Estimation of the proportion of treatment effect explained by a high-dimensional surrogate

Surrogate Marker Evaluation: A Tutorial Using R

The practice of using a surrogate marker to replace a primary outcome in clinical studies has become widespread. Typically, the primary outcome requires long-term patient follow-up, is expensive, or is invasive or burdensome for patients to measure, while the surrogate marker is not (or less so). Of course, a surrogate marker must be validated before it should be used to make a decision about the effectiveness of a treatment. There has been a tremendous amount of statistical and clinical research focused on evaluating and validating surrogate markers over the past 35 years. Although there is ongoing debate over the optimal evaluation method, the development of new approaches and insights has greatly enriched the field. In this tutorial, we describe available statistical frameworks for evaluating a surrogate marker and specifically focus on the practical implementation of the proportion of treatment effect explained framework. We consider both uncensored and censored outcomes, parametric and non-parametric estimation, evaluating multiple surrogates, heterogeneity in the utility of the surrogate marker, surrogate evaluation from a prediction perspective, and the surrogate paradox. We include R code to implement these procedures with a follow-along R markdown. We close with a discussion on open problems in this research area, particularly in terms of using the surrogate marker to test for treatment in a future study, which is the ultimate goal of surrogate marker evaluation.

Circulating fatty acid profiles impact total, cardiovascular disease, and cancer mortality in a population-based prospective cohort study

BACKGROUND

Evidence linking circulating fatty acids (FAs) to mortality from age-related chronic diseases was limited and inconsistent. We aimed to investigate the associations of plasma FAs with total, cardiovascular disease (CVD), and cancer mortality and explore the potential mechanism.

METHODS

117,871 individuals were prospectively followed in the UK Biobank. Circulating FAs were measured by a high-throughput NMR-based metabolic platform. Causes and dates of death were collected from death certificates according to the code of International Statistical Classification of Diseases (ICD-10).

RESULTS

Over a median follow-up of 11.9 years, 7805 (6.6 %) deaths occurred. Plasma saturated FAs (SFAs) were positively associated with total mortality risk while plasma polyunsaturated FAs (PUFAs) exhibited an inverse association. For cause-specific mortality, circulating PUFAs, linoleic acid (LA), and n-3 PUFAs were associated with 34 %, 30 %, and 37 % lower risk of CVD mortality, respectively. Moreover, plasma n-3 PUFAs were related to a 24 % lower risk of cancer mortality. However, circulating non-LA n-6 PUFAs were associated with 11 % and 22 % higher risk of total and cancer mortality, respectively. Serum levels of C-reactive protein (CRP) and apolipoprotein A (ApoA) had significant mediation effects on these associations. Additionally, the inverse association of plasma n-6 PUFAs with total mortality only existed among carriers of the GG genotype at rs16966952 and the inverse association of plasma PUFAs with CVD mortality was only observed among TT genotype carriers at rs174547.

CONCLUSIONS

Circulating PUFAs, particularly n-3 PUFAs and LA, were inversely related to premature death from chronic diseases and longevity. Inflammatory and lipid metabolism partially explained these associations. Genetic interactions with rs16966952 and rs174547 further modified these associations.

Circulating fatty acids and risk of severe non-alcoholic fatty liver disease in the UK biobank: a prospective cohort of 116 223 individuals

Fatty acid (FA) metabolism plays an important role in the development of nonalcoholic fatty liver disease (NAFLD). However, data on the relationship between circulating FAs and NAFLD risk are limited. This study aims to assess the associations between specific circulating FAs and severe NAFLD risk among the general population. Overall 116 223 participants without NAFLD and other liver diseases from the UK Biobank were enrolled between 2006 and 2010 and were followed up until the end of 2021. Plasma concentrations of saturated fatty acids (SFAs), monounsaturated fatty acids (MUFAs), and polyunsaturated fatty acids (PUFAs) were analyzed using an NMR-based biomarker profiling platform. Hazard ratios (HRs) and 95% confidence intervals (CIs) of NAFLD risk were estimated using Cox proportional-hazard models adjusted for other potential confounders. During a mean follow-up of 12.3 years, we documented 1394 cases of severe NAFLD. After multivariate adjustment, plasma SFAs and MUFAs were associated with a higher risk of severe NAFLD, whereas plasma n-3 PUFAs, n-6 PUFAs, and linoleic acid (LA) were associated with a lower risk. As compared with the lowest quartile, HRs (95% CIs) of severe NAFLD risk in the highest quartiles were 1.85 (1.45-2.36) for SFAs, 1.74 (1.23-2.44) for MUFAs, 0.79 (0.65-0.97) for n-3 PUFAs, 0.68 (0.48-0.96) for n-6 PUFAs, and 0.73 (0.54-0.99) for LA. The significant relationships were mainly mediated by serum TG for SFAs, HDL-C for MUFAs and n-6 PUFAs, and C-reactive protein for n-3 PUFAs. Plasma SFAs were associated with a more pronounced increase in the risk of severe NAFLD among participants with fewer SFA-associated alleles (P interaction = 0.032). Dietary recommendations for reducing plasma SFAs and MUFAs while increasing n-3 and n-6 PUFAs may be protective for severe NAFLD, which could be mediated by lipid metabolism and inflammation.

Doubly robust evaluation of high-dimensional surrogate markers

When evaluating the effectiveness of a treatment, policy, or intervention, the desired measure of efficacy may be expensive to collect, not routinely available, or may take a long time to occur. In these cases, it is sometimes possible to identify a surrogate outcome that can more easily, quickly, or cheaply capture the effect of interest. Theory and methods for evaluating the strength of surrogate markers have been well studied in the context of a single surrogate marker measured in the course of a randomized clinical study. However, methods are lacking for quantifying the utility of surrogate markers when the dimension of the surrogate grows. We propose a robust and efficient method for evaluating a set of surrogate markers that may be high-dimensional. Our method does not require treatment to be randomized and may be used in observational studies. Our approach draws on a connection between quantifying the utility of a surrogate marker and the most fundamental tools of causal inference-namely, methods for robust estimation of the average treatment effect. This connection facilitates the use of modern methods for estimating treatment effects, using machine learning to estimate nuisance functions and relaxing the dependence on model specification. We demonstrate that our proposed approach performs well, demonstrate connections between our approach and certain mediation effects, and illustrate it by evaluating whether gene expression can be used as a surrogate for immune activation in an Ebola study.

Association of Fish Oil Supplementation with Risk of Coronary Heart Disease in Individuals with Diabetes and Prediabetes: A Prospective Study in the UK Biobank

This study aimed to explore the association between habitual intake of fish oil supplementation and the risk of developing CHD in patients with prediabetes and diabetes. Habitual use of fish oil was assessed by repeated questionnaires. Cox proportional hazard models were applied to calculate hazard ratios (HRs) and 95% confidence intervals (CIs). Over a median follow-up of 11.6 years, 4304 and 3294 CHD cases were documented among 47,663 individuals with prediabetes and 22,146 patients with diabetes in the UK Biobank, respectively. After multivariable adjustment, the HRs (95% CI) of CHD were 0.91 (0.85-0.98) and 0.87 (0.80-0.95) for individuals utilizing fish oil supplementation compared with non-users among the participants with prediabetes and diabetes, respectively. Furthermore, we identified an inverse relationship between fish oil use and CHD incidence, which was significantly mediated by serum C-reactive protein (CRP) levels in individuals with prediabetes and by very-low-density lipoprotein cholesterol (VLDL-C) in patients with diabetes at baseline. The inverse associations were consistent in the analyses stratified by potential confounders. In conclusion, the consumption of fish oil supplements was linked to decreased serum CRP and VLDL-C levels and subsequent CHD risk among adults with prediabetes and diabetes. Our findings highlight the important role of the habitual intake of fish oil supplements in preventing CHD in individuals with impaired glucose metabolism.