Selection Bias When Estimating Average Treatment Effects Using One-sample Instrumental Variable Analysis

The impact of excess body weight on employment outcomes: A systematic review of the evidence

BACKGROUND

Excess body weight has been recognised as an important factor in influencing labour market outcomes. Several hypotheses explain the causal effect of excess body weight on employment outcomes, including productivity, labour supply, and discrimination. In this review, we provide a systematic synthesis of the evidence on the causal impact of excess body weight on labour market outcomes worldwide.

METHODS

We searched Econ Lit, and Web of Science databases for relevant studies published from 1st Jan 2010-20 th Jan 2023. Studies were included if they were either longitudinal analysis, pooled cross-sectional or cross-sectional studies if they used instrumental variable methodology based on Mendelian Randomisation. Only studies with measures of body weight and employment outcomes were included.

RESULTS

The number of potentially relevant studies constituted 4321 hits. A total of 59 studies met the inclusion criteria and were qualitatively reviewed by the authors. Most of the included studies were conducted in the USA (N=18), followed by the UK (N=9), Germany (N=6), Finland (N=4), and non-EU countries (N=22). Evidence from the included studies suggests that the effect of excess weight differs by gender, ethnicity, country, and time period. White women with excess weight in the USA, the UK, Germany, Canada, and in the EU (multi-country analyses) are less likely to be employed, and when employed they face lower wages compared to normal weight counterparts. For men there is no effect of excess weight on employment outcomes or the magnitude of the effect is much smaller or even positive in some cases.

CONCLUSIONS

This review has shown that despite ample research on the relationship between excess weight and employment status and wages, robust causal evidence of the effects of excess weight on employment outcomes remains scarce and relies significantly on strong statistical and theoretical assumptions. Further research into these relationships outside of USA and Western Europe context is needed.

Using Genetics to Investigate Relationships between Phenotypes: Application to Endometrial Cancer

Genome-wide association studies (GWAS) have accelerated the exploration of genotype-phenotype associations, facilitating the discovery of replicable genetic markers associated with specific traits or complex diseases. This narrative review explores the statistical methodologies developed using GWAS data to investigate relationships between various phenotypes, focusing on endometrial cancer, the most prevalent gynecological malignancy in developed nations. Advancements in analytical techniques such as genetic correlation, colocalization, cross-trait locus identification, and causal inference analyses have enabled deeper exploration of associations between different phenotypes, enhancing statistical power to uncover novel genetic risk regions. These analyses have unveiled shared genetic associations between endometrial cancer and many phenotypes, enabling identification of novel endometrial cancer risk loci and furthering our understanding of risk factors and biological processes underlying this disease. The current status of research in endometrial cancer is robust; however, this review demonstrates that further opportunities exist in statistical genetics that hold promise for advancing the understanding of endometrial cancer and other complex diseases.

Impaired GK-GKRP interaction rather than direct GK activation worsens lipid profiles and contributes to long-term complications: a Mendelian randomization study

BACKGROUND

Glucokinase (GK) plays a key role in glucose metabolism. In the liver, GK is regulated by GK regulatory protein (GKRP) with nuclear sequestration at low plasma glucose level. Some GK activators (GKAs) disrupt GK-GKRP interaction which increases hepatic cytoplasmic GK level. Excess hepatic GK activity may exceed the capacity of glycogen synthesis with excess triglyceride formation. It remains uncertain whether hypertriglyceridemia associated with some GKAs in previous clinical trials was due to direct GK activation or impaired GK-GKRP interaction.

METHODS

Using publicly available genome-wide association study summary statistics, we selected independent genetic variants of GCKR and GCK associated with fasting plasma glucose (FPG) as instrumental variables, to mimic the effects of impaired GK-GKRP interaction and direct GK activation, respectively. We applied two-sample Mendelian Randomization (MR) framework to assess their causal associations with lipid-related traits, risks of metabolic dysfunction-associated steatotic liver disease (MASLD) and cardiovascular diseases. We verified these findings in one-sample MR analysis using individual-level statistics from the Hong Kong Diabetes Register (HKDR).

RESULTS

Genetically-proxied impaired GK-GKRP interaction increased plasma triglycerides, low-density lipoprotein cholesterol and apolipoprotein B levels with increased odds ratio (OR) of 14.6 (95% CI 4.57-46.4) per 1 mmol/L lower FPG for MASLD and OR of 2.92 (95% CI 1.78-4.81) for coronary artery disease (CAD). Genetically-proxied GK activation was associated with decreased risk of CAD (OR 0.69, 95% CI 0.54-0.88) and not with dyslipidemia. One-sample MR validation in HKDR showed consistent results.

CONCLUSIONS

Impaired GK-GKRP interaction, rather than direct GK activation, may worsen lipid profiles and increase risks of MASLD and CAD. Development of future GKAs should avoid interfering with GK-GKRP interaction.

Methodological approaches, challenges, and opportunities in the application of Mendelian randomisation to lifecourse epidemiology: A systematic literature review

Diseases diagnosed in adulthood may have antecedents throughout (including prenatal) life. Gaining a better understanding of how exposures at different stages in the lifecourse influence health outcomes is key to elucidating the potential benefits of disease prevention strategies. Mendelian randomisation (MR) is increasingly used to estimate causal effects of exposures across the lifecourse on later life outcomes. This systematic literature review explores MR methods used to perform lifecourse investigations and reviews previous work that has utilised MR to elucidate the effects of factors acting at different stages of the lifecourse. We conducted searches in PubMed, Embase, Medline and MedRXiv databases. Thirteen methodological studies were identified. Four studies focused on the impact of time-varying exposures in the interpretation of "standard" MR techniques, five presented methods for repeat measures of the same exposure, and four described methodological approaches to handling multigenerational exposures. A further 127 studies presented the results of an applied research question. Over half of these estimated effects in a single generation and were largely confined to the exploration of questions regarding body composition. The remaining mostly estimated maternal effects. There is a growing body of research focused on the development and application of MR methods to address lifecourse research questions. The underlying assumptions require careful consideration and the interpretation of results rely on select conditions. Whilst we do not advocate for a particular strategy, we encourage practitioners to make informed decisions on how to approach a research question in this field with a solid understanding of the limitations present and how these may be affected by the research question, modelling approach, instrument selection, and data availability.

Causal Estimation of Long-term Intervention Cost-effectiveness Using Genetic Instrumental Variables: An Application to Cancer

BACKGROUND

This article demonstrates a means of assessing long-term intervention cost-effectiveness in the absence of data from randomized controlled trials and without recourse to Markov simulation or similar types of cohort simulation.

METHODS

Using a Mendelian randomization study design, we developed causal estimates of the genetically predicted effect of bladder, breast, colorectal, lung, multiple myeloma, ovarian, prostate, and thyroid cancers on health care costs and quality-adjusted life-years (QALYs) using outcome data drawn from the UK Biobank cohort. We then used these estimates in a simulation model to estimate the cost-effectiveness of a hypothetical population-wide preventative intervention based on a repurposed class of antidiabetic drugs known as sodium-glucose cotransporter-2 (SGLT2) inhibitors very recently shown to reduce the odds of incident prostate cancer.

RESULTS

Genetic liability to prostate cancer and breast cancer had material causal impacts on either or both health care costs and QALYs. Mendelian randomization results for the less common cancers were associated with considerable uncertainty. SGLT2 inhibition was unlikely to be a cost-effective preventative intervention for prostate cancer, although this conclusion depended on the price at which these drugs would be offered for a novel anticancer indication.

IMPLICATIONS

Our new causal estimates of cancer exposures on health economic outcomes may be used as inputs into decision-analytic models of cancer interventions such as screening programs or simulations of longer-term outcomes associated with therapies investigated in randomized controlled trials with short follow-ups. Our method allowed us to rapidly and efficiently estimate the cost-effectiveness of a hypothetical population-scale anticancer intervention to inform and complement other means of assessing long-term intervention value.

HIGHLIGHTS

The article demonstrates a novel method of assessing long-term intervention cost-effectiveness without relying on randomized controlled trials or cohort simulations.Mendelian randomization was used to estimate the causal effects of certain cancers on health care costs and quality-adjusted life-years (QALYs) using data from the UK Biobank cohort.Given causal data on the association of different cancer exposures on costs and QALYs, it was possible to simulate the cost-effectiveness of an anticancer intervention.Genetic liability to prostate cancer and breast cancer significantly affected health care costs and QALYs, but the hypothetical intervention using SGLT2 inhibitors for prostate cancer may not be cost-effective, depending on the drug's price for the new anticancer indication. The methods we propose and implement can be used to efficiently estimate intervention cost-effectiveness and to inform decision making in all manner of preventative and therapeutic contexts.

Causal Relationships Between Screen Use, Reading, and Brain Development in Early Adolescents

The rise of new media has greatly changed the lifestyles, leading to increased time on these platforms and less time spent reading. This shift has particularly profound impacts on early adolescents, who are in a critical stage of brain development. Previous studies have found associations between screen use and mental health, but it remains unclear whether screen use is the direct cause of the outcomes. Here, the Adolescent Brain Cognitive Development (ABCD) dataset is utlized to examine the causal relationships between screen use and brain development. The results revealed adverse causal effects of screen use on language ability and specific behaviors in early adolescents, while reading has positive causal effects on their language ability and brain volume in the frontal and temporal regions. Interestingly, increased screen use is identified as a result, rather than a cause, of certain behaviors such as rule-breaking and aggressive behaviors. Furthermore, the analysis uncovered an indirect influence of screen use, mediated by changes in reading habits, on brain development. These findings provide new evidence for the causal influences of screen use on brain development and highlight the importance of monitoring media use and related habit change in children.

The causal effects of education on adult health, mortality and income: evidence from Mendelian randomization and the raising of the school leaving age

BACKGROUND

On average, educated people are healthier, wealthier and have higher life expectancy than those with less education. Numerous studies have attempted to determine whether education causes differences in later health outcomes or whether another factor ultimately causes differences in education and subsequent outcomes. Previous studies have used a range of natural experiments to provide causal evidence. Here we compare two natural experiments: a policy reform, raising the school leaving age in the UK in 1972; and Mendelian randomization.

METHODS

We used data from 334 974 participants of the UK Biobank, sampled between 2006 and 2010. We estimated the effect of an additional year of education on 25 outcomes, including mortality, measures of morbidity and health, ageing and income, using multivariable adjustment, the policy reform and Mendelian randomization. We used a range of sensitivity analyses and specification tests to assess the plausibility of each method's assumptions.

RESULTS

The three different estimates of the effects of educational attainment were largely consistent in direction for diabetes, stroke and heart attack, mortality, smoking, income, grip strength, height, body mass index (BMI), intelligence, alcohol consumption and sedentary behaviour. However, there was evidence that education reduced rates of moderate exercise and increased alcohol consumption. Our sensitivity analyses suggest that confounding by genotypic or phenotypic confounders or specific forms of pleiotropy are unlikely to explain our results.

CONCLUSIONS

Previous studies have suggested that the differences in outcomes associated with education may be due to confounding. However, the two independent sources of exogenous variation we exploit largely imply consistent causal effects of education on outcomes later in life.

Chronic inflammation does not mediate the effect of adiposity on grip strength: results from a multivariable Mendelian randomization study

The relationship between adiposity and grip strength (GS) is complex. We investigated whether one pathway through which adiposity affects GS was via chronic inflammation. 367,583 UK Biobank participants had body mass index (BMI), waist-hip-ratio (WHR), C-reactive protein (CRP) and GS data. Univariable Mendelian randomization (MR) and multivariable Mendelian randomization (MVMR) analyses (using inverse variance weighted (IVW) weighted median estimates (WME) and MR-Egger models) estimated total, direct and indirect effects of adiposity traits on GS using genetic instruments for BMI and WHR (exposures) and CRP (mediator). Observational findings suggested higher BMI was associated with stronger grip, e.g., in males, per standard deviation (SD) higher BMI, GS was higher by 0.48 kg (95% confidence interval(CI):0.44,0.51), independent of CRP. For males MR estimates were directionally consistent; for females, estimates were consistent with the null. Observational findings for WHR suggested that higher WHR was associated with weaker grip. In multivariable MR-IVW analyses, effects in males were consistent with the null. In females, there were consistent effects such that higher WHR was associated with stronger grip, e.g., 1-SD higher WHR was associated with 1.25 kg (MVMR-Egger; 95% CI:0.72,1.78) stronger grip, independent of CRP. Across sexes and adiposity indicators, CRP's mediating role was minor. Greater adiposity may increase GS in early old age, but effects vary by sex and adiposity location. There was no evidence that inflammation mediated these effects.

Guidelines for performing Mendelian randomization investigations: update for summer 2023

This paper provides guidelines for performing Mendelian randomization investigations. It is aimed at practitioners seeking to undertake analyses and write up their findings, and at journal editors and reviewers seeking to assess Mendelian randomization manuscripts. The guidelines are divided into ten sections: motivation and scope, data sources, choice of genetic variants, variant harmonization, primary analysis, supplementary and sensitivity analyses (one section on robust statistical methods and one on other approaches), extensions and additional analyses, data presentation, and interpretation. These guidelines will be updated based on feedback from the community and advances in the field. Updates will be made periodically as needed, and at least every 24 months.

Linear and Nonlinear Associations Between Vitamin D and Grip Strength: A Mendelian Randomization Study in UK Biobank

BACKGROUND

Low vitamin D status is a widespread phenomenon. Similarly, muscle weakness, often indicated by low grip strength, is another public health concern; however, the vitamin D-grip strength relationship is equivocal. It is important to understand whether variation in vitamin D status causally influences muscle strength to elucidate whether supplementation may help prevent/treat muscle weakness.

METHODS

UK Biobank participants, aged 37-73 years, with valid data on Vitamin D status (circulating 25-hydroxyvitamin D [25(OH)D] concentration) and maximum grip strength were included (N = 368,890). We examined sex-specific cross-sectional associations between 25(OH)D and grip strength. Using Mendelian randomization (MR), we estimated the strength of the 25(OH)D-grip strength associations using genetic instruments for 25(OH)D as our exposure. Crucially, because potential effects of vitamin D supplementation on strength could vary by underlying 25(OH)D status, we allowed for nonlinear relationships between 25(OH)D and strength in all analyses.

RESULTS

Mean (SD) of 25(OH)D was 50 (21) nmol/L in males and females. In cross-sectional analyses, there was evidence of nonlinear associations between 25(OH)D and strength, for example, compared to males with 50 nmol/L circulating 25(OH)D, males with 75 nmol/L had 0.36 kg (0.31,0.40) stronger grip; males with 25 nmol/L had 1.01 kg (95% confidence interval [CI]: 0.93, 1.08) weaker grip. In MR analyses, linear and nonlinear models fitted the data similarly well, for example, 25 nmol/L higher circulating 25(OH)D in males was associated with 0.25 kg (-0.05, 0.55) greater grip (regardless of initial 25(OH)D status). Results were similar, albeit weaker, for females.

CONCLUSIONS

Using two different methods to triangulate evidence, our findings suggest moderate to small causal links between circulating 25(OH)D and grip strength.

Age-specific effects of weight-based body size on fracture risk in later life: a lifecourse Mendelian randomisation study

Musculoskeletal conditions, including fractures, can have severe and long-lasting consequences. Higher body mass index in adulthood is widely acknowledged to be protective for most fracture sites. However, sources of bias induced by confounding factors may have distorted previous findings. Employing a lifecourse Mendelian randomisation (MR) approach by using genetic instruments to separate effects at different life stages, this investigation aims to explore how prepubertal and adult body size independently influence fracture risk in later life.Using data from a large prospective cohort, univariable and multivariable MR were conducted to simultaneously estimate the effects of age-specific genetic proxies for body size (n = 453,169) on fracture risk (n = 416,795). A two-step MR framework was additionally applied to elucidate potential mediators. Univariable and multivariable MR indicated strong evidence that higher body size in childhood reduced fracture risk (OR, 95% CI: 0.89, 0.82 to 0.96, P = 0.005 and 0.76, 0.69 to 0.85, P = 1 × 10- 6, respectively). Conversely, higher body size in adulthood increased fracture risk (OR, 95% CI: 1.08, 1.01 to 1.16, P = 0.023 and 1.26, 1.14 to 1.38, P = 2 × 10- 6, respectively). Two-step MR analyses suggested that the effect of higher body size in childhood on reduced fracture risk was mediated by its influence on higher estimated bone mineral density (eBMD) in adulthood.This investigation provides novel evidence that higher body size in childhood reduces fracture risk in later life through its influence on increased eBMD. From a public health perspective, this relationship is complex since obesity in adulthood remains a major risk factor for co-morbidities. Results additionally indicate that higher body size in adulthood is a risk factor for fractures. Protective effect estimates previously observed are likely attributed to childhood effects.

The impact of fatty acids biosynthesis on the risk of cardiovascular diseases in Europeans and East Asians: a Mendelian randomization study

Despite early interest, the evidence linking fatty acids to cardiovascular diseases (CVDs) remains controversial. We used Mendelian randomization to explore the involvement of polyunsaturated (PUFA) and monounsaturated (MUFA) fatty acids biosynthesis in the etiology of several CVD endpoints in up to 1 153 768 European (maximum 123 668 cases) and 212 453 East Asian (maximum 29 319 cases) ancestry individuals. As instruments, we selected single nucleotide polymorphisms mapping to genes with well-known roles in PUFA (i.e. FADS1/2 and ELOVL2) and MUFA (i.e. SCD) biosynthesis. Our findings suggest that higher PUFA biosynthesis rate (proxied by rs174576 near FADS1/2) is related to higher odds of multiple CVDs, particularly ischemic stroke, peripheral artery disease and venous thromboembolism, whereas higher MUFA biosynthesis rate (proxied by rs603424 near SCD) is related to lower odds of coronary artery disease among Europeans. Results were unclear for East Asians as most effect estimates were imprecise. By triangulating multiple approaches (i.e. uni-/multi-variable Mendelian randomization, a phenome-wide scan, genetic colocalization and within-sibling analyses), our results are compatible with higher low-density lipoprotein (LDL) cholesterol (and possibly glucose) being a downstream effect of higher PUFA biosynthesis rate. Our findings indicate that PUFA and MUFA biosynthesis are involved in the etiology of CVDs and suggest LDL cholesterol as a potential mediating trait between PUFA biosynthesis and CVDs risk.

Investigating causal relations between sleep duration and risks of adverse pregnancy and perinatal outcomes: linear and nonlinear Mendelian randomization analyses

BACKGROUND

Observational studies have reported maternal short/long sleep duration to be associated with adverse pregnancy and perinatal outcomes. However, it remains unclear whether there are nonlinear causal effects. Our aim was to use Mendelian randomization (MR) and multivariable regression to examine nonlinear effects of sleep duration on stillbirth (MR only), miscarriage (MR only), gestational diabetes, hypertensive disorders of pregnancy, perinatal depression, preterm birth and low/high offspring birthweight.

METHODS

We used data from European women in UK Biobank (N=176,897), FinnGen (N=~123,579), Avon Longitudinal Study of Parents and Children (N=6826), Born in Bradford (N=2940) and Norwegian Mother, Father and Child Cohort Study (MoBa, N=14,584). We used 78 previously identified genetic variants as instruments for sleep duration and investigated its effects using two-sample, and one-sample nonlinear (UK Biobank only), MR. We compared MR findings with multivariable regression in MoBa (N=76,669), where maternal sleep duration was measured at 30 weeks.

RESULTS

In UK Biobank, MR provided evidence of nonlinear effects of sleep duration on stillbirth, perinatal depression and low offspring birthweight. Shorter and longer duration increased stillbirth and low offspring birthweight; shorter duration increased perinatal depression. For example, longer sleep duration was related to lower risk of low offspring birthweight (odds ratio 0.79 per 1 h/day (95% confidence interval: 0.67, 0.93)) in the shortest duration group and higher risk (odds ratio 1.40 (95% confidence interval: 1.06, 1.84)) in the longest duration group, suggesting shorter and longer duration increased the risk. These were supported by the lack of evidence of a linear effect of sleep duration on any outcome using two-sample MR. In multivariable regression, risks of all outcomes were higher in the women reporting <5 and ≥10 h/day sleep compared with the reference category of 8-9 h/day, despite some wide confidence intervals. Nonlinear models fitted the data better than linear models for most outcomes (likelihood ratio P-value=0.02 to 3.2×10-52), except for gestational diabetes.

CONCLUSIONS

Our results show shorter and longer sleep duration potentially causing higher risks of stillbirth, perinatal depression and low offspring birthweight. Larger studies with more cases are needed to detect potential nonlinear effects on hypertensive disorders of pregnancy, preterm birth and high offspring birthweight.

Associations between insomnia and pregnancy and perinatal outcomes: Evidence from mendelian randomization and multivariable regression analyses

BACKGROUND

Insomnia is common and associated with adverse pregnancy and perinatal outcomes in observational studies. However, those associations could be vulnerable to residual confounding or reverse causality. Our aim was to estimate the association of insomnia with stillbirth, miscarriage, gestational diabetes (GD), hypertensive disorders of pregnancy (HDP), perinatal depression, preterm birth (PTB), and low/high offspring birthweight (LBW/HBW).

METHODS AND FINDINGS

We used 2-sample mendelian randomization (MR) with 81 single-nucleotide polymorphisms (SNPs) instrumenting for a lifelong predisposition to insomnia. Our outcomes included ever experiencing stillbirth, ever experiencing miscarriage, GD, HDP, perinatal depression, PTB (gestational age <37 completed weeks), LBW (<2,500 grams), and HBW (>4,500 grams). We used data from women of European descent (N = 356,069, mean ages at delivery 25.5 to 30.0 years) from UK Biobank (UKB), FinnGen, Avon Longitudinal Study of Parents and Children (ALSPAC), Born in Bradford (BiB), and the Norwegian Mother, Father and Child Cohort (MoBa). Main MR analyses used inverse variance weighting (IVW), with weighted median and MR-Egger as sensitivity analyses. We compared MR estimates with multivariable regression of insomnia in pregnancy on outcomes in ALSPAC (N = 11,745). IVW showed evidence of an association of genetic susceptibility to insomnia with miscarriage (odds ratio (OR): 1.60, 95% confidence interval (CI): 1.18, 2.17, p = 0.002), perinatal depression (OR 3.56, 95% CI: 1.49, 8.54, p = 0.004), and LBW (OR 3.17, 95% CI: 1.69, 5.96, p < 0.001). IVW results did not support associations of insomnia with stillbirth, GD, HDP, PTB, and HBW, with wide CIs including the null. Associations of genetic susceptibility to insomnia with miscarriage, perinatal depression, and LBW were not observed in weighted median or MR-Egger analyses. Results from these sensitivity analyses were directionally consistent with IVW results for all outcomes, with the exception of GD, perinatal depression, and PTB in MR-Egger. Multivariable regression showed associations of insomnia at 18 weeks of gestation with perinatal depression (OR 2.96, 95% CI: 2.42, 3.63, p < 0.001), but not with LBW (OR 0.92, 95% CI: 0.69, 1.24, p = 0.60). Multivariable regression with miscarriage and stillbirth was not possible due to small numbers in index pregnancies. Key limitations are potential horizontal pleiotropy (particularly for perinatal depression) and low statistical power in MR, and residual confounding in multivariable regression.

CONCLUSIONS

In this study, we observed some evidence in support of a possible causal relationship between genetically predicted insomnia and miscarriage, perinatal depression, and LBW. Our study also found observational evidence in support of an association between insomnia in pregnancy and perinatal depression, with no clear multivariable evidence of an association with LBW. Our findings highlight the importance of healthy sleep in women of reproductive age, though replication in larger studies, including with genetic instruments specific to insomnia in pregnancy are important.

The causes and consequences of Alzheimer's disease: phenome-wide evidence from Mendelian randomization

Alzheimer's disease (AD) has no proven causal and modifiable risk factors, or effective interventions. We report a phenome-wide association study (PheWAS) of genetic liability for AD in 334,968 participants of the UK Biobank study, stratified by age. We also examined the effects of AD genetic liability on previously implicated risk factors. We replicated these analyses in the HUNT study. PheWAS hits and previously implicated risk factors were followed up in a Mendelian randomization (MR) framework to identify the causal effect of each risk factor on AD risk. A higher genetic liability for AD was associated with medical history and cognitive, lifestyle, physical and blood-based measures as early as 39 years of age. These effects were largely driven by the APOE gene. The follow-up MR analyses were primarily null, implying that most of these associations are likely to be a consequence of prodromal disease or selection bias, rather than the risk factor causing the disease.

Estimating the causal effect of liability to disease on healthcare costs using Mendelian Randomization

Accurate measurement of the effects of disease status on healthcare costs is important in the pragmatic evaluation of interventions but is complicated by endogeneity bias. Mendelian Randomization, the use of random perturbations in germline genetic variation as instrumental variables, can avoid these limitations. We used a novel Mendelian Randomization analysis to model the causal impact on inpatient hospital costs of liability to six prevalent diseases and health conditions: asthma, eczema, migraine, coronary heart disease, Type 2 diabetes, and depression. We identified genetic variants from replicated genome-wide associations studies and estimated their association with inpatient hospital costs on over 300,000 individuals. There was concordance of findings across varieties of sensitivity analyses, including stratification by sex and methods robust to violations of the exclusion restriction. Results overall were imprecise and we could not rule out large effects of liability to disease on healthcare costs. In particular, genetic liability to coronary heart disease had substantial impacts on costs.

Sample-constrained partial identification with application to selection bias

Many partial identification problems can be characterized by the optimal value of a function over a set where both the function and set need to be estimated by empirical data. Despite some progress for convex problems, statistical inference in this general setting remains to be developed. To address this, we derive an asymptotically valid confidence interval for the optimal value through an appropriate relaxation of the estimated set. We then apply this general result to the problem of selection bias in population-based cohort studies. We show that existing sensitivity analyses, which are often conservative and difficult to implement, can be formulated in our framework and made significantly more informative via auxiliary information on the population. We conduct a simulation study to evaluate the finite sample performance of our inference procedure and conclude with a substantive motivating example on the causal effect of education on income in the highly-selected UK Biobank cohort. We demonstrate that our method can produce informative bounds using plausible population-level auxiliary constraints. We implement this method in the R package selectioninterval.

Avoiding collider bias in Mendelian randomization when performing stratified analyses

Mendelian randomization (MR) uses genetic variants as instrumental variables to investigate the causal effect of a risk factor on an outcome. A collider is a variable influenced by two or more other variables. Naive calculation of MR estimates in strata of the population defined by a collider, such as a variable affected by the risk factor, can result in collider bias. We propose an approach that allows MR estimation in strata of the population while avoiding collider bias. This approach constructs a new variable, the residual collider, as the residual from regression of the collider on the genetic instrument, and then calculates causal estimates in strata defined by quantiles of the residual collider. Estimates stratified on the residual collider will typically have an equivalent interpretation to estimates stratified on the collider, but they are not subject to collider bias. We apply the approach in several simulation scenarios considering different characteristics of the collider variable and strengths of the instrument. We then apply the proposed approach to investigate the causal effect of smoking on bladder cancer in strata of the population defined by bodyweight. The new approach generated unbiased estimates in all the simulation settings. In the applied example, we observed a trend in the stratum-specific MR estimates at different bodyweight levels that suggested stronger effects of smoking on bladder cancer among individuals with lower bodyweight. The proposed approach can be used to perform MR studying heterogeneity among subgroups of the population while avoiding collider bias.

Exploring and mitigating potential bias when genetic instrumental variables are associated with multiple non-exposure traits in Mendelian randomization

With the increasing size and number of genome-wide association studies, individual single nucleotide polymorphisms are increasingly found to associate with multiple traits. Many different mechanisms could result in proposed genetic IVs for an exposure of interest being associated with multiple non-exposure traits, some of which could bias MR results. We describe and illustrate, through causal diagrams, a range of scenarios that could result in proposed IVs being related to non-exposure traits in MR studies. These associations could occur due to five scenarios: (i) confounding, (ii) vertical pleiotropy, (iii) horizontal pleiotropy, (iv) reverse causation and (v) selection bias. For each of these scenarios we outline steps that could be taken to explore the underlying mechanism and mitigate any resulting bias in the MR estimation. We recommend MR studies explore possible IV-non-exposure associations across a wider range of traits than is usually the case. We highlight the pros and cons of relying on sensitivity analyses without considering particular pleiotropic paths versus systematically exploring and controlling for potential pleiotropic or other biasing paths via known traits. We apply our recommendations to an illustrative example of the effect of maternal insomnia on offspring birthweight in UK Biobank.

Adiposity and grip strength: a Mendelian randomisation study in UK Biobank

BACKGROUND

Muscle weakness, which increases in prevalence with age, is a major public health concern. Grip strength is commonly used to identify weakness and an improved understanding of its determinants is required. We aimed to investigate if total and central adiposity are causally associated with grip strength.

METHODS

Up to 470,786 UK Biobank participants, aged 38-73 years, with baseline data on four adiposity indicators (body mass index (BMI), body fat percentage (BF%), waist circumference (WC) and waist-hip-ratio (WHR)) and maximum grip strength were included. We examined sex-specific associations between each adiposity indicator and grip strength. We explored whether associations varied by age, by examining age-stratified associations (< 50 years, 50-59 years, 60-64 years,65 years +). Using Mendelian randomisation (MR), we estimated the strength of the adiposity-grip strength associations using genetic instruments for each adiposity trait as our exposure.

RESULTS

In males, observed and MR associations were generally consistent: higher BMI and WC were associated with stronger grip; higher BF% and WHR were associated with weaker grip: 1-SD higher BMI was associated with 0.49 kg (95% CI: 0.45 kg, 0.53 kg) stronger grip; 1-SD higher WHR was associated with 0.45 kg (95% CI:0.41 kg, 0.48 kg) weaker grip (covariate adjusted observational analyses). Associations of BMI and WC with grip strength were weaker at older ages: in males aged < 50 years and 65 years + , 1-SD higher BMI was associated with 0.93 kg (95% CI: 0.84 kg, 1.01 kg) and 0.13 kg (95% CI: 0.05 kg, 0.21 kg) stronger grip, respectively. In females, higher BF% was associated with weaker grip and higher WC was associated with stronger grip; other associations were inconsistent.

CONCLUSIONS

Using different methods to triangulate evidence, our findings suggest causal links between adiposity and grip strength. Specifically, higher BF% (in both sexes) and WHR (males only) were associated with weaker grip strength.

Understanding the assumptions underlying Mendelian randomization

With the rapidly increasing availability of large genetic data sets in recent years, Mendelian Randomization (MR) has quickly gained popularity as a novel secondary analysis method. Leveraging genetic variants as instrumental variables, MR can be used to estimate the causal effects of one phenotype on another even when experimental research is not feasible, and therefore has the potential to be highly informative. It is dependent on strong assumptions however, often producing biased results if these are not met. It is therefore imperative that these assumptions are well-understood by researchers aiming to use MR, in order to evaluate their validity in the context of their analyses and data. The aim of this perspective is therefore to further elucidate these assumptions and the role they play in MR, as well as how different kinds of data can be used to further support them.

Impact of nonrandom selection mechanisms on the causal effect estimation for two-sample Mendelian randomization methods

Nonrandom selection in one-sample Mendelian Randomization (MR) results in biased estimates and inflated type I error rates only when the selection effects are sufficiently large. In two-sample MR, the different selection mechanisms in two samples may more seriously affect the causal effect estimation. Firstly, we propose sufficient conditions for causal effect invariance under different selection mechanisms using two-sample MR methods. In the simulation study, we consider 49 possible selection mechanisms in two-sample MR, which depend on genetic variants (G), exposures (X), outcomes (Y) and their combination. We further compare eight pleiotropy-robust methods under different selection mechanisms. Results of simulation reveal that nonrandom selection in sample II has a larger influence on biases and type I error rates than those in sample I. Furthermore, selections depending on X+Y, G+Y, or G+X+Y in sample II lead to larger biases than other selection mechanisms. Notably, when selection depends on Y, bias of causal estimation for non-zero causal effect is larger than that for null causal effect. Especially, the mode based estimate has the largest standard errors among the eight methods. In the absence of pleiotropy, selections depending on Y or G in sample II show nearly unbiased causal effect estimations when the casual effect is null. In the scenarios of balanced pleiotropy, all eight MR methods, especially MR-Egger, demonstrate large biases because the nonrandom selections result in the violation of the Instrument Strength Independent of Direct Effect (InSIDE) assumption. When directional pleiotropy exists, nonrandom selections have a severe impact on the eight MR methods. Application demonstrates that the nonrandom selection in sample II (coronary heart disease patients) can magnify the causal effect estimation of obesity on HbA1c levels. In conclusion, nonrandom selection in two-sample MR exacerbates the bias of causal effect estimation for pleiotropy-robust MR methods.

It is well known that nonrandom selection in one-sample Mendelian Randomization (MR) can result in biased estimates and inflated type I error rates. Actually, two-sample MR analyses are more prone to be affected by nonrandom selection than one-sample MR analyses, because two samples for genome-wide association studies (GWAS) may be selected each under different mechanisms from the source population. Summary-level genetic association statistics in two-sample MR may be derived from different study designs such as case-control, case-only and cohort studies, which further inevitably affect the causal effect estimation of exposure on outcome. In this study, we firstly propose a theorem for causal effect invariance under different selection mechanisms. In the simulation, we design 49 combinations of nonrandom selection mechanisms in sample I and sample II, which are widespread in practical applications. The simulation results reveal that the selection mechanisms in sample II have a larger influence on biases and type I error rates than those in sample I. As an illustrative example, we find the nonrandom selection in sample II (coronary heart disease patients) can magnify the causal effect estimation of obesity on the HbA1c levels.

Mendelian randomization

Mendelian randomization (MR) is a term that applies to the use of genetic variation to address causal questions about how modifiable exposures influence different outcomes. The principles of MR are based on Mendel's laws of inheritance and instrumental variable estimation methods, which enable the inference of causal effects in the presence of unobserved confounding. In this Primer, we outline the principles of MR, the instrumental variable conditions underlying MR estimation and some of the methods used for estimation. We go on to discuss how the assumptions underlying an MR study can be assessed and give methods of estimation that are robust to certain violations of these assumptions. We give examples of a range of studies in which MR has been applied, the limitations of current methods of analysis and the outlook for MR in the future. The difference between the assumptions required for MR analysis and other forms of non-interventional epidemiological studies means that MR can be used as part of a triangulation across multiple sources of evidence for causal inference.

Mendelian Randomization: Concepts and Scope

Mendelian randomization (MR) is a method of studying the causal effects of modifiable exposures (i.e., potential risk factors) on health, social, and economic outcomes using genetic variants associated with the specific exposures of interest. MR provides a more robust understanding of the influence of these exposures on outcomes because germline genetic variants are randomly inherited from parents to offspring and, as a result, should not be related to potential confounding factors that influence exposure-outcome associations. The genetic variant can therefore be used as a tool to link the proposed risk factor and outcome, and to estimate this effect with less confounding and bias than conventional epidemiological approaches. We describe the scope of MR, highlighting the range of applications being made possible as genetic data sets and resources become larger and more freely available. We outline the MR approach in detail, covering concepts, assumptions, and estimation methods. We cover some common misconceptions, provide strategies for overcoming violation of assumptions, and discuss future prospects for extending the clinical applicability, methodological innovations, robustness, and generalizability of MR findings.

Investigating the Association Between Seven Sleep Traits and Nonalcoholic Fatty Liver Disease: Observational and Mendelian Randomization Study

Background and Aim: Aberrant sleep parameters are associated with the risk of nonalcoholic fatty liver disease (NAFLD). However, existing information is inconsistent among studies and involves reverse causation. Therefore, we aimed to investigate the observational associations and causations between sleep traits and NAFLD. Methods: We performed multivariable regression to assess observational associations of seven sleep traits (sleep duration, easiness of getting up in the morning, chronotype, nap during day, snoring, insomnia, and narcolepsy), and NAFLD in the UK Biobank (1,029 NAFLD). The Cox proportional hazards model was applied to derive hazard ratios and 95% confidence intervals (CIs). Furthermore, a bidirectional two-sample Mendelian randomization (MR) approach was used to explore the causal relationships between sleep traits and NAFLD. Results: In the multivariable regression model adjusted for potential confounders, getting up in the morning not at all easy (HR, 1.51; 95% CI, 1.27-1.78) and usually insomnia (HR, 1.46; 95% CI, 1.21-1.75) were associated with the risk of NAFLD. Furthermore, the easiness of getting up in the morning and insomnia showed a dose-response association with NAFLD (P_trend <0.05). MR analysis found consistent causal effects of NAFLD on easiness of getting up in the morning (OR, 0.995; 95% CI, 0.990-0.999; p = 0.033) and insomnia (OR, 1.006; 95% CI, 1.001-1.011; p = 0.024). These results were robust to weak instrument bias, pleiotropy, and heterogeneity. Conclusions: Findings showed consistent evidence of observational analyses and MR analyses that trouble getting up in the morning and insomnia were associated with an increased risk of NAFLD. Bidirectional MR demonstrated causal effects of NAFLD on sleep traits.

Mendelian Randomization Analyses Suggest Childhood Body Size Indirectly Influences End Points From Across the Cardiovascular Disease Spectrum Through Adult Body Size

Background Obesity is associated with long-term health consequences including cardiovascular disease. Separating the independent effects of childhood and adulthood obesity on cardiovascular disease risk is challenging as children with obesity typically remain overweight throughout the lifecourse. Methods and Results This study used 2-sample univariable and multivariable Mendelian randomization to estimate the effect of childhood body size both independently and after accounting for adult body size on 12 endpoints across the cardiovascular disease disease spectrum. Univariable analyses identified strong evidence of a total effect between genetically predicted childhood body size and increased risk of atherosclerosis, atrial fibrillation, coronary artery disease, heart failure, hypertension, myocardial infarction, peripheral artery disease, and varicose veins. However, evidence of a direct effect was weak after accounting for adult body size using multivariable Mendelian randomization, suggesting that childhood body size indirectly increases risk of these 8 disease outcomes via the pathway involving adult body size. Conclusions These findings suggest that the effect of genetically predicted childhood body size on the cardiovascular disease outcomes analyzed in this study are a result of larger body size persisting into adulthood. Further research is necessary to ascertain the critical timepoints where, if ever, the detrimental impact of obesity initiated in early life begins to become immutable.

The use of negative control outcomes in Mendelian randomization to detect potential population stratification

A key assumption of Mendelian randomization (MR) analysis is that there is no association between the genetic variants used as instruments and the outcome other than through the exposure of interest. One way in which this assumption can be violated is through population stratification, which can introduce confounding of the relationship between the genetic variants and the outcome and so induce an association between them. Negative control outcomes are increasingly used to detect unobserved confounding in observational epidemiological studies. Here we consider the use of negative control outcomes in MR studies to detect confounding of the genetic variants and the exposure or outcome. As a negative control outcome in an MR study, we propose the use of phenotypes which are determined before the exposure and outcome but which are likely to be subject to the same confounding as the exposure or outcome of interest. We illustrate our method with a two-sample MR analysis of a preselected set of exposures on self-reported tanning ability and hair colour. Our results show that, of the 33 exposures considered, genome-wide association studies (GWAS) of adiposity and education-related traits are likely to be subject to population stratification that is not controlled for through adjustment, and so any MR study including these traits may be subject to bias that cannot be identified through standard pleiotropy robust methods. Negative control outcomes should therefore be used regularly in MR studies to detect potential population stratification in the data used.

Methodological Challenges and Statistical Approaches in the COMprehensive Post-Acute Stroke Services Study

BACKGROUND

The COMprehensive Post-Acute Stroke Services study was a cluster-randomized pragmatic trial designed to evaluate a comprehensive care transitions model versus usual care. The data collected during this trial were complex and analysis methodology was required that could simultaneously account for the cluster-randomized design, missing patient-level covariates, outcome nonresponse, and substantial nonadherence to the intervention.

OBJECTIVE

The objective of this study was to discuss an array of complementary statistical methods to evaluate treatment effectiveness that appropriately addressed the challenges presented by the complex data arising from this pragmatic trial.

METHODS

We utilized multiple imputation combined with inverse probability weighting to account for missing covariate and outcome data in the estimation of intention-to-treat effects (ITT). The ITT estimand reflects the effectiveness of assignment to the COMprehensive Post-Acute Stroke Services intervention compared with usual care (ie, it does not take into account intervention adherence). Per-protocol analyses provide complementary information about the effect of treatment, and therefore are relevant for patients to inform their decision-making. We describe estimation of the complier average causal effect using an instrumental variables approach through 2-stage least squares estimation. For all preplanned analyses, we also discuss additional sensitivity analyses.

DISCUSSION

Pragmatic trials are well suited to inform clinical practice. Care should be taken to proactively identify the appropriate balance between control and pragmatism in trial design. Valid estimation of ITT and per-protocol effects in the presence of complex data requires application of appropriate statistical methods and concerted efforts to ensure high-quality data are collected.

Long-term cost-effectiveness of interventions for obesity: A mendelian randomisation study

BACKGROUND

The prevalence of obesity has increased in the United Kingdom, and reliably measuring the impact on quality of life and the total healthcare cost from obesity is key to informing the cost-effectiveness of interventions that target obesity, and determining healthcare funding. Current methods for estimating cost-effectiveness of interventions for obesity may be subject to confounding and reverse causation. The aim of this study is to apply a new approach using mendelian randomisation for estimating the cost-effectiveness of interventions that target body mass index (BMI), which may be less affected by confounding and reverse causation than previous approaches.

METHODS AND FINDINGS

We estimated health-related quality-adjusted life years (QALYs) and both primary and secondary healthcare costs for 310,913 men and women of white British ancestry aged between 39 and 72 years in UK Biobank between recruitment (2006 to 2010) and 31 March 2017. We then estimated the causal effect of differences in BMI on QALYs and total healthcare costs using mendelian randomisation. For this, we used instrumental variable regression with a polygenic risk score (PRS) for BMI, derived using a genome-wide association study (GWAS) of BMI, with age, sex, recruitment centre, and 40 genetic principal components as covariables to estimate the effect of a unit increase in BMI on QALYs and total healthcare costs. Finally, we used simulations to estimate the likely effect on BMI of policy relevant interventions for BMI, then used the mendelian randomisation estimates to estimate the cost-effectiveness of these interventions. A unit increase in BMI decreased QALYs by 0.65% of a QALY (95% confidence interval [CI]: 0.49% to 0.81%) per year and increased annual total healthcare costs by £42.23 (95% CI: £32.95 to £51.51) per person. When considering only health conditions usually considered in previous cost-effectiveness modelling studies (cancer, cardiovascular disease, cerebrovascular disease, and type 2 diabetes), we estimated that a unit increase in BMI decreased QALYs by only 0.16% of a QALY (95% CI: 0.10% to 0.22%) per year. We estimated that both laparoscopic bariatric surgery among individuals with BMI greater than 35 kg/m2, and restricting volume promotions for high fat, salt, and sugar products, would increase QALYs and decrease total healthcare costs, with net monetary benefits (at £20,000 per QALY) of £13,936 (95% CI: £8,112 to £20,658) per person over 20 years, and £546 million (95% CI: £435 million to £671 million) in total per year, respectively. The main limitations of this approach are that mendelian randomisation relies on assumptions that cannot be proven, including the absence of directional pleiotropy, and that genotypes are independent of confounders.

CONCLUSIONS

Mendelian randomisation can be used to estimate the impact of interventions on quality of life and healthcare costs. We observed that the effect of increasing BMI on health-related quality of life is much larger when accounting for 240 chronic health conditions, compared with only a limited selection. This means that previous cost-effectiveness studies have likely underestimated the effect of BMI on quality of life and, therefore, the potential cost-effectiveness of interventions to reduce BMI.

Association of Genetic Variants for Plasma LRG1 With Rapid Decline in Kidney Function in Patients With Type 2 Diabetes

CONTEXT

Elevated levels of plasma leucine-rich α-2-glycoprotein 1 (LRG1), a component of transforming growth factor beta signaling, are associated with development and progression of chronic kidney disease in patients with type 2 diabetes (T2D). However, whether this relationship is causal is uncertain.

OBJECTIVES

To identify genetic variants associated with plasma LRG1 levels and determine whether genetically predicted plasma LRG1 contributes to a rapid decline in kidney function (RDKF) in patients with T2D.

DESIGN AND PARTICIPANTS

We performed a genome-wide association study of plasma LRG1 among 3694 T2D individuals [1881 (983 Chinese, 420 Malay, and 478 Indian) discovery from Singapore Study of Macro-angiopathy and Micro-vascular Reactivity in Type 2 Diabetes cohort and 1813 (Chinese) validation from Diabetic Nephropathy cohort]. One- sample Mendelian randomization analysis was performed among 1337 T2D Chinese participants with preserved glomerular filtration function [baseline estimated glomerular filtration rate (eGFR) ≥60 mL/min/1.73 m2)]. RDKF was defined as an eGFR decline of 3 mL/min/1.73 m2/year or greater.

RESULTS

We identified rs4806985 variant near LRG1 locus robustly associated with plasma LRG1 levels (meta P = 6.66 × 10-16). Among 1337 participants, 344 (26%) developed RDKF, and the rs4806985 variant was associated with higher odds of RDKF (meta odds ratio = 1.23, P = 0.030 adjusted for age and sex). Mendelian randomization analysis provided evidence for a potential causal effect of plasma LRG1 on kidney function decline in T2D (P < 0.05).

CONCLUSION

We demonstrate that genetically influenced plasma LRG1 increases the risk of RDKF in T2D patients, suggesting plasma LRG1 as a potential treatment target. However, further studies are warranted to elucidate underlying pathways to provide insight into diabetic kidney disease prevention.

Body size and composition and risk of site-specific cancers in the UK Biobank and large international consortia: A mendelian randomisation study

BACKGROUND

Evidence for the impact of body size and composition on cancer risk is limited. This mendelian randomisation (MR) study investigates evidence supporting causal relationships of body mass index (BMI), fat mass index (FMI), fat-free mass index (FFMI), and height with cancer risk.

METHODS AND FINDINGS

Single nucleotide polymorphisms (SNPs) were used as instrumental variables for BMI (312 SNPs), FMI (577 SNPs), FFMI (577 SNPs), and height (293 SNPs). Associations of the genetic variants with 22 site-specific cancers and overall cancer were estimated in 367,561 individuals from the UK Biobank (UKBB) and with lung, breast, ovarian, uterine, and prostate cancer in large international consortia. In the UKBB, genetically predicted BMI was positively associated with overall cancer (odds ratio [OR] per 1 kg/m2 increase 1.01, 95% confidence interval [CI] 1.00-1.02; p = 0.043); several digestive system cancers: stomach (OR 1.13, 95% CI 1.06-1.21; p < 0.001), esophagus (OR 1.10, 95% CI 1.03, 1.17; p = 0.003), liver (OR 1.13, 95% CI 1.03-1.25; p = 0.012), and pancreas (OR 1.06, 95% CI 1.01-1.12; p = 0.016); and lung cancer (OR 1.08, 95% CI 1.04-1.12; p < 0.001). For sex-specific cancers, genetically predicted elevated BMI was associated with an increased risk of uterine cancer (OR 1.10, 95% CI 1.05-1.15; p < 0.001) and with a lower risk of prostate cancer (OR 0.97, 95% CI 0.94-0.99; p = 0.009). When dividing cancers into digestive system versus non-digestive system, genetically predicted BMI was positively associated with digestive system cancers (OR 1.04, 95% CI 1.02-1.06; p < 0.001) but not with non-digestive system cancers (OR 1.01, 95% CI 0.99-1.02; p = 0.369). Genetically predicted FMI was positively associated with liver, pancreatic, and lung cancer and inversely associated with melanoma and prostate cancer. Genetically predicted FFMI was positively associated with non-Hodgkin lymphoma and melanoma. Genetically predicted height was associated with increased risk of overall cancer (OR per 1 standard deviation increase 1.09; 95% CI 1.05-1.12; p < 0.001) and multiple site-specific cancers. Similar results were observed in analyses using the weighted median and MR-Egger methods. Results based on consortium data confirmed the positive associations between BMI and lung and uterine cancer risk as well as the inverse association between BMI and prostate cancer, and, additionally, showed an inverse association between genetically predicted BMI and breast cancer. The main limitations are the assumption that genetic associations with cancer outcomes are mediated via the proposed risk factors and that estimates for some lower frequency cancer types are subject to low precision.

CONCLUSIONS

Our results show that the evidence for BMI as a causal risk factor for cancer is mixed. We find that BMI has a consistent causal role in increasing risk of digestive system cancers and a role for sex-specific cancers with inconsistent directions of effect. In contrast, increased height appears to have a consistent risk-increasing effect on overall and site-specific cancers.

Preference-based instrumental variables in health research rely on important and underreported assumptions: a systematic review

OBJECTIVE

Preference-based instrumental variables (PP IV) designs can identify causal effects when patients receive treatment due to variation in providers' treatment preference. We offer a systematic review and methodological assessment of PP IV applications in health research.

STUDY DESIGN AND SETTING

We included studies that applied PP IV for evaluation of any treatment in any population in health research (PROSPERO: CRD42020165014). We searched within four databases (Medline, Web of Science, ScienceDirect, SpringerLink) and four journals (including full-text and title and abstract sources) between January 1, 1998, and March 5, 2020. We extracted data on areas of applications and methodology, including assumptions using Swanson and Hernan's (2013) guideline.

RESULTS

We included 185 of 1087 identified studies. The use of PP IV has increased, being predominantly used for treatment effects in cancer, cardiovascular disease, and mental health. The most common PP IV was treatment variation at the facility-level, followed by physician- and regional-level. Only 12 percent of applications report the four main assumptions for PP IV. Selection on treatment may be a potential issue in 46 percent of studies.

CONCLUSION

The assumptions of PP IV are not sufficiently reported in existing work. PP IV-studies should use reporting guidelines.

Causal Inference Methods to Integrate Omics and Complex Traits

Major biotechnological advances have facilitated a tremendous boost to the collection of (gen-/transcript-/prote-/methyl-/metabol-)omics data in very large sample sizes worldwide. Coordinated efforts have yielded a deluge of studies associating diseases with genetic markers (genome-wide association studies) or with molecular phenotypes. Whereas omics-disease associations have led to biologically meaningful and coherent mechanisms, the identified (non-germline) disease biomarkers may simply be correlates or consequences of the explored diseases. To move beyond this realm, Mendelian randomization provides a principled framework to integrate information on omics- and disease-associated genetic variants to pinpoint molecular traits causally driving disease development. In this review, we show the latest advances in this field, flag up key challenges for the future, and propose potential solutions.

Proxy gene-by-environment Mendelian randomization study confirms a causal effect of maternal smoking on offspring birthweight, but little evidence of long-term influences on offspring health

Background

A lack of genetic data across generations makes transgenerational Mendelian randomization (MR) difficult. We used UK Biobank and a novel proxy gene-by-environment MR to investigate effects of maternal smoking heaviness in pregnancy on offspring health, using participants’ (generation one: G1) genotype (rs16969968 in CHRNA5) as a proxy for their mothers’ (G0) genotype.

Methods

We validated this approach by replicating an established effect of maternal smoking heaviness on offspring birthweight. Then we applied this approach to explore effects of maternal (G0) smoking heaviness on offspring (G1) later life outcomes and on birthweight of G1 women’s children (G2).

Results

Each additional smoking-increasing allele in offspring (G1) was associated with a 0.018 [95% confidence interval (CI): -0.026, -0.009] kg lower G1 birthweight in maternal (G0) smoking stratum, but no meaningful effect (-0.002 kg; 95% CI: -0.008, 0.003) in maternal non-smoking stratum (interaction P-value = 0.004). The differences in associations of rs16969968 with grandchild’s (G2) birthweight between grandmothers (G0) who did, versus did not, smoke were heterogeneous (interaction P-value = 0.042) among mothers (G1) who did (-0.020 kg/allele; 95% CI: -0.044, 0.003), versus did not (0.007 kg/allele; 95% CI: -0.005, 0.020), smoke in pregnancy.

Conclusions

Our study demonstrated how offspring genotype can be used to proxy for the mother’s genotype in gene-by-environment MR. We confirmed the causal effect of maternal (G0) smoking on offspring (G1) birthweight, but found little evidence of an effect on G1 longer-term health outcomes. For grandchild’s (G2) birthweight, the effect of grandmother’s (G0) smoking heaviness in pregnancy may be modulated by maternal (G1) smoking status in pregnancy.

Effect of Education on Myopia: Evidence from the United Kingdom ROSLA 1972 Reform

Purpose

Cross-sectional and longitudinal studies have consistently reported an association between education and myopia. However, conventional observational studies are at risk of bias due to confounding by factors such as socioeconomic position and parental educational attainment. The current study aimed to estimate the causal effect of education on refractive error using regression discontinuity analysis.

Methods

Regression discontinuity analysis was applied to assess the influence on refractive error of the raising of the school leaving age (ROSLA) from 15 to 16 years introduced in England and Wales in 1972. For comparison, a conventional ordinary least squares (OLS) analysis was performed. The analysis sample comprised 21,548 UK Biobank participants born in a nine-year interval centered on September 1957, the date of birth of those first affected by ROSLA.

Results

In OLS analysis, the ROSLA 1972 reform was associated with a -0.29 D (95% confidence interval [CI]: -0.36 to -0.21, P < 0.001) more negative refractive error. In other words, the refractive error of the study sample became more negative by -0.29 D during the transition from a minimum school leaving age of 15 to 16 years of age. Regression discontinuity analysis estimated the causal effect of the ROSLA 1972 reform on refractive error as -0.77 D (95% CI: -1.53 to -0.02, P = 0.04).

Conclusions

Additional compulsory schooling due to the ROSLA 1972 reform was associated with a more negative refractive error, providing additional support for a causal relationship between education and myopia.

Use of Multivariable Mendelian Randomization to Address Biases Due to Competing Risk Before Recruitment

Background: Mendelian randomization (MR) provides unconfounded estimates. MR is open to selection bias when the underlying sample is selected on surviving to recruitment on the genetically instrumented exposure and competing risk of the outcome. Few methods to address this bias exist. Methods: We show that this selection bias can sometimes be addressed by adjusting for common causes of survival and outcome. We use multivariable MR to obtain a corrected MR estimate for statins on stroke. Statins affect survival, and stroke typically occurs later in life than ischemic heart disease (IHD), making estimates for stroke open to bias from competing risk. Results: In univariable MR in the UK Biobank, genetically instrumented statins did not protect against stroke [odds ratio (OR) 1.33, 95% confidence interval (CI) 0.80-2.20] but did in multivariable MR (OR 0.81, 95% CI 0.68-0.98) adjusted for major causes of survival and stroke [blood pressure, body mass index (BMI), and smoking initiation] with a multivariable Q-statistic indicating absence of selection bias. However, the MR estimate for statins on stroke using MEGASTROKE remained positive and the Q statistic indicated pleiotropy. Conclusion: MR studies of harmful exposures on late-onset diseases with shared etiology need to be conceptualized within a mechanistic understanding so as to identify any potential bias due to survival to recruitment on both genetically instrumented exposure and competing risk of the outcome, which may then be investigated using multivariable MR or estimated analytically and results interpreted accordingly.

Causal effect of children's secondary education on parental health outcomes: findings from a natural experiment in Botswana

OBJECTIVES

A growing literature highlights the intergenerational transmission of human capital from parents to children. However, far less is known about 'upward transmission' from children to parents. In this study, we use a 1996 Botswana education policy reform as a natural experiment to identify the causal effect of children's secondary schooling on their parents' health.

SETTING

Botswana's decennial census (2001 and 2011). Data were obtained through the Integrated Public Use Microdata Series and are 10% random samples of the complete population in each of these census years.

PARTICIPANTS

Survey respondents who were citizens born in Botswana, at least 18 years old at the time of the census and born in or after 1975 (n=89 721).

PRIMARY AND SECONDARY OUTCOME MEASURES

Parental survival and disability at the time of the census, separately for mothers and fathers.

RESULTS

The 1996 reform caused a large increase in grade 10 enrolment, inducing an additional 0.4 years of schooling for the first cohorts affected (95% CI 0.3 to 0.5, p<0.001). The reform, however, had no effect on parental survival and disability by the time exposed child cohorts reach age 30. Results were robust to a wide array of sensitivity analyses.

CONCLUSIONS

This study found little evidence that parents' survival and disability were affected by their offspring's educational attainment in Botswana. Parents' health may not be necessarily affected by increasing their offspring's educational attainment.

The causal effects of health conditions and risk factors on social and socioeconomic outcomes: Mendelian randomization in UK Biobank

BACKGROUND

We aimed to estimate the causal effect of health conditions and risk factors on social and socioeconomic outcomes in UK Biobank. Evidence on socioeconomic impacts is important to understand because it can help governments, policy makers and decision makers allocate resources efficiently and effectively.

METHODS

We used Mendelian randomization to estimate the causal effects of eight health conditions (asthma, breast cancer, coronary heart disease, depression, eczema, migraine, osteoarthritis, type 2 diabetes) and five health risk factors [alcohol intake, body mass index (BMI), cholesterol, systolic blood pressure, smoking] on 19 social and socioeconomic outcomes in 336 997 men and women of White British ancestry in UK Biobank, aged between 39 and 72 years. Outcomes included annual household income, employment, deprivation [measured by the Townsend deprivation index (TDI)], degree-level education, happiness, loneliness and 13 other social and socioeconomic outcomes.

RESULTS

Results suggested that BMI, smoking and alcohol intake affect many socioeconomic outcomes. For example, smoking was estimated to reduce household income [mean difference = -£22 838, 95% confidence interval (CI): -£31 354 to -£14 321] and the chance of owning accommodation [absolute percentage change (APC) = -20.8%, 95% CI: -28.2% to -13.4%], of being satisfied with health (APC = -35.4%, 95% CI: -51.2% to -19.5%) and of obtaining a university degree (APC = -65.9%, 95% CI: -81.4% to -50.4%), while also increasing deprivation (mean difference in TDI = 1.73, 95% CI: 1.02 to 2.44, approximately 216% of a decile of TDI). There was evidence that asthma decreased household income, the chance of obtaining a university degree and the chance of cohabiting, and migraine reduced the chance of having a weekly leisure or social activity, especially in men. For other associations, estimates were null.

CONCLUSIONS

Higher BMI, alcohol intake and smoking were all estimated to adversely affect multiple social and socioeconomic outcomes. Effects were not detected between health conditions and socioeconomic outcomes using Mendelian randomization, with the exceptions of depression, asthma and migraines. This may reflect true null associations, selection bias given the relative health and age of participants in UK Biobank, and/or lack of power to detect effects.

Association between major depressive disorder and multiple disease outcomes: a phenome-wide Mendelian randomisation study in the UK Biobank

Depression affects all aspects of an individual's life but evidence relating to the causal effects on health is limited. We used information from 337,536 UK Biobank participants and performed hypothesis-free phenome-wide association analyses between major depressive disorder (MDD) genetic risk score (GRS) and 925 disease outcomes. GRS-disease outcome associations passing the multiple-testing corrected significance threshold (P < 1.9 × 10^-3) were followed by Mendelian randomisation (MR) analyses to test for causality. MDD GRS was associated with 22 distinct diseases in the phenome-wide discovery stage, with the strongest signal observed for MDD diagnosis and related co-morbidities including anxiety and sleep disorders. In inverse-variance weighted MR analyses, MDD was associated with several inflammatory and haemorrhagic gastrointestinal diseases, including oesophagitis (OR 1.32, 95% CI 1.18-1.48), non-infectious gastroenteritis (OR 1.25, 95% CI 1.06-1.48), gastrointestinal haemorrhage (OR 1.26, 95% CI 1.11-1.43) and intestinal E.coli infections (OR 3.24, 95% CI 1.74-6.02). Signals were also observed for symptoms/disorders of the urinary system (OR 1.36, 95% CI 1.19-1.56), asthma (OR 1.23, 95% CI 1.06-1.44), and painful respiration (OR 1.28, 95% CI 1.14-1.44). MDD was associated with disorders of lipid metabolism (OR 1.22, 95% CI 1.12-1.34) and ischaemic heart disease (OR 1.30, 95% CI 1.15-1.47). Sensitivity analyses excluding pleiotropic variants provided consistent associations. Our study indicates a causal link between MDD and a broad range of diseases, suggesting a notable burden of co-morbidity. Early detection and management of MDD is important, and treatment strategies should be selected to also minimise the risk of related co-morbidities.

Within family Mendelian randomization studies

Mendelian randomization (MR) is increasingly used to make causal inferences in a wide range of fields, from drug development to etiologic studies. Causal inference in MR is possible because of the process of genetic inheritance from parents to offspring. Specifically, at gamete formation and conception, meiosis ensures random allocation to the offspring of one allele from each parent at each locus, and these are unrelated to most of the other inherited genetic variants. To date, most MR studies have used data from unrelated individuals. These studies assume that genotypes are independent of the environment across a sample of unrelated individuals, conditional on covariates. Here we describe potential sources of bias, such as transmission ratio distortion, selection bias, population stratification, dynastic effects and assortative mating that can induce spurious or biased SNP-phenotype associations. We explain how studies of related individuals such as sibling pairs or parent-offspring trios can be used to overcome some of these sources of bias, to provide potentially more reliable evidence regarding causal processes. The increasing availability of data from related individuals in large cohort studies presents an opportunity to both overcome some of these biases and also to evaluate familial environmental effects.

Guidelines for performing Mendelian randomization investigations

This paper provides guidelines for performing Mendelian randomization investigations. It is aimed at practitioners seeking to undertake analyses and write up their findings, and at journal editors and reviewers seeking to assess Mendelian randomization manuscripts. The guidelines are divided into nine sections: motivation and scope, data sources, choice of genetic variants, variant harmonization, primary analysis, supplementary and sensitivity analyses (one section on robust statistical methods and one on other approaches), data presentation, and interpretation. These guidelines will be updated based on feedback from the community and advances in the field. Updates will be made periodically as needed, and at least every 18 months.

Mendelian Randomization analysis of the causal effect of adiposity on hospital costs

Estimates of the marginal effect of measures of adiposity such as body mass index (BMI) on healthcare costs are important for the formulation and evaluation of policies targeting adverse weight profiles. Most estimates of this association are affected by endogeneity bias. We use a novel identification strategy exploiting Mendelian Randomization - random germline genetic variation modelled using instrumental variables - to identify the causal effect of BMI on inpatient hospital costs. Using data on over 300,000 individuals, the effect size per person per marginal unit of BMI per year varied according to specification, including £21.22 (95% confidence interval (CI): £14.35-£28.07) for conventional inverse variance weighted models to £18.85 (95% CI: £9.05-£28.65) for penalized weighted median models. Effect sizes from Mendelian Randomization models were larger in most cases than non-instrumental variable multivariable adjusted estimates (£13.47, 95% CI: £12.51-£14.43). There was little evidence of non-linearity. Within-family estimates, intended to address dynastic biases, were imprecise.

Mendel's laws, Mendelian randomization and causal inference in observational data: substantive and nomenclatural issues

We respond to criticisms of Mendelian randomization (MR) by Mukamal, Stampfer and Rimm (MSR). MSR consider that MR is receiving too much attention and should be renamed. We explain how MR links to Mendel's laws, the origin of the name and our lack of concern regarding nomenclature. We address MSR's substantive points regarding MR of alcohol and cardiovascular disease, an issue on which they dispute the MR findings. We demonstrate that their strictures with respect to population stratification, confounding, weak instrument bias, pleiotropy and confounding have been addressed, and summarise how the field has advanced in relation to the issues they raise. We agree with MSR that "the hard problem of conducting high-quality, reproducible epidemiology" should be addressed by epidemiologists. However we see more evidence of confrontation of this issue within MR, as opposed to conventional observational epidemiology, within which the same methods that have demonstrably failed in the past are simply rolled out into new areas, leaving their previous failures unexamined.

Guidelines for performing Mendelian randomization investigations

This paper provides guidelines for performing Mendelian randomization investigations. It is aimed at practitioners seeking to undertake analyses and write up their findings, and at journal editors and reviewers seeking to assess Mendelian randomization manuscripts. The guidelines are divided into nine sections: motivation and scope, data sources, choice of genetic variants, variant harmonization, primary analysis, supplementary and sensitivity analyses (one section on robust methods and one on other approaches), data presentation, and interpretation. These guidelines will be updated based on feedback from the community and advances in the field. Updates will be made periodically as needed, and at least every 18 months.

Assessing causal treatment effect estimation when using large observational datasets

Background

Recently, there has been a heightened interest in developing and evaluating different methods for analysing observational data. This has been driven by the increased availability of large data resources such as Electronic Health Record (EHR) data alongside known limitations and changing characteristics of randomised controlled trials (RCTs). A wide range of methods are available for analysing observational data. However, various, sometimes strict, and often unverifiable assumptions must be made in order for the resulting effect estimates to have a causal interpretation. In this paper we will compare some common approaches to estimating treatment effects from observational data in order to highlight the importance of considering, and justifying, the relevant assumptions prior to conducting an observational analysis.

Methods

A simulation study was conducted based upon a small cohort of patients with chronic obstructive pulmonary disease. Two-stage least squares instrumental variables, propensity score, and linear regression models were compared under a range of different scenarios including different strengths of instrumental variable and unmeasured confounding. The effects of violating the assumptions of the instrumental variables analysis were also assessed. Sample sizes of up to 200,000 patients were considered.

Results

Two-stage least squares instrumental variable methods can yield unbiased treatment effect estimates in the presence of unmeasured confounding provided the sample size is sufficiently large. Adjusting for measured covariates in the analysis reduces the variability in the two-stage least squares estimates. In the simulation study, propensity score methods produced very similar results to linear regression for all scenarios. A weak instrument or strong unmeasured confounding led to an increase in uncertainty in the two-stage least squares instrumental variable effect estimates. A violation of the instrumental variable assumptions led to bias in the two-stage least squares effect estimates. Indeed, these were sometimes even more biased than those from a naïve linear regression model.

Conclusions

Instrumental variable methods can perform better than naïve regression and propensity scores. However, the assumptions need to be carefully considered and justified prior to conducting an analysis or performance may be worse than if the problem of unmeasured confounding had been ignored altogether.

MR-pheWAS with stratification and interaction: Searching for the causal effects of smoking heaviness identified an effect on facial aging

Mendelian randomization (MR) is an established approach to evaluate the effect of an exposure on an outcome. The gene-by-environment (GxE) study design can be used to determine whether the genetic instrument affects the outcome through pathways other than via the exposure of interest (horizontal pleiotropy). MR phenome-wide association studies (MR-pheWAS) search for the effects of an exposure, and can be conducted in UK Biobank using the PHESANT package. In this proof-of-principle study, we introduce the novel GxE MR-pheWAS approach, that combines MR-pheWAS with the use of GxE interactions. This method aims to identify the presence of effects of an exposure while simultaneously investigating horizontal pleiotropy. We systematically test for the presence of causal effects of smoking heaviness-stratifying on smoking status (ever versus never)-as an exemplar. If a genetic variant is associated with smoking heaviness (but not smoking initiation), and this variant affects an outcome (at least partially) via tobacco intake, we would expect the effect of the variant on the outcome to differ in ever versus never smokers. We used PHESANT to test for the presence of effects of smoking heaviness, instrumented by genetic variant rs16969968, among never and ever smokers respectively, in UK Biobank. We ranked results by the strength of interaction between ever and never smokers. We replicated previously established effects of smoking heaviness, including detrimental effects on lung function. Novel results included a detrimental effect of heavier smoking on facial aging. We have demonstrated how GxE MR-pheWAS can be used to identify potential effects of an exposure, while simultaneously assessing whether results may be biased by horizontal pleiotropy.

Understanding the consequences of education inequality on cardiovascular disease: mendelian randomisation study

OBJECTIVES

To investigate the role of body mass index (BMI), systolic blood pressure, and smoking behaviour in explaining the effect of education on the risk of cardiovascular disease outcomes.

DESIGN

Mendelian randomisation study.

SETTING

UK Biobank and international genome-wide association study data.

PARTICIPANTS

Predominantly participants of European ancestry.

EXPOSURE

Educational attainment, BMI, systolic blood pressure, and smoking behaviour in observational analysis, and randomly allocated genetic variants to instrument these traits in mendelian randomisation.

MAIN OUTCOMES MEASURE

The risk of coronary heart disease, stroke, myocardial infarction, and cardiovascular disease (all subtypes; all measured in odds ratio), and the degree to which this is mediated through BMI, systolic blood pressure, and smoking behaviour respectively.

RESULTS

Each additional standard deviation of education (3.6 years) was associated with a 13% lower risk of coronary heart disease (odds ratio 0.86, 95% confidence interval 0.84 to 0.89) in observational analysis and a 37% lower risk (0.63, 0.60 to 0.67) in mendelian randomisation analysis. As a proportion of the total risk reduction, BMI was estimated to mediate 15% (95% confidence interval 13% to 17%) and 18% (14% to 23%) in the observational and mendelian randomisation estimates, respectively. Corresponding estimates were 11% (9% to 13%) and 21% (15% to 27%) for systolic blood pressure and 19% (15% to 22%) and 34% (17% to 50%) for smoking behaviour. All three risk factors combined were estimated to mediate 42% (36% to 48%) and 36% (5% to 68%) of the effect of education on coronary heart disease in observational and mendelian randomisation analyses, respectively. Similar results were obtained when investigating the risk of stroke, myocardial infarction, and cardiovascular disease.

CONCLUSIONS

BMI, systolic blood pressure, and smoking behaviour mediate a substantial proportion of the protective effect of education on the risk of cardiovascular outcomes and intervening on these would lead to reductions in cases of cardiovascular disease attributable to lower levels of education. However, more than half of the protective effect of education remains unexplained and requires further investigation.