A Unified Survey of Treatment Effect Heterogeneity Modelling and Uplift Modelling

A central question in many fields of scientific research is to determine how an outcome is affected by an action, i.e., to estimate the causal effect or treatment effect of an action. In recent years, in areas such as personalised healthcare, sociology, and online marketing, a need has emerged to estimate heterogeneous treatment effects with respect to individuals of different characteristics. To meet this need, two major approaches have been taken: treatment effect heterogeneity modelling and uplifting modelling. Researchers and practitioners in different communities have developed algorithms based on these approaches to estimate the heterogeneous treatment effects. In this article, we present a unified view of these two seemingly disconnected yet closely related approaches under the potential outcome framework. We provide a structured survey of existing methods following either of the two approaches, emphasising their inherent connections and using unified notation to facilitate comparisons. We also review the main applications of the surveyed methods in personalised marketing, personalised medicine, and sociology. Finally, we summarise and discuss the available software packages and source codes in terms of their coverage of different methods and applicability to different datasets, and we provide general guidelines for method selection.

Inflammatory Cytokines and Risk of Ischemic Stroke: A Mendelian Randomization Study

Background: Observational studies have revealed the association between some inflammatory cytokines and the occurrence of ischemic stroke, but the causal relationships remain unclear.Methods: We conducted a two-sample Mendelian randomization (MR) analysis to assess the causal effects of thirty inflammatory cytokines and the risk of ischemic stroke. For exposure data, we collected genetic variants associated with inflammatory cytokines as instrumental variables (IVs) from a genome-wide association study (GWAS) meta-analysis from Finland (sample size up to 8,293). For the outcome data, we collected summary data of ischemic stroke from a large-scale GWAS meta-analysis involved 17 studies (34,217 cases and 406,111 controls). We further performed a series of sensitivity analyses as validation of primary MR results.Results: According to the primary MR estimations and further sensitivity analyses, we established one robust association after Bonferroni correction: the odds ratio (95% CI) per unit change in genetically increased IL-4 was 0.84 (0.89–0.95) for ischemic stroke. The chemokine MCP3 showed a nominally significant association with ischemic stroke risk (OR: 0.93, 95% CI: 0.88–0.99, unadjusted p < 0.05). There was no evidence of a causal effect of other inflammatory cytokines and the risk of ischemic stroke.Conclusions: Our study suggested that genetically increased IL-4 levels showed a protective effect on the risk of ischemic stroke, which provides important new insights into the potential therapeutic target for preventing ischemic stroke.

A sub-group evaluation of the multi-month dispensing strategy for differentiated HIV care: is personalization of care guidelines warranted in Haiti?

Background Differentiated care strategies are rapidly becoming the norm for HIV care delivery globally. Building upon an interest in tailoring antiretroviral therapy (ART) delivery for client-centered needs, the Ministry of Health and Population in Haiti formally endorsed multiple-month dispenses (MMD) in the 2016 national ART guidelines This study explores heterogeneity in retention in care with MMD for specific Haitian populations living with HIV and evaluates if a targeted algorithm for optimal ART prescription intervals is warranted in Haiti. Methods This study included ART-naïve individuals who started ART on or after January 1st, 2017 in Haiti. To identify subgroups in which to explore heterogeneity of retention, we implemented a double-lasso regression method to determine which individual characteristics would define the subgroups. Characteristics evaluated for potential subgroup definition included: sex, age category, WHO clinical stage, and body mass index category. We employed instrumental variable models to estimate the causal effect of increasing ART dispensing length on ART retention, by client subgroup. The outcome of interest was retention in care after one year in treatment. We then estimated the marginal effect of a 30-day increase to ART dispensing length to retention in care for each of these subgroups. Results There was evidence for heterogeneity in the effect of extending ART dispensing intervals on retention by WHO clinical stage. We observed significant improvements to retention in care at one year with a 30-day increase in ART dispense length for all subgroups defined by WHO clinical stages 1-4. The effects ranged from a 14.7% increase (95% CI: 12.4-17.0) to the likelihood of retention for people with HIV in WHO stage 1 to a 21.6% increase (95% CI: 18.7-24.5) to the likelihood of retention for those in WHO stage 3. Conclusions All the subgroups defined by WHO clinical stage experienced a benefit of extending ART intervals to retention in care at one year. Though the effect did differ slightly by WHO stage, the effects went in the same direction and were of similar magnitude. Therefore, a standardized recommendation for MMD among those living with HIV and new on ART is appropriate for Haiti treatment guidelines.

Polynomial Mendelian Radomization reveals widespread non-linear causal effects in the UK Biobank

Causal inference is a critical step in improving our understanding of biological processes and Mendelian randomisation (MR) has emerged as one of the foremost methods to efficiently interrogate diverse hypotheses using large-scale, observational data from biobanks. Although many extensions have been developed to address the three core assumptions of MR-based causal inference (relevance, exclusion restriction, and exchangeability), most approaches implicitly assume that any putative causal effect is linear. Here we propose PolyMR, an MR-based method which provides a polynomial approximation of an (arbitrary) causal function between an exposure and an outcome. We show that this method provides accurate inference of the shape and magnitude of causal functions with greater accuracy than existing methods. We applied this method to data from the UK Biobank, testing for effects between anthropometric traits and continuous health-related phenotypes and found most of these (84%) to have causal effects which deviate significantly from linear. These deviations ranged from slight attenuation at the extremes of the exposure distribution, to large changes in the magnitude of the effect across the range of the exposure (e.g. a 1 kg/m2 change in BMI having stronger effects on glucose levels if the initial BMI was higher), to non-monotonic causal relationships (e.g. the effects of BMI on cholesterol forming an inverted U shape). Finally, we show that the linearity assumption of the causal effect may lead to the misinterpretation of health risks at the individual level or heterogeneous effect estimates when using cohorts with differing average exposure levels.

GWAS and ExWAS of blood Mitochondrial DNA copy number identifies 71 loci and highlights a potential causal role in dementia

Background: Mitochondrial DNA copy number (mtDNA-CN) is an accessible blood-based measurement believed to capture underlying mitochondrial function. The specific biological processes underpinning its regulation, and whether those processes are causative for disease, is an area of active investigation.Methods: We developed a novel method for array-based mtDNA-CN estimation suitable for biobank-scale studies, called 'AutoMitoC'. We applied AutoMitoC to 395,781 UKBiobank study participants and performed genome and exome-wide association studies, identifying novel common and rare genetic determinants. Finally, we performed two-sample Mendelian Randomization to assess whether genetically low mtDNA-CN influenced select mitochondrial phenotypes.Results: Overall, genetic analyses identified 71 loci for mtDNA-CN, which implicated several genes involved in rare mtDNA depletion disorders, dNTP metabolism, and the mitochondrial central dogma. Rare variant analysis identified SAMHD1 mutation carriers as having higher mtDNA-CN (beta=0.23 SDs; 95% CI, 0.18- 0.29; P=2.6x10-19), a potential therapeutic target for patients with mtDNA depletion disorders, but at increased risk of breast cancer (OR=1.91; 95% CI, 1.52-2.40; P=2.7x10-8). Finally, Mendelian randomization analyses suggest a causal effect of low mtDNA-CN on dementia risk (OR=1.94 per 1 SD decrease in mtDNA-CN; 95% CI, 1.55-2.32; P=7.5x10-4).Conclusions: Altogether, our genetic findings indicate that mtDNA-CN is a complex biomarker reflecting specific mitochondrial processes related to mtDNA regulation, and that these processes are causally related to human diseases.Funding: No funds supported this specific investigation. Awards and positions supporting authors include: Canadian Institutes of Health Research (CIHR) Frederick Banting and Charles Best Canada Graduate Scholarships Doctoral Award (MC, PM); CIHR Post-Doctoral Fellowship Award (RM); Wellcome Trust Grant number: 099313/B/12/A; Crasnow Travel Scholarship; Bongani Mayosi UCT-PHRI Scholarship 2019/2020 (TM); Wellcome Trust Health Research Board Irish Clinical Academic Training (ICAT) Programme Grant Number: 203930/B/16/Z (CJ); European Research Council COSIP Grant Number: 640580 (MO); E.J. Moran Campbell Internal Career Research Award (MP); CISCO Professorship in Integrated Health Systems and Canada Research Chair in Genetic and Molecular Epidemiology (GP).

Is there a causal relationship between executive function and liability to mental health and substance use? A Mendelian randomisation approach

Background: Executive function consists of several cognitive control processes that are able to regulate lower level processes. Poorer performance in tasks designed to test executive function is associated with a range of psychopathologies such as schizophrenia, major depressive disorder (MDD) and anxiety, as well as with smoking and alcohol consumption. Despite these well-documented associations, whether they reflect causal relationships, and if so in what direction, remains unclear. We aimed to establish whether there is a causal relationship between a latent factor for performance on multiple executive function tasks - which we refer to as common executive function (cEF) - and liability to schizophrenia, MDD, anxiety, smoking initiation, alcohol consumption, alcohol dependence and cannabis use disorder (CUD), and the directionality of any relationship observed. Methods: We used a two-sample bidirectional Mendelian randomisation (MR) approach using genome-wide association study (GWAS) summary data from large cohorts (N=17,310 to 848,460) to examine whether causal relationships exist, and if so in which direction. Results: We found evidence of a causal effect of increased cEF on reduced schizophrenia liability (IVW: OR=0.10; 95% CI 0.05 to 0.19; p-value=3.43x10-12), reduced MDD liability (IVW: OR=0.52; 95% CI 0.38 to 0.72; p-value=5.23x10-05), decreased drinks per week (IVW: β=-0.06; 95% CI -0.10 to -0.02; p-value=0.003), and reduced CUD liability (IVW: OR=0.27; 95% CI 0.12 to 0.61; p-value=1.58x10-03). We also found evidence of a causal effect of increased schizophrenia liability on decreased cEF (IVW: β=-0.04; 95% CI -0.04 to -0.03; p-value=3.25x10-27), as well as smoking initiation on decreased cEF (IVW: β=-0.06; 95%CI -0.09 to -0.03; p-value=6.11x10-05). Conclusion: Our results indicate a potential bidirectional causal relationship between a latent factor measure of executive function (cEF) and schizophrenia liability, a possible causal effect of increased cEF on reduced MDD liability, CUD liability, and alcohol consumption, and a possible causal effect of smoking initiation on decreased cEF. These results suggest that executive function should be considered as a potential risk factor for some mental health and substance use outcomes, and may also be impacted by mental health (particularly schizophrenia). Further studies are required to improve our understanding of the underlying mechanisms of these effects, but our results suggest that executive function may be a promising intervention target. These results may therefore inform the prioritisation of experimental medicine studies (e.g., of executive function interventions), for both mental health and substance use outcomes, to improve the likelihood of successful translation.

Genetic liability to rheumatoid arthritis on autism and autistic traits: polygenic risk score and Mendelian randomization analyses

Higher prevalence of autism in offspring born to mothers with rheumatoid arthritis has been reported in observational studies. We investigated (a) the associations between maternal and offspring’s own genetic liability for rheumatoid arthritis and autism-related outcomes in the offspring using polygenic risk scores (PRS) and (b) whether the effects were causal using Mendelian randomization (MR). Using the latest genome-wide association (GWAS) summary data on rheumatoid arthritis and individual-level data from the Avon Longitudinal Study of Parents and Children, United Kingdom, we constructed PRSs for maternal and offspring genetic liability for rheumatoid arthritis (single-nucleotide polymorphism [SNP] p-value threshold 0.05). We investigated associations with autism, and autistic traits: social and communication difficulties, coherence, repetitive behaviours and sociability. We used modified Poisson regression with robust standard errors. In two-sample MR analyses, we used 40 genome-wide significant SNPs for rheumatoid arthritis and investigated the causal effects on risk for autism, in 18,381 cases and 27,969 controls of the Psychiatric Genetics Consortium and iPSYCH. Sample size ranged from 4992 to 7849 in PRS analyses. We found little evidence of associations between rheumatoid arthritis PRSs and autism-related phenotypes in the offspring (maternal PRS on autism: RR 0.89, 95%CI 0.73–1.07, p = 0.21; offspring’s own PRS on autism: RR 1.11, 95%CI 0.88–1.39, p = 0.39). MR results provided little evidence for a causal effect (IVW OR 1.01, 95%CI 0.98–1.04, p = 0.56). There was little evidence for associations between genetic liability for rheumatoid arthritis on autism-related outcomes in offspring. Lifetime risk for rheumatoid arthritis has no causal effects on autism.