On making causal claims: A review and recommendations

Social scientists often estimate models from correlational data, where the independent variable has not been exogenously manipulated; they also make implicit or explicit causal claims based on these models. When can these claims be made? We answer this question by first discussing design and estimation conditions under which model estimates can be interpreted, using the randomized experiment as the gold standard. We show how endogeneity--which includes omitted variables, omitted selection, simultaneity, common methods bias, and measurement error--renders estimates causally uninterpretable. Second, we present methods that allow researchers to test causal claims in situations where randomization is not possible or when causal interpretation is confounded, including fixed-effects panel, sample selection, instrumental variable, regression discontinuity, and difference-in-differences models. Third, we take stock of the methodological rigor with which causal claims are being made in a social sciences discipline by reviewing a representative sample of 110 articles on leadership published in the previous 10 years in top-tier journals. Our key finding is that researchers fail to address at least 66 % and up to 90 % of design and estimation conditions that make causal claims invalid. We conclude by offering 10 suggestions on how to improve non-experimental research.

Impact of tamoxifen dose on tamoxifen and its active metabolites exposure in breast cancer patients: preliminary results from a prospective, open-label trial

Background: CYP2D6 is the key enzyme responsible for tamoxifen bioactivation mainly into endoxifen. This gene is highly polymorphic and breast cancer patients classified as CYP2D6 poor metabolizers (PM) or intermediate metabolizers (IM) appear to show low concentrations of endoxifen and to achieve less benefit from tamoxifen treatment. Purpose: This prospective, open-label trial aimed to assess how the increase of tamoxifen dose influences the level of endoxifen in the different genotype groups (poor-, intermediate-, and extensive-metabolizers (EM)). We examined the impact of doubling tamoxifen dose to 20mg twice daily on endoxifen plasma concentrations across these genotype groups. Patients and methods: Patients were assayed for CYP2D6 genotype and phenotype using dextromethorphan test. Tamoxifen, N-desmethyltamoxifen, 4-hydroxytamoxifen and endoxifen plasma levels were determined on 2 occasions at baseline (20mg/day of tamoxifen) and at day 30, 90 and 120 after dose increase (20 mg twice daily) using liquid chromatography-tandem-mass spectrometry. Endoxifen plasma levels were measured 6 to 24 hours after last drug intake to evaluate its accumulation before and after doubling tamoxifen dosage. ANOVA was used to evaluate endoxifen levels increase and difference between genotype groups. Results: 63 patients are available for analysis to date. Tamoxifen, N-desmethyltamoxifen, 4-hydroxytamoxifen and endoxifen plasma reached steady state at 30 day after tamoxifen dose escalation, with a significant increase compared to baseline by 1.6 to 1.8 fold : geometric mean plasma concentrations (CV %) were 140 ng/mL (45%) at baseline vs 255 (47%) at day 30 for tamoxifen (P < 0.0001); 256 (49%) vs 408 (64%) for N-desmethyltamoxifen (P < 0.0001); 2.4 (46%) vs 3.9 (51%) for 4-OH-tamoxifen (P < 0.0001); and 20 (91%) vs 33 (91%) for endoxifen (P < 0.02). On baseline, endoxifen levels tended to be lower in PM: 7 ng/mL (36%), than IM: 16 ng/mL (70%), P=0.08, and EM: 24 ng/mL (71%), P<0.001. After doubling tamoxifen dosage, endoxifen concentrations rose similarly in PM, IM and EM with respectively, 1.5 (18%), 1.5 (28%) and 1.7 (30%) fold increase from baseline, P=0.18. Conclusion: Endoxifen exposure varies widely under standard tamoxifen dosage, with CYP2D6 genotype explaining only a minor part of this variability. It increases consistently on doubling tamoxifen dose, similarly across genotypes. This would enable exposure optimization based on concentration monitoring.

On making causal claims: A review and recommendations

Social scientists often estimate models from correlational data, where the independent variable has not been exogenously manipulated; they also make implicit or explicit causal claims based on these models. When can these claims be made? We answer this question by first discussing design and estimation conditions under which model estimates can be interpreted, using the randomized experiment as the gold standard. We show how endogeneity--which includes omitted variables, omitted selection, simultaneity, common methods bias, and measurement error--renders estimates causally uninterpretable. Second, we present methods that allow researchers to test causal claims in situations where randomization is not possible or when causal interpretation is confounded, including fixed-effects panel, sample selection, instrumental variable, regression discontinuity, and difference-in-differences models. Third, we take stock of the methodological rigor with which causal claims are being made in a social sciences discipline by reviewing a representative sample of 110 articles on leadership published in the previous 10 years in top-tier journals. Our key finding is that researchers fail to address at least 66 % and up to 90 % of design and estimation conditions that make causal claims invalid. We conclude by offering 10 suggestions on how to improve non-experimental research.

Genome size rather than content might affect call properties in toads of three ploidy levels (Anura: Bufonidae: Bufo viridis subgroup).

In vertebrates, genome size has been shown to correlate with nuclear and cell sizes, and influences phenotypic features, such as brain complexity. In three different anuran families, advertisement calls of polyploids exhibit longer notes and intervals than diploids, and difference in cellular dimensions have been hypothesized to cause these modifications. We investigated this phenomenon in green toads (Bufo viridis subgroup) of three ploidy levels, in a different call type (release calls) that may evolve independently from advertisement calls, examining 1205 calls, from ten species, subspecies, and hybrid forms. Significant differences between pulse rates of six diploid and four polyploid (3n, 4n) green toad forms across a range of temperatures from 7 to 27 °C were found. Laboratory data supported differences in pulse rates of triploids vs. tetraploids, but failed to reach significance when including field recordings. This study supports the idea that genome size, irrespective of call type, phylogenetic context, and geographical background, might affect call properties in anurans and suggests a common principle governing this relationship. The nuclear-cell size ratio, affected by genome size, seems the most plausible explanation. However, we cannot rule out hypotheses under which call-influencing genes from an unexamined diploid ancestral species might also affect call properties in the hybrid-origin polyploids.

Assessment of skinfold thickness equations in estimating body composition in children with inflammatory bowel

Introduction: Growth is a central process in paediatrics. Weight and height evaluation are therefore routine exams for every child but in some situation, particularly inflammatory bowel disease (IBD), a wider evaluation of nutritional status needs to be performed. The assessment of body composition is essential in order to maintain acceptable growth using the following techniques: Dual-energy X-ray absorptiometry (DEXA), bio-impedance-analysis (BIA) and anthropometric measurements (skinfold thickness skin), the latter being most easily available and most cost effective. Objectives: To assess the accuracy of skinfold equations in estimating percentage body fat (%BF) in children with inflammatory bowel disease (IBD), compared with assessment of body fat dual energy X-ray absorptiometry (DEXA). Methods: Twenty-one patients (11 females, 10 males; mean age: 14.3 years, range 12 - 16 years) with IBD (Crohn's disease n = 15, ulcerative colitis n = 6)). Estimated%BF was computed using 6 established equations based on the triceps, biceps, subscapular and suprailiac skinfolds (Deurenberg, Weststrate, Slaughter, Durnin & Rahaman, Johnston, Brook) and compared to DEXA. Concordance analysis was performed using Lin's concordance correlation and the Bland-Altman limits of agreement method. Results: Durnin & Rahaman's equation shows a higher Lin's concordance coefficient with a small difference amongst raw values for skinfolds and DEXA compared to the other equations. Correlation coefficient between mean and difference is close to zero with a non-significant Bradley-Blackwood test. Conclusion: Body composition in paediatric IBD patients using the Durnin & Rahaman skinfold-equation adequately reflects values obtained by DEXA.

Apparent versus true gene expression changes of three hypoxia-related genes in autopsy derived tissue and the importance of normalisation.

The aim of our work was to show how a chosen normal-isation strategy can affect the outcome of quantitative gene expression studies. As an example, we analysed the expression of three genes known to be upregulated under hypoxic conditions: HIF1A, VEGF and SLC2A1 (GLUT1). Raw RT-qPCR data were normalised using two different strategies: a straightforward normalisation against a single reference gene, GAPDH, using the 2(-ΔΔCt) algorithm and a more complex normalisation against a normalisation factor calculated from the quantitative raw data from four previously validated reference genes. We found that the two different normalisation strategies revealed contradicting results: normalising against a validated set of reference genes revealed an upregulation of the three genes of interest in three post-mortem tissue samples (cardiac muscle, skeletal muscle and brain) under hypoxic conditions. Interestingly, we found a statistically significant difference in the relative transcript abundance of VEGF in cardiac muscle between donors who died of asphyxia versus donors who died from cardiac death. Normalisation against GAPDH alone revealed no upregulation but, in some instances, a downregulation of the genes of interest. To further analyse this discrepancy, the stability of all reference genes used were reassessed and the very low expression stability of GAPDH was found to originate from the co-regulation of this gene under hypoxic conditions. We concluded that GAPDH is not a suitable reference gene for the quantitative analysis of gene expression in hypoxia and that validation of reference genes is a crucial step for generating biologically meaningful data.

Sex difference and the role of leptin in the association between high-sensitivity C-reactive protein and adiposity in two different populations.

Elevated high-sensitivity C-reactive protein (hs-CRP) concentration is associated with an increased risk of cardiovascular disease but this association seems to be largely mediated via conventional cardiovascular risk factors. In particular, the association between hs-CRP and obesity has been extensively demonstrated and correlations are stronger in women than men. We used fractional polynomials-a method that allows flexible modeling of non linear relations-to investigate the dose/response mathematical relationship between hs-CRP and several indicators of adiposity in Caucasians (Switzerland) and Africans (Seychelles) surveyed in two population-based studies. This relationship was non-linear exhibiting a steeper slope for low levels of hs-CRP and a higher level in women. The observed sex difference in the relationship between hs-CRP and adiposity almost disappeared upon adjustment for leptin, suggesting that these sex differences might be partially mediated, by leptin. All these relationship were similar in Caucasians and Africans. This is the first report on a non-linear relation, stratified by gender, between hs-CRP and adiposity.

Violent adolescents and their educational environment : a multilevel analysis ; et, Youths carrying a weapon or using a weapon in a fight : what makes the difference ?

Rapport de synthèse : Cette recherche s'intéresse (1) au port et à l'utilisation d'armes chez les adolescents ainsi que (2) aux rôles des facteurs environnementaux et individuels dans la violence juvénile. Les données étaient tirés de SMASH 2002 (Swiss multicenter adolescent survey on health 2002), étude dans laquelle un échantillon représentatif de 7548 étudiants et apprentis âgés entre 16 et 20 ans vivant en Suisse ont été interrogés Dans une première étude, les adolescents ayant porté une arme (couteau, masse, coup de poing américain, pistolet/autre arme à feu, spray) durant l'année précédant l'enquête étaient comparés avec ceux n'ayant pas porté d'arme. Ensuite, dans le sous-échantillon de porteurs d'armes, ceux ayant uniquement porté l'arme étaient comparés avec ceux ayant utilisé une arme dans une bagarre. Des facteurs individuels, familiaux, scolaires et sociaux ont été étudiés à l'aide d'analyses bivariées et multivariées. 13.7% des jeunes vivant en Suisse ont porté une arme dans l'année précédant l'enquête. 6.2% des filles porteuses d'armes et 19.9% des garçons porteurs d'armes ont fait usage de l'arme dans une bagarre. Chez les garçons et chez les filles, les porteurs d'armes étaient plus souvent délinquants et victimes de violence physique. Les garçons porteurs d'armes étaient plus souvent des apprentis, à la recherche de sensations fortes, porteurs de tatouages, avaient une mauvaise relation avec leurs parents, étaient dans des bagarres sous l'influence de substances, et avaient des relations sexuelles à risque. En comparaison avec les porteuses d'armes, les filles utilisatrices d'armes étaient plus souvent fumeuses quotidiennes. Les garçons ayant utilisé leur arme étaient plus souvent nés à l'étranger, vivaient dans un milieu urbain, étaient des apprentis, avaient un mauvais contexte scolaire, avaient des relations sexuelles à risque et étaient impliqués dans des bagarres sous l'influence de substances. Nos résultats montrent que porter une arme est un comportement relativement fréquent chez les adolescents vivant en Suisse et qu'une proportion non négligeable de ces porteurs d'armes ont utilisé l'arme dans une bagarre. De ce fait, une discussion sur le port d'arme devrait être incluse dans l'entretien clinique ainsi que dans les programmes de prévention visant les adolescents. Dans une deuxième étude, la violence juvénile était définie comme présente si l'adolescent avait commis au moins un des quatre délits suivants durant l'année précédant l'enquête: attaquer un adulte, arracher ou voler quelque chose, porter une arme ou utiliser une arme dans une bagarre. Des niveaux écologiques étaient testés et résultaient en un modèle à trois niveaux pour les garçons (niveau individuel, niveau classe et niveau école) et, à cause d'une faible prévalence de la violence chez les filles, en un modèle à un niveau (individuel) pour les filles. Des variables dépendantes étaient attribuées à chaque niveau, en se basant sur la littérature. Le modèle multiniveaux des garçons montrait que le niveau école (10%) et le niveau classe (24%) comptaient pour plus d'un tiers de la variance inter-individuelle dans le comportement violent. Les facteurs associés à ce comportement chez les filles étaient être victime de violence physique et la recherche de sensations fortes. Pour les garçons, les facteurs explicatifs de la violence étaient pratiquer des relations sexuelles à risque, être à la recherche de sensations fortes, être victime de violence physique, avoir une mauvaise relation avec les parents, être déprimé et vivre dans une famille monoparentale au niveau individuel, la violence et les actes antisociaux au niveau de la classe et être apprenti au niveau de l'école. Des interventions au niveau de la classe ainsi qu'un règlement explicit en ce qui concerne la violence et d'autres comportements à risque dans des écoles devraient être prioritaires pour la prévention de la violence chez les adolescents. En outre, la prévention devrait tenir compte des différences entre les sexes.

The difference between the basal metabolic rate and the sleeping metabolic rate in Japanese.

Previous studies have demonstrated the difference between the basal metabolic rate (BMR) and the sleeping metabolic rate (SMR): however, the difference in the Japanese population has not yet been explored. This study examined the relationship between the BMR and SMR in ninety-four healthy Japanese subjects (37 males and 57 females, 39 +/- 12 y of age and 22.0 +/- 7.4% body fat) in a respiratory chamber. The SMR was significantly lower than the BMR (1416 +/- 245 vs. 1492 +/- 256 kcal/d): however, there was a highly significant correlation between the two (r = 0.867; p < 0.001). The ratio of SMR/BMR largely varied among individuals (0.95 +/-0.08, 8.4% of the coefficient of variation). The ratio was significantly lower in males than in females (0.93 +/- 0.10 vs. 0.97 +/- 0.06, p < 0.05). None of the anthropometric measures (age, weight, body mass index, body surface area or percent body fat) correlated with the ratio. These results showed that SMR was 95%, of BMR on average in a healthy Japanese group. However, when applied over a longer time period (24 h or more), the difference tends to become negligible for most analyses in a group. Although the difference between SMR and BMR will induce a 5% gap of physical activity level defined as the total energy expenditure divided by the BMR or SMR, this factor seems to have little practical importance in epidemiological research.

Gender difference in the response to an angiotensin-converting enzyme inhibitor and a diurectic in hypertensive patients of African descent

Résumé de la thèse en français Titre : Différence entre hommes et femmes dans la réponse à un inhibiteur de l'enzyme de conversion de l'angiotensine et un diurétique chez des patients hypertendus d'origine Africaine. Introduction: L'efficacité des inhibiteurs de l'enzyme de conversion de l'angiotensine (ACEI) dans le traitement de l'hypertension artérielle chez les patients africains est controversée. Objectif: Nous avons examiné la baisse de la tension artérielle ambulatoire (ABP) en réponse à un diurétique et un ACEI chez des patients hypertendus d'origine africaine et nous avons évalué les différentes caractéristiques déterminant l'efficacité du traitement. Méthodes: Etude en simple-aveugle randomisée, en crossover AB/BA. Arrangement Familles hypertendues d'origine africaine de la population générale des Seychelles. Participants : 52 patients (29 hommes et 23 femmes) sur 62 patients hypertendus éligibles ont été inclus. Le principal résultat était la mesure de la réponse de l'ABP à 20 mg de lisinopril (LIS) ou 25 mg d'hydrochlorothiazide (HCT) quotidiennement pendant quatre semaines. Résultats: Le jour, la réponse systolique/diastolique de l'ABP sous HCT était de 4.9 (95% intervalle de confiance (IC) 1.2-8.6)/3.6 (1.0-6.2) mm Hg pour les hommes et 12.9 (9.216.6)/6.3 (3.7-8.8) mm Hg pour les femmes. Sous LIS, la réponse était de 18.8 (15.022.5)/14.6 (12.0-17.1) mm Hg pour les hommes et de 12.4 (8.7-16.2)/7.7 (5.1-10.2) mm Hg pour les femmes. La nuit, la réponse systolique/diastolique sous HCT était de 5.0 (0.6-9.4)/2.7 ((-0.4)-5.7) mm Hg pour les hommes et de 11.5 (7.1-16.0)/5.7 (2.6-8.8) mm Hg pour les femmes, et sous LIS était de 18.7 (14.2-22.1)/15.4 (12.4-18.5) mm Hg pour les hommes et de 3.5 ((-1.0)-7.9)/2.3 ((-0.8)-5.4) mm Hg pour les femmes. L'analyse de régression linéaire multiple a montré que le sexe est un prédicteur indépendant de la réponse tensionnelle à l'HCT et au LIS. Conclusions : Les patients hypertendus d'origine africaine ont présenté une baisse tensionnelle plus grande en réponse au LIS qu'à l'HCT. Les hommes ont mieux répondu au LIS qu'à l'HCT alors que les femmes ont répondu de manière similaire aux deux traitements.

Gender difference in the response to an angiotensin-converting enzyme inhibitor and a diuretic in hypertensive patients of African descent.

BACKGROUND: The efficacy of angiotensin-converting enzyme (ACE) inhibitors in decreasing blood pressure in African patients is controversial. OBJECTIVE: We examined the ambulatory blood pressure (ABP) response to a diuretic and an ACE inhibitor in hypertensive patients of East African descent and evaluated the individual characteristics that determined treatment efficacy. DESIGN: A single-blind randomized AB/BA crossover design. SETTING: Hypertensive families of East African descent from the general population in the Seychelles. PARTICIPANTS: Fifty-two (29 men and 23 women) out of 62 eligible hypertensive patients were included.Main outcome measures ABP response to 20 mg lisinopril (LIS) daily and 25 mg hydrochlorothiazide (HCT) daily given for a 4-week period.Results The daytime systolic/diastolic ABP response to HCT was 4.9 [95% confidence interval (CI) 1.2-8.6]/3.6 (1.0-6.2) mmHg for men and 12.9 (9.2-16.6)/6.3 (3.7-8.8) mmHg for women. With LIS the response was 18.8 (15.0-22.5)/14.6 (12.0-17.1) mmHg for men and 12.4 (8.7-16.2)/7.7 (5.1-10.2) mmHg for women. The night-time systolic/diastolic response to HCT was 5.0 (0.6-9.4)/2.7 [(-0.4)-5.7] mmHg for men and 11.5 (7.1-16.0)/5.7 (2.6-8.8) mmHg for women, and to LIS was 18.7 (14.2-22.1)/15.4 (12.4-18.5) mmHg for men and 3.5 [(-1.0)-7.9]/2.3 [(-0.8)-5.4] mmHg for women. Linear regression analyses showed that gender is an independent predictor of the ABP responses to HCT and to LIS. CONCLUSIONS: Hypertensive patients of African descent responded better to LIS than to HCT. Men responded better to LIS than to HCT and women responded similarly to both drugs.