Catherine Kreatsoulas

A global perspective on psychosocial risk factors for cardiovascular disease.

Worldwide, there is variation in the incidence CVD with the greater burden being borne by low and middle-income countries. Traditional risk factors do not fully explain the CVD risk in populations, and there is increasing awareness of the impact the social environment and psychological factors have on CVD incidence and outcomes. The measurement of psychosocial variables is uniquely complex as variables are difficult to define objectively and local understanding of psychosocial risk factors may be subject to cultural influences. Notwithstanding this, there is a growing evidence base for the independent role they play in the pathogenesis of CVD. Consistent associations have been seen for general psychological stress, work-related stress, locus of control and depression with CVD risk. Despite the strength of this association the results from behavioural and pharmacological interventions have not clearly resulted in improved outcomes.

Reconstructing angina: cardiac symptoms are the same in women and men.

clude potassium content on the NFP. One might expect items with high levels of potassium to more often be labeled; however, categories where potassium information was most available had a wide range of potassium content, while some categories with expected high levels of potassium, such as canned whole tomatoes and canned beans, had low potassium information availability. The lack of potassium information on the NFP presents a problem for patients and consumers trying to make informed decisions when purchasing foods, particularly those motivated to minimize their risk of cardiovascular disease and those for whom potassium intake must be restricted. Lack of potassium information is also a problem for researchers and policy makers interested in understanding the overall and potassium-specific nutritional content of the packaged food supply. The New York City Department of Health and Mental Hygiene, along with 35 health authorities and health organizations, has called for a publicly accessible, productspecific nutrition database of packaged food products. The creation of such a database would allow for analyzing nutritional trends, which would inform recommendations to improve nutritional intake. While such a database is valuable with existing NFP information, its public health value would be enhanced if potassium information was consistently available. A limitation of this study is that the NSRI database was originally created to assess changes in sodium concentration; therefore, it includes a wide range of products, but only in food categories with significant potential for sodium reduction (eg, some desserts and beverages are not included). Also, only the top 80% of products sold within each category are represented in the analyses. The FDA is planning to revise the NFP, and potential improvements to its content and format are under consideration. The addition of potassium content and percent daily value to required NFP information could remedy the described deficit in publicly available nutrition information. Providing this important information to consumers, patients, and researchers would allow a more detailed understanding of the food supply, which would complement existing strategies to improve population nutritional intake.

Identifying women with severe angiographic coronary disease.

Abstract.  Kreatsoulas C, Natarajan MK, Khatun R, Velianou JL, Anand SS (McMaster University; CARING Network, McMaster University; Population Health Research Institute, McMaster University and Hamilton Health Sciences; Interventional Cardiology, Hamilton Health Sciences; Eli Lilly Canada–May Cohen Chair in Womens Health, McMaster University; Michael G. DeGroote‐Heart and Stroke Foundation of Ontario Chair in Population Health Research, McMaster University; Population Genomics Program, McMaster University; McMaster University, Hamilton, ON, Canada). Identifying women with severe angiographic coronary disease. J Intern Med 2010; 268:66–74.

Accuracy of Self-Reported Height and Weight to Determine Body MassIndex Among Youth

Background: Self-reported height and weight has important economic, clinical and research value however little is known on the accuracy of self-reporting BMI among youth. Our objective was to determine the accuracy of self-reported height and weight estimates compared to measured height and weight used to determine BMI, among youth. Methods: Youth ages 15-25 were recruited from primary care urban hospital clinical where a trained clinical assistant measured the participant’s height and weight. The youth were asked to self-report their height and weight as part of a larger computerized survey. Continuous variables were compared using t-tests, and dichotomous variables using chi-square tests. BMI correlation was determined using Pearson’s r and agreement using a weighted kappa test. Results: Among 355 youth, the mean measured BMI for men: 27.3+7.0 kg/m2 compared to women: 28.9+8.7 kg/m2 (p=0.08). 58% of youth had an above normal BMI. There was high correlation between measured and selfreported BMI when calculated using an adjusted r2=0.84 (p<0.01). Agreement was also high between BMI categories (weighted kappa=0.88, p<0.01). Conclusions: Youth can accurately self-report height and weight to derive meaningful BMI scores for BMI categorization during this important period of body transition in the life course cycle. BMI is often conceptualized as categories and the weighted kappa test is a sensitive test capturing ordinal levels of BMI categorical agreement.

Evaluation of the role of abciximab (Reopro) as a rescue agent during percutaneous coronary interventions : In-hospital and six-month outcomes

Abciximab is effective for the prevention of complications when administered prior to percutaneous coronary intervention (PCI). The efficacy and safety of abciximab as an unplanned or rescue agent for complications of PCI is unknown. Rescue versus planned use was compared in 186 consecutive patients. Primary or rescue PCI for acute myocardial infarction (MI) and shock were excluded. Rescue abciximab use was undertaken in 101 patients (54.3%) and planned abciximab was used in 85 (45.7%). The rescue abciximab patients had a lower incidence of previous MI, preprocedural thrombus, multivessel, and vein graft intervention. In‐hospital endpoints in the rescue versus planned abciximab patients were death (1.0% vs. 1.2%, P = 1.0), Q‐wave MI (2.0% vs. 2.4%, P = 1.0), any MI (14.9% vs. 9.4%, P = 0.3), target vessel revascularization (TVR; 0% vs. 1.2%, P = 1.0), and composite (15.8% vs. 10.6%, P = 0.3). At 6 months, events were death (4.0% vs. 2.3%, P = 0.69), MI (14.9% vs. 9.4%, P = 0.26), TVR (20.8% vs. 4.7%, P = 0.001), and composite (30.7% vs. 15.3%, P = 0.01). In‐hospital complications between the rescue and planned abciximab patients of major bleed (1.0% vs. 1.8%, P = NS), stroke (0% vs. 1.8%, P = NS), and thrombocytopenia (3.0% vs. 1.8%, P = NS) were similar. There was a significantly higher procedural time (99.6 min vs. 86.1 min, P = 0.02), contrast volume (278.8 ml vs. 223.5 ml, P = 0.04), and heparin use (8984 u vs. 6003 u, P = 0.0006) in the rescue group. In this nonrandomized comparison, rescue abciximab allowed for the safe discharge from hospital in the majority of patients. However, during a 6‐month follow‐up, more patients treated with rescue abciximab required TVR with either repeat PCI or CABG. Further studies are warranted to evaluate the overall strategy of rescue abciximab use in PCI. Cathet. Cardiovasc. Intervent. 51:138–144, 2000.

An emerging double burden of disease: the prevalence of individuals with cardiovascular disease and cancer.

Cardiovascular disease (CVD) and cancer are the two leading causes of death in the United States; at the same time, the number of survivors is increasing as therapies continue to improve. The primary objective of this study is to determine the prevalence and characteristics of individuals affected by both CVD and cancer.

Food Insecurity Screening in Pediatric Primary Care: Can Offering Referrals Help Identify Families in Need?

OBJECTIVE To describe a clinical approach for food insecurity screening incorporating a menu offering food-assistance referrals, and to examine relationships between food insecurity and referral selection. METHODS Caregivers of 3- to 10-year-old children presenting for well-child care completed a self-administered questionnaire on a laptop computer. Items included the US Household Food Security Survey Module: 6-Item Short Form (food insecurity screen) and a referral menu offering assistance with: 1) finding a food pantry, 2) getting hot meals, 3) applying for Supplemental Nutrition Assistance Program (SNAP), and 4) applying for Special Supplemental Nutrition Program for Women, Infants, and Children (WIC). Referrals were offered independent of food insecurity status or eligibility. We examined associations between food insecurity and referral selection using multiple logistic regression while adjusting for covariates. RESULTS A total of 340 caregivers participated; 106 (31.2%) reported food insecurity, and 107 (31.5%) selected one or more referrals. Forty-nine caregivers (14.4%) reported food insecurity but selected no referrals; 50 caregivers (14.7%) selected one or more referrals but did not report food insecurity; and 57 caregivers (16.8%) both reported food insecurity and selected one or more referrals. After adjustment, caregivers who selected one or more referrals had greater odds of food insecurity compared to caregivers who selected no referrals (adjusted odds ratio 4.0; 95% confidence interval 2.4-7.0). CONCLUSIONS In this sample, there was incomplete overlap between food insecurity and referral selection. Offering referrals may be a helpful adjunct to standard screening for eliciting family preferences and identifying unmet social needs.

Interpreting angina: symptoms along a gender continuum

Background ‘Typical’ angina is often used to describe symptoms common among men, while ‘atypical’ angina is used to describe symptoms common among women, despite a higher prevalence of angina among women. This discrepancy is a source of controversy in cardiac care among women. Objectives To redefine angina by (1) qualitatively comparing angina symptoms and experiences in women and men and (2) to propose a more meaningful construct of angina that integrates a more gender-centred approach. Methods Patients were recruited between July and December 2010 from a tertiary cardiac care centre and interviewed immediately prior to their first angiogram. Symptoms were explored through in-depth semi-structured interviews, transcribed verbatim and analysed concurrently using a modified grounded theory approach. Angiographically significant disease was assessed at ≥70% stenosis of a major epicardial vessel. Results Among 31 total patients, 13 men and 14 women had angiograpically significant CAD. Patients describe angina symptoms according to 6 symptomatic subthemes that array along a ‘gender continuum’. Gender-specific symptoms are anchored at each end of the continuum. At the centre of the continuum, are a remarkably large number of symptoms commonly expressed by both men and women. Conclusions The ‘gender continuum’ offers new insights into angina experiences of angiography candidates. Notably, there is more overlap of shared experiences between men and women than conventionally thought. The gender continuum can help researchers and clinicians contextualise patient symptom reports, avoiding the conventional ‘typical’ versus ‘atypical’ distinction that can misrepresent gendered angina experiences.

Design of the Magnetic Resonance Imaging Evaluation of Mineralocorticoid Receptor Antagonism in Diabetic Atherosclerosis (MAGMA) Trial

Mineralocorticoid receptor (MR) activation plays an essential role in promoting inflammation, fibrosis, and target organ damage. Currently, no studies are investigating MR antagonism in patients with type 2 diabetes mellitus (T2DM) with chronic kidney disease, at high risk for cardiovascular complications, who are otherwise not candidates for MR antagonism by virtue of heart failure. Further, there is limited information on candidate therapies that may demonstrate differential benefit from this therapy. We hypothesized that MR antagonism may provide additional protection from atherosclerosis progression in higher‐risk patients who otherwise may not be candidates for such a therapeutic approach. In this double‐blind, randomized, placebo‐controlled trial, subjects with T2DM with chronic kidney disease (≥ stage 3) will be randomized in a 1:1 manner to placebo or spironolactone (12.5 mg with eventual escalation to 25 mg daily over a 4‐week period). The co‐primary efficacy endpoint will be percentage change in total atheroma volume in thoracic aorta and left ventricular mass at 52 weeks in patients treated with spironolactone vs placebo. Secondary outcomes include 24‐hour mean systolic blood pressure, central aortic blood pressure, and insulin resistance (HOMA‐IR) at 6 weeks. A novel measure in the study will be changes in candidate miRNAs that regulate expression of NR3C2 (MR gene) as well as measuring monocyte/macrophage polarization in response to therapy with spironolactone. We envision that our strategy of simultaneously probing the effects of a drug combined with analysis of mechanisms of action and predictive response will likely provide key information with which to design event‐based trials.

Machine learning in social epidemiology: Learning from experience

Now over 60 years later, with many momentous accolades achieved in parallel with exponential advances in computing, applications of machine learning have infiltrated, improved and continue to augment many aspects of our daily lives. Today machine learning is a mainstay in business, finance, manufacturing, retail, science, technology, mobile computing, social media affecting our behaviours as consumers and creators of data, each interaction deepening our digital footprint. Medicine and disciplines related to health have become the new frontier for machine learning and big data. In particular, fields such as social epidemiology seem well suited to tap into the vast amounts of social data (Gruebner et al. 2017) including credit scores and social networks that could potentially shed some new insights to understanding health behaviours and how social determinants of health may operate. While successful examples of mainstream applications of machine learning offer much excitement for adaptation in the social sciences, we are at a critical moment in history where we can learn from successful machine learning applications, limitations and the potential dangers of mal-adapting these techniques. Machine learning is “a set of methods that can automatically detect patterns in data, and then use the uncovered patterns to predict future data, or to perform other kinds of decision-making under uncertainty”(Murphy, 2013). And while methods from machine learning are closely related to the type of statistics traditionally used in social health research, they differ in probabilistic inference and modeling. The paper by Seligman, Tuljapurkar and Rehkopf (2018) sought to compare four machine learning algorithms with a traditional regression to determine if 1) machine learning algorithms lead to better predictions and 2) do they enhance our understanding of how social determinants may result in differences in health outcomes. The authors conclude that traditional regression historically used in social health research faired well when compared to several machine learning methods; neural networks faired best due to their robust ability to allow for interactions and nonlinearity among input variables. However, the interpretation of neural networks is complicated, and the authors base their conclusions almost exclusively on the r square value obtained in cross-validation, a process in itself laden with inherent limitations. While the authors successfully compare results between the different methodologies, it is unclear how these methods enhance our understanding of health outcomes, particularly when the fundamental goal of machine learning is to generalize beyond the algorithm training set. Arguably, this may not necessarily be the fault with this study per se but rather a consequence of the infancy of these techniques in the social epidemiologic space. As quantitative social scientists process and often collate multiple sources of data, there are many alluring features from various techniques in machine learning that offer new methodologic ideas in how to handle and merge structured and unstructured datasets. A distinct advantage of machine learning methods includes the robust handling of large numbers of variables combined in interactive linear and nonlinear ways to detect patterns in the data for prediction. While there is a vast array of learning algorithms available, all machine learning algorithms consist of combinations of three key components: 1) representation of the input of data, where a classifier can learn in the hypothesis space, 2) evaluation of the classifiers, and lastly, 3) optimization, a search among classifiers to find the best performing one (Domingos, 2012). In supervised learning, the goal is prediction and includes techniques such as regression and classification or pattern recognition whereas in unsupervised learning, the goal is to find patterns in the data which is sometimes called knowledge discovery (Murphy, 2013). Reinforcement learning, while not as commonly used, is useful for learning how to act or behave when given occasional reward or punishment signals (Murphy, 2013). Table 1 outlines some of the strengths and limitations associated with this comparative study of machine learning methods used to evaluate health outcomes from the Health and Retirement Study dataset (Seligman et al., 2018). While each type of machine learning offers distinct advantages and