7, 11, and 12 respectively among a group of 12 significant variables. In order to examine the likelihood that drug use was a cause of job injuries, the relationship between job injuries and alcohol problems, licit drug use, and illicit drug use was examined across categories of third variables. For age, the relationship between drug use and injuries remained strong for the youngest age group, but disappeared for the oldest age group. Logistic regression analysis confirmed the plausibility of noncausal explanations of job injuries for illicit drug use, but not for alcohol problems or licit drug use. Overall, the results indicated that illicit drug use does not appear to be a major cause of job injuries. The implications of these results for drug-testing programs are discussed.

A Counterfactual Theory of Causal Explanation

We commonly explain certain occurrences in the natural world, on the basis of a shared stock of information available to the non-specialised layman as well as to the natural scientist. That shared stock of information is our folk physics. The snow melted because spring had finally come; the car was wrecked because it was hit by a bus; the match lit because it was struck. Such explanations are, or appear to be, causal explanations. Many of our other folk explanations may be noncausal explanations, but I consider only the former sort. I assume, following Davidson, that laws are strict, i.e., exceptionless. If it is a deterministic law that Fs are Gs, then every F without exception must be a G. If it is a stochastic law that Pr(G, F)=p, where O<p<1, then, for any finite sample, s, it may well be that the proportion of Fs in s which are Gs may not be exactly p. However, as one examines Fs indefinitely, the proportion which are found to be Gs will tend to p, and in the limit, the proportion of Fs which are Gs will be

A simplified logic of causal inference.

This paper provides a simplified method for evaluating the evidence in favour of a causal claim. It analyses the evidence bearing upon such a claim in terms of two questions: Do the putative cause and effect covary, and can alternative non-causal explanations of the relationship be ruled out? The different research designs for assessing covariation are outlined, as are the ways in which these designs permit a researcher to decide between alternative explanations of the relationship.

CHLORAMPHENICOL USE AND CHILDHOOD LEUKAEMIA IN SHANGHAI

A population-based case-control interview study of 309 childhood leukaemia cases and 618 age and sex matched controls showed a significant dose-response relation between chloramphenicol and risk of both acute lymphocytic leukaemia (ALL) and acute non-lymphocytic leukaemia (ANLL), treatment for more than 10 days being associated with risks of 11.0 and 12.0, respectively. A significant risk of ANLL was also observed with the use of syntomycin, a drug pharmacologically related to chloramphenicol. The risks remained high when analyses were limited to either first or latest use of these antibiotics more than 2 years before diagnosis. Although the association may have non-causal explanations, the results warrant cautious prescribing patterns and further investigations into the leukaemogenic potential of chloramphenicol.

A Type of Non-Causal Explanation1

There is a type of explanation frequently employed in the sciences, particularly the quantitative ones, that I intend to focus on. It seeks to understand some phenomenon or regularity that occurs in a type of system S by deriving an equation describing that phenomenon or regularity from a set of equations that govern S. I want to propose a way of understanding such explanations.

Parental reports of depressives: An investigation of several explanations

Associations linking parental reports of low care and/or overprotection with several grades of depressive experience have been reported in a number of studies. Three non-causal explanations of those associations are examined in the present paper: that depression levels might influence judgement of parental characteristics, that reported parental characteristics may not reflect actual parental characteristics, and that those with a potentially depressive temperament of dependency might elicit less parental care and greater parental overprotection. Varying levels of depression in a clinical group did not influence scores on a self-report measure of parental characteristics, the Parental Bonding Instrument (PBI). Subjects' scores on the PBI (i.e. perceived parental characteristics) correlated significantly with scores returned by family observers and judges in 2 separate studies, supporting the validity of the PBI as a measure of actual parental characteristics. Depression scores in subjects were strongly linked with lower maternal care and with maternal overprotection, whether the maternal characteristics were judged by the subjects or by the mothers themselves. There was no evidence to support the view that those with marked dependency traits elicit different parental responses.