Med 106 Research Methods and Medical Statistics (2024) - PDF

MED 106 Research Methods and Medical Statistics Introduction to Confounding I: The multifactorial nature of disease and the concept of the confounder Dr Constantinos Koshiaris Assistant Professor in Medical Statistics and Epidemiology Learning Objectives Recognize the multifactorial nature of health...

MED 106 Research Methods and Medical Statistics Introduction to Confounding I: The multifactorial nature of disease and the concept of the confounder Dr Constantinos Koshiaris Assistant Professor in Medical Statistics and Epidemiology Learning Objectives Recognize the multifactorial nature of health and disease Describe the concept of the confounder Describe the presence of potential confounding in different research scenarios Multifactorial nature of disease Multifactorial nature of disease Most diseases are caused by high exposure to certain risk factors and/or low exposure to certain protective factors. These factors are collectively called disease determinants and differ from disease to disease. However, many diseases share common determinants. This applies particularly for chronic diseases, such as cardiovascular and cerebrovascular diseases, type 2 diabetes, cancer, and neurodegenerative diseases Multifactorial nature of disease Disease determinants may include: Genetic factors Biological/physiological factors (e.g. ageing, other disease status, biochemical factors) Lifestyle factors (e.g. diet, physical activity, smoking, alcohol consumption, sleeping patterns) Psychosocial factors (e.g. stress, social isolation) Sociodemographic (e.g. socioeconomic position, educational attainment, urbanization) Wider environmental factors (e.g. air pollution, impure water, toxic agents, radiation) Disease determinants: Cardiovascular disease high blood pressure ageing unhealthy diet hypercholesterolemia smoking Cardiovascular disease hyperglycemia low physical activity high alcohol consumption obesity genetic predisposition Disease determinants: Type 2 diabetes high blood pressure Ageing unhealthy diet Smoking Hypercholesterolemia Type 2 diabetes low physical activity high alcohol consumption obesity genetic predisposition Disease determinants: Alzheimer’s Disease high blood pressure ageing social isolation low education hypercholesterolemia smoking? Alzheimer’s Disease hyperglycemia low physical activity low cognitive activity obesity genetic predisposition Disease determinants: Colorectal cancer Obesity Age Alcohol Smoking Diabetes Processed meat Family history of colorectal cancer Colorectal cancer Inflammatory bowel disease History of colorectal polyps Introduction to confounding Introduction to Confounding In the previous sessions we have seen how chance (random error) and bias (systematic error) can introduce error in research findings (estimates). Even in the absence of chance (i.e. large study) and bias (i.e. random sampling and accurate assessment of exposure and outcome), we still have to deal with a third parameter which may introduce error. This parameter is called confounding and its role is inherently linked with the multifactorial nature of disease. Exposure-Outcome association Exposure Outcome Exposure-Outcome association Rate of ice cream consumption Number of sunburns Exposure-Outcome association Hot temperature Rate of ice cream consumption Number of sunburns Exposure-Outcome association dietary intake smoking physical activity obesity alcohol intake type 2 diabetes Exposure-Outcome association age body weight alcohol consumption other diseases Factor X Exposure meat intake fruit and veg intake fish intake Outcome colorectal cancer type 2 diabetes CVD Confounding Confounding: a third factor (confounder) explains all or part of the association between an exposure and an outcome Implications of confounding: 1. In general, it distorts the relationship between an exposure and an outcome 2. It can “create” associations 3. It can “mask” associations 4. It can overestimate/underestimate associations Confounding A potential confounder needs to fulfil the following criteria: 1. Has to be associated with the outcome of interest 2. Has to be associated with the exposure of interest 3. Should not lie in the causal pathway between exposure and outcome Confounding (Example 1) Confounding: a third factor (confounder) explains all or part of the association between an exposure and an outcome smoking alcohol consumption lung cancer Deciding on whether a factor is a potential confounder (Example) A study investigated the association between alcohol consumption and lung cancer and found that the Odds Ratio for lung cancer was 2.50 (95% CI: 2.30 - 2.70) comparing those with high alcohol consumption to those with low alcohol consumption The results are displayed in the next slide Deciding on whether a factor is a potential confounder (Example) Lung cancer Alcohol consumption Yes Heavy 535 No Odds in the exposed 2721 Odds in the non-exposed Odds Ratio No or low 1182 15562 535/2721 = 0.20 1182/15562 = 0.08 0.20/0.80 = 2.50 Deciding on whether a factor is a potential confounder (Example) The authors speculated that smoking could be a potential confounder in the association between alcohol consumption and lung cancer What steps should we take in order to decide whether smoking is a potential confounder?? In order to decide this we have to answer the 3 questions …. Deciding on whether a factor is a potential confounder (Example) 1. Is there an association between the potential confounder (smoking) and the outcome of interest (lung cancer)? 2. Is there an association between the potential confounder (smoking) and the exposure of interest (alcohol consumption)? 3. Does the potential confounder (smoking) lie in the causal pathway between exposure (alcohol consumption) and outcome (lung cancer)? Deciding on whether a factor is a potential confounder (Example) 1. Is there an association between the potential confounder (smoking) and the outcome of interest (lung cancer)? Odds Ratio (95% CIs) for the association between smoking and lung cancer = 6.50 (6.30 - 6.70) YES! Deciding on whether a factor is a potential confounder (Example) 2. Is there an association between the potential confounder (smoking) and the exposure of interest (alcohol consumption)? Odds Ratio (95% CIs) for the association between smoking and high alcohol consumption = 3.10 (2.95 - 3.25) YES! Deciding on whether a factor is a potential confounder (Example) 3. Does the potential confounder (smoking) lie in the causal pathway between exposure (alcohol consumption) and outcome (lung cancer)? alcohol consumption NO! smoking lung cancer Deciding on whether a factor is a potential confounder (Example) When the answer in the following 3 questions is: YES 1. Is there an association between the potential confounder (smoking) and the outcome of interest (lung cancer)? YES 2. Is there an association between the potential confounder (smoking) and the exposure of interest (alcohol consumption)? NO 3. Does the potential confounder (smoking) lie in the causal pathway between exposure (alcohol consumption) and outcome (lung cancer)? we can conclude that we have a potential confounder and we proceed in investigating the extent of confounding Deciding on whether a factor is a potential confounder (Example) Note 1: If a factor is associated with the outcome but not with the exposure (or vice versa), then this factor cannot be a confounder in the specific association! Note 2: If a factor is associated with both the outcome and the exposure, but lies in the causal pathway between the two, then this factor is not called a confounder but a mediator! Some examples of mediation (not confounding!) salt intake blood pressure stroke physical activity insulin sensitivity type 2 diabetes obesity blood cholesterol CVD Further Examples on deciding about potential confounding.. A team of researchers investigated the association between diet and colorectal cancer and found that individuals with an unhealthy dietary pattern had 85% higher risk of developing the disease compared to those with a healthy dietary pattern (OR: 1.85; 95% CI: 1.40 - 2.30) Potential confounders? physical activity alcohol consumption smoking age etc.. Further Examples on deciding about potential confounding.. A team of researchers investigated the association between smoking and Alzheimer’s Disease and found that smokers had 20% lower risk of developing the disease compared to non-smokers (OR: 0.80; 95%CI: 0.64 0.96) Potential confounders? age educational attainment social isolation physical activity etc. Learning Objectives Recognize the multifactorial nature of health and disease Describe the concept of the confounder Describe the presence of potential confounding in different research scenarios Further reading (optional) Petrie A. & Sabin C. Medical Statistics at a Glance, 3rd Edition, Chapter 34 [ISBN : 978-1-40518051-1] Buring EJ. Epidemiology in Medicine, Chapters 3, 9 [ISBN : 978- 0316356367] http://www.ncbi.nlm.nih.gov/pubmed/11274518 http://www.healthknowledge.org.uk/node/803