JFJohn Ferguson

John Ferguson

*This is an algorithmically generated author record based on publicly available information.

Current Organizations / Institutes

The organizations/institutes from the scholar's latest article

Previous Organizations / Institutes

The organizations / institutes from the scholar's other articles in the past ten years


John Ferguson has a total of 90 co-authors. Here are the top co-authors by number of shared publications:


Latest Publications

  • Top Cited Papers
    Open Access

    Estimating and displaying population attributable fractions using the R package: graphPAF

    by John Ferguson, Maurice O’cOnnell

    Here we introduce graphPAF, a comprehensive R package designed for estimation, inference and display of population attributable fractions (PAF) and impact fractions. In addition to allowing inference for standard population attributable fractions and impact fractions, graphPAF facilitates display of attributable fractions over multiple risk factors using fan-plots and nomograms, calculations of attributable fractions for continuous exposures, inference for attributable fractions appropriate for specific risk factor $$\rightarrow $$ → mediator $$\rightarrow $$ → outcome pathways (pathway-specific attributable fractions) and Bayesian network-based calculations and inference for joint, sequential and average population attributable fractions in multi-risk factor scenarios. This article can be used as both a guide to the theory of attributable fraction estimation and a tutorial regarding how to use graphPAF in practical examples.


  • Open Access

    Bias assessment and correction for Levin’s population attributable fraction in the presence of confounding

    by John Ferguson, Alberto Alvarez, Martin Mulligan, Conor Judge, Martin O’dOnnell

    In 1953, Morton Levin introduced a simple approach to estimating population attributable fractions (PAF) depending only on risk factor prevalence and relative risk. This formula and its extensions are still in widespread use today, particularly to estimate PAF in populations where individual data is unavailable. Unfortunately, Levin’s approach is known to be asymptotically biased for the PAF when the risk factor-disease relationship is confounded even if relative risks that are correctly adjusted for confounding are used in the estimator. Here we describe a simple re-expression of Miettinen’s estimand that depends on the causal relative risk, the unadjusted relative risk and the population risk factor prevalence. While this re-expression is not new, it has been underappreciated in the literature, and the associated estimator may be useful in estimating PAF in populations when individual data is unavailable provided estimated adjusted and unadjusted relative risks can be transported to the population of interest. Using the re-expressed estimand, we develop novel analytic formulae for the relative and absolute asymptotic bias in Levin’s formula, solidifying earlier work by Darrow and Steenland that used simulations to investigate this bias. We extend all results to settings with non-binary valued risk factors and continuous exposures and discuss the utility of these results in estimating PAF in practice.


  • Open Access

    Bias assessment and correction for Levin’s population attributable fraction in the presence of confounding

    by John Ferguson, Alberto Alvarez, Martin Mulligan, Conor Judge, Martin O'Donnell

    In 1953, Morton Levin introduced a simple approach to estimating population attributable fractions (PAF) depending only on risk factor prevalence and relative risk. This formula and its extensions are still in widespread use today, particularly to estimate PAF in populations where individual data is unavailable. Unfortunately, Levin’s approach is known to be asymptotically biased for the PAF when the risk factor-disease relationship is confounded even if relative risks that are correctly adjusted for confounding are used in the estimator. An alternative estimator, first introduced by Miettinen in 1972, is unbiased for the PAF provided the true relative risk is invariant across confounder strata. However, despite its statistical superiority, Miettinen’s estimator is seldom used in practice, as its direct application requires an estimate of risk factor prevalence within disease cases rather than an estimate of risk factor prevalence in the general population. Here we describe a simple re-expression of Miettinen’s estimand that depends on the causal relative risk, the unadjusted relative risk and the population risk factor prevalence. While this re-expression is not new, it has been underappreciated in the literature, and the associated estimator may be useful in estimating PAF in populations when individual data is unavailable provided estimated adjusted and unadjusted relative risks can be transported to the population of interest. Using the re-expressed estimand, we develop novel analytic formulae for the relative and absolute asymptotic bias in Levin’s formula, solidifying earlier work by Darrow and Steenland that used simulations to investigate this bias. We extend all results to settings with non-binary valued risk factors and continuous exposures and discuss the utility of these results in estimating PAF in practice.

  • Depressive Symptoms and Risk of Acute Stroke

    by Robert P. Murphy, Catriona Reddin, Annika Rosengren, Conor Judge, Graeme J. Hankey, John Ferguson, Alberto Alvarez-Iglesias, Shahram Oveisgharan, Mohammad Wasay, Clodagh McDermott, et al.

    Background and Objectives Depression has been reported to be a risk factor for acute stroke, based largely on studies in high-income countries. In the INTERSTROKE study, we explored the contribution of depressive symptoms to acute stroke risk and 1-month outcome across regions of the world, within subpopulations and by stroke type Methods INTERSTROKE is an international case-control study of risk factors for first acute stroke, conducted in 32 countries. Cases were patients with CT or MRI confirmed incident acute hospitalized stroke, and controls were matched for age, sex, and within sites. Standardized questions asked about self-reported depressive symptoms during the previous 12 months and use of prescribed antidepressant medications were recorded. Multivariable conditional logistic regression was used to determine the association of pre-stroke depressive symptoms with acute stroke risk. Adjusted ordinal logistic regression was used to explore the association of pre-stroke depressive symptoms with post-stroke functional outcome, measured with the modified-Rankin scale at 1-month after stroke. Results Of 26,877 participants, 40.4% were women, the mean age was 61.7 ± 13.4 years. The prevalence of depressive symptoms within the last 12 months was higher in cases compared to controls (18.3%vs.14.1%,p < 0.001) and differed by region (pinteraction < 0.001), with lowest prevalence in China (6.9% in controls) and highest in South America (32.2% of controls). In multivariable analyses pre-stroke depressive symptoms were associated with greater odds of acute stroke (OR 1.46, 95%CI 1.34–1.58), which was significant for both intracerebral hemorrhage (OR 1.56, 95%CI 1.28–1.91) and ischemic stroke (OR 1.44, 95%CI 1.31–1.58). A larger magnitude of association with stroke was seen in patients with a greater burden of depressive symptoms. While pre-admission depressive symptoms were not associated with a greater odds of worse baseline stroke severity (OR 1.02, 95%CI 0.94–1.10), they were associated with a greater odds of poor functional outcome at 1-month after acute stroke (OR 1.09, 95%CI 1.01–1.19). Discussion In this global study we recorded that depressive symptoms are an important risk factor for acute stroke, including both ischemic and hemorrhagic stroke. Pre-admission depressive symptoms were associated with poorer functional outcome, but not baseline stroke severity, suggesting an adverse role of depressive symptoms in post-stroke recovery.


  • Open Access

    Estimating and displaying population attributable fractions using the R package graphPAF

    by John Ferguson, Maurice O'COnnell

    GraphPAF is a comprehensive R package designed for estimation, inference and display of population attributable fractions (PAF)s and impact fractions. In addition to allowing inference for standard PAFs and impact fractions, graphPAF facilitates display of PAFs over multiple risk factors using fan plots and nomograms, calculations of PAFs for continuous exposures, inference for PAFs appropriate for specific risk factor → mediator → outcome pathways (pathway-specific PAFs) and Bayesian network based calculations and inference for joint, sequential and average PAFs in scenarios where multiple risk factors are of interest. In summary, graphPAF extends and consolidates existing packages for PAF estimation in multiple ways. This article serves as a broad overview of theory and estimation approaches appropriate for attributable fractions, as well as a guide regarding how to use the graphPAF package in practice.


  • Open Access

    Bias assessment and correction for Levin’s population attributable fraction in the presence of confounding

    by John Ferguson, Alberto Alvarez, Martin Mulligan, Conor Judge, Martin O’donnell

    In 1953, Morton Levin introduced a simple approach to estimating population attributable fractions (PAF) depending only on risk factor prevalence and relative risk. This formula and its extensions are still in widespread use today, particularly to estimate PAF in populations where individual data is unavailable. Unfortunately, Levin’s approach is known to be asymptotically biased for the PAF when the risk factor-disease relationship is confounded even if relative risks that are correctly adjusted for confounding are used in the estimator. An alternative estimator, first introduced by Miettinen in 1972, is unbiased for the PAF provided the true relative risk is invariant across confounder strata. However, despite its statistical superiority, Miettinen’s estimator is seldom used in practice, as its direct application requires an estimate of risk factor prevalence within disease cases rather than an estimate of risk factor prevalence in the general population. Here we describe a simple re-expression of Miettinen’s estimand that depends on the causal relative risk, the unadjusted relative risk and the population risk factor prevalence. While this re-expression is not new, it has been underappreciated in the literature, and the associated estimator may be useful in estimating PAF in populations when individual data is unavailable provided estimated adjusted and unadjusted relative risks can be transported to the population of interest. Using the re-expressed estimand, we develop novel analytic formulae for the relative and absolute asymptotic bias in Levin’s formula, solidifying earlier work by Darrow and Steenland that used simulations to investigate this bias. We extend all results to settings with non-binary valued risk factors and continuous exposures and discuss the utility of these results in estimating PAF in practice.


  • Open Access

    Association of Lipids, Lipoproteins, and Apolipoproteins with Stroke Subtypes in an International Case Control Study (INTERSTROKE)

    by Martin J. O’dOnnell, Matthew McQueen, Allan Sniderman, Guillaume Pare, Xingyu Wang, Graeme J. Hankey, Sumathy Rangarajan, Siu Lim Chin, Purnima Rao-Melacini, John Ferguson, et al.

    Background and Purpose The association of dyslipidemia with stroke has been inconsistent, which may be due to differing associations within etiological stroke subtypes. We sought to determine the association of lipoproteins and apolipoproteins within stroke subtypes.Methods Standardized incident case-control STROKE study in 32 countries. Cases were patients with acute hospitalized first stroke, and matched by age, sex and site to controls. Concentrations of total cholesterol, high-density lipoprotein cholesterol (HDL-C), low-density lipoprotein cholesterol (LDL-C), apolipoprotein A1 (apoA1), and apoB were measured. Non-HDL-C was calculated. We estimated multivariable odds ratio (OR) and population attributable risk percentage (PAR%). Outcome measures were all stroke, ischemic stroke (and subtypes), and intracerebral hemorrhage (ICH).Results Our analysis included 11,898 matched case-control pairs; 77.3% with ischemic stroke and 22.7% with ICH. Increasing apoB (OR, 1.10; 95% confidence interval [CI], 1.06 to 1.14 per standard deviation [SD]) and LDL-C (OR, 1.06; 95% CI, 1.02 to 1.10 per SD) were associated with an increase in risk of ischemic stroke, but a reduced risk of ICH. Increased apoB was significantly associated with large vessel stroke (PAR 13.4%; 95% CI, 5.6 to 28.4) and stroke of undetermined cause. Higher HDL-C (OR, 0.75; 95% CI, 0.72 to 0.78 per SD) and apoA1 (OR, 0.63; 95% CI, 0.61 to 0.66 per SD) were associated with ischemic stroke (and subtypes). While increasing HDL-C was associated with an increased risk of ICH (OR, 1.20; 95% CI, 1.14 to 1.27 per SD), apoA1 was associated with a reduced risk (OR, 0.80; 95% CI, 0.75 to 0.85 per SD). ApoB/A1 (OR, 1.38; 95% CI, 1.32 to 1.44 per SD) had a stronger magnitude of association than the ratio of LDL-C/HDL-C (OR, 1.26; 95% CI, 1.21 to 1.31 per SD) with ischemic stroke (P<0.0001). Conclusions The pattern and magnitude of association of lipoproteins and apolipoproteins with stroke varies by etiological stroke subtype. While the directions of association for LDL, HDL, and apoB were opposing for ischemic stroke and ICH, apoA1 was associated with a reduction in both ischemic stroke and ICH. The ratio of apoB/A1 was the best lipid predictor of ischemic stroke risk.


  • Open Access

    Pathway-specific population attributable fractions

    by Maurice M O’cOnnell, John P Ferguson

    Introduction: A population attributable fraction represents the relative change in disease prevalence that one might expect if a particular exposure was absent from the population. Often, one might be interested in what percentage of this effect acts through particular pathways. For instance, the effect of a sedentary lifestyle on stroke risk may be mediated by blood pressure, body mass index and several other intermediate risk factors. Methods: We define a new metric, the pathway-specific population attributable fraction (PS-PAF), for mediating pathways of interest. PS-PAFs can be informally defined as the relative change in disease prevalence from an intervention that shifts the distribution of the mediator to its expected distribution if the risk factor were eliminated, and sometimes more simply as the relative change in disease prevalence if the mediating pathway were disabled. A potential outcomes framework is used for formal definitions and associated estimands are derived via relevant identifiability conditions. Computationally efficient estimators for PS-PAFs are derived based on these identifiability conditions. Results: Calculations are demonstrated using INTERSTROKE—an international case–control study designed to quantify disease burden attributable to a number of known causal risk factors. The applied results suggest that mediating pathways from physical activity through blood pressure, blood lipids and body size explain comparable proportions of stroke disease burden, but a large proportion of the disease burden due to physical inactivity may be explained by alternative pathways. Conclusion: PS-PAFs measure disease burden attributable to differing mediating pathways and can generate insights into the dominant mechanisms by which a risk factor affects disease at a population level.


  • Open Access

    Association of Oral Anticoagulation With Stroke in Atrial Fibrillation or Heart Failure

    by Catriona Reddin, Conor Judge, Elaine Loughlin, Robert Murphy, Maria Costello, Alberto Alvarez, John Ferguson, Andrew Smyth, Michelle Canavan, Martin J. O’dOnnell

    Background and Purpose: Atrial fibrillation and heart failure with reduced ejection fraction (HFrEF) are common sources of cardioembolism. While oral anticoagulation is strongly recommended for atrial fibrillation, there are marked variations in guideline recommendations for HFrEF due to uncertainty about net clinical benefit. This systematic review and meta-analysis evaluates the comparative association of oral anticoagulation with stroke and other cardiovascular risk in populations with atrial fibrillation or HFrEF in sinus rhythm and identify factors mediating different estimates of net clinical benefit. Methods: PubMed and Embase were searched from database inception to November 20, 2019 for randomized clinical trials comparing oral anticoagulation to control. A random-effects meta-analysis was used to estimate a pooled treatment-effect overall and within atrial fibrillation and HFrEF trials. Differences in treatment effect were assessed by estimating I 2 among all trials and testing the between-trial-population P -interaction. The primary outcome measure was all stroke. Secondary outcome measures were ischemic stroke, hemorrhagic stroke, mortality, myocardial infarction, and major hemorrhage. Results: Twenty-one trials were eligible for inclusion, 15 (n=19 332) in atrial fibrillation (mean follow-up: 23.1 months), and 6 (n=9866) in HFrEF (mean follow-up: 23.9 months). There were differences in primary outcomes between trial populations, with all-cause mortality included for 95.2% of HFrEF trial population versus 0.38% for atrial fibrillation. Mortality was higher in controls groups of HFrEF populations (19.0% versus 9.6%) but rates of stroke lower (3.1% versus 7.0%) compared with atrial fibrillation. The association of oral anticoagulation with all stroke was consistent for atrial fibrillation (odds ratio, 0.51 [95% CI, 0.42–0.63]) and HFrEF (odds ratio, 0.61 [95% CI, 0.47–0.79]; I 2 =12.4%; P interaction=0.31). There were no statistically significant differences in the association of oral anticoagulation with cardiovascular events, mortality or bleeding between populations. Conclusions: The relative association of oral anticoagulation with stroke risk, and other cardiovascular outcomes, is similar for patients with atrial fibrillation and HFrEF. Differences in the primary outcomes employed by trials in HFrEF, compared with atrial fibrillation, may have contributed to differing conclusions of the relative efficacy of oral anticoagulation.