Ateneo - UPO

The course introduces and delves into the main biostatistical and epidemiological methods used in the analysis of biomedical, clinical, and population data, integrating theoretical aspects with practical applications using R software.

After an initial realignment on the fundamental concepts of descriptive and inferential statistics, the main epidemiological study designs, frequency and association measures and the statistical models most used in the biomedical field will be addressed, with particular reference to linear regression, logistic regression, Poisson models and survival analyses using Kaplan-Meier curves and Cox model.

Particular attention will be paid to the interpretation of multivariate models, confounding, effect modification, and criteria for choosing adjustment variables.

Part of the course will also be dedicated to the main statistical issues associated with omics and high-dimensional data, with particular reference to biological and technical variability, preprocessing, multiple comparisons, and False Discovery Rate control.

The theoretical concepts will be progressively applied in R through the analysis and interpretation of biomedical and epidemiological datasets.

The main material for exam preparation will consist of handouts, presentations, datasets, R scripts, and scientific articles made available by the instructor. Official documentation of R and the packages used during the course. Further bibliographical references and scientific articles may be indicated during the lessons.

Consultation texts:
W.W. Daniele, C.L. Cross — Biostatistics: A Basis for Analysis in the Health Sciences, 10th ed., Wiley
Lash TL, VanderWeele TJ, Haneuse S, Rothman KJ. Modern Epidemiology. 4th ed. Philadelphia: Wolters
Kluwer; 2021.

The course aims to provide students with the methodological foundations of biostatistics and epidemiology

necessary to understand, analyze and interpret biomedical, clinical and epidemiological data, with
particular attention to the correct application of quantitative methods in data research and analysis.

The main goal is to develop the ability to recognize the nature of the statistical problem,
choose the most appropriate method or model in relation to the scientific question, the design of the
study and to the nature of the outcome, apply it using the R software and critically interpret its
results.

The course will also provide the epidemiological tools needed to understand the main study designs
observational and experimental, including their characteristics, indications, strengths and limitations. They will be
addressed the concepts of population and sample, incidence and prevalence, measures of association and effect,
temporality of exposure and outcome, selection of comparison groups and main sources of error
systematic.

The course also aims to develop the capacity to recognise the main sources of bias and
confounding, understanding the meaning of multivariate adjustment, identifying variables
relevant to model building and distinguish between statistical association, prediction and
causal interpretation.

A specific part of the course is devoted to the main statistical problems specific to data
omics and high-dimensional, with the aim of providing the basic conceptual tools for their
analysis and interpretation.

Basic knowledge of biology, probability, and descriptive statistics is useful.

Given the heterogeneity of students' educational paths, the fundamental statistical and biological concepts necessary to follow the course will be recalled and consolidated in the first lessons.

No prior knowledge of epidemiology or advanced statistical analysis is required.

A basic understanding of the R language is useful but not indispensable. The commands and procedures necessary for carrying out the exercises will be introduced and used during the course

The course includes integrated theoretical-practical lessons.
Statistical and epidemiological concepts will be introduced through problems and examples from biomedical, clinical, and epidemiological research and will be immediately accompanied by their application using R software.

Activities will include:
lectures;
guided exercises in R;
analysis of biomedical and epidemiological datasets;
interpretation of statistical outputs, tables and graphs;
discussion of examples from the scientific literature.

The teaching approach will privilege the understanding, appropriate choice and interpretation of statistical methods over the mathematical derivation of formulas.

R scripts, datasets used during exercises, presentations, and any supplementary materials will be made available to students through the course's teaching platform.

For practical tasks, the availability of a computer with R and RStudio installed is recommended.

The use of R will primarily aim to apply and interpret the statistical methods presented during the course; prior programming knowledge is therefore not required.

Learning assessment aims to assess understanding of biostatistical and epidemiological principles and the ability to apply and interpret them in problems related to biomedical, clinical, and epidemiological data.

The test may include closed-ended and open-ended questions, simple application exercises, and questions based on the interpretation of datasets, tables, graphs, and statistical outputs, including those produced using R.

• recognize the scientific question, the study design and the nature of the data;
• distinguish major epidemiological study designs, including cross-sectional, case-control, cohort, and experimental studies, including their characteristics, strengths, and limitations;
• correctly interpret the main measures of frequency and epidemiological association;
• recognize major sources of bias and potential confounders;
• choose the appropriate statistical method or model based on the outcome, study design and data structure;
• distinguish between univariate and multivariate analyses and motivate the choice of variables to include in adjustment models;
• distinguish between raw and adjusted estimates and interpret coefficients, measures of association, confidence intervals and p-values;
• interpret the results of major regression models, including linear, logistic, Poisson, and Cox regression, and survival analyses using Kaplan-Meier curves;
• critically evaluate the results in the biomedical and epidemiological context, distinguishing between statistical association, prediction and causal interpretation;
• understand the key statistical issues associated with omics and high-dimensional data, including multiple comparisons and False Discovery Rate control

Storing the syntax of R commands will not be required. Instead, the ability to understand the purpose of the analyses performed, read the outputs correctly, choose the appropriate model, understand how to adjust it, and interpret the results will be evaluated.

The evaluation of the module contributes to the determination of the final grade of integrated teaching.

1. Fundamentals of Biostatistics and Introduction to Data Analysis with R
Population and sample. Statistical unit. Parameters and estimates. Types of variables. Outcomes, exposures, predictors and covariates. Distribution of data. Frequencies and proportions. Mean, median, quantiles, standard deviation, and interquartile range. Graphical representation of data. Biological and sample variability.
Introduction to the structure of biomedical datasets and the R environment. Importing, exploring, describing, and representing data in R.

2. Statistical inference and linear regression
Sample variability and standard error. Point estimate and confidence interval. Hypothesis testing logic. P-value. Errors of I and II type. Statistical significance and biological or clinical relevance.
Introduction to the linear regression model. Outcome continuous. Interpretation of coefficients. Continuous and categorical predictors. Reference category. Univariate and multivariate models.
Application and interpretation of linear models in R.

3. Epidemiology and logistic regression
Main epidemiological study designs: cross-sectional studies, case-control studies, cohort studies, and experimental studies.
Population at risk. Prevalence, incidence, risk and rate.
Measures of association and effect: difference in risk, relative risk and odds ratio.
Logistic regression for binary outcomes. Odds and odds ratio. Interpretation of coefficients and confidence intervals. Univariate and multivariate logistic models.
Applications in R to clinical and epidemiological data.

4. Confounding, effect modification and model adjustment
Concept of confounding. Distinction between exposure, predictor, confounder, mediator, and effect modifier.
Raw estimates and adjusted estimates. Introduction to interaction and effect modification.
Principles for choosing adjustment variables based on biological and clinical knowledge, scientific literature, and study design.
Limitations of variable selection approaches based solely on statistical significance.
Conceptual introduction to causal diagrams.
Comparison in R between raw and adjusted models and interpretation of changes in estimates.

5. Models for count data and rates
Count data, number of events and person-time. Incidence rates.
Introduction to the Poisson distribution and Poisson regression. Rate ratio. Using offset for rate analysis. Nods to the problem of overdispersion.
Application in R of Poisson models to epidemiological data and interpretation of estimates.

6. Survival analysis
Time-event data and censorship. Survival function. Kaplan-Meier curves. Comparison of survival curves.
Concept of hazard.
Cox regression model. Hazard ratio. Univariate and multivariate models. Interpretation of hazard ratio and confidence intervals. Concept of proportionality of hazards.
Construction and interpretation in R of Kaplan-Meier curves and Cox models.

7. Statistical principles for the analysis of omics data
Characteristics of omics and high-dimensional data. Biological and technical variability. Principles of data preprocessing and normalization.
Problems related to the large number of variables compared to the sample size.
Exploratory analysis of omics data.
Multiple comparisons and control of the False Discovery Rate.
General principles of biomarker identification and interpretation of high-dimensional analysis results.
Application using R of the main statistical concepts addressed in the course to omics datasets.

At the end of the course the student will be able to:

• recognize the structure and main characteristics of biomedical, clinical and epidemiological data;
• appropriately describe and represent quantitative and qualitative variables;
• understand and interpret point estimates, confidence intervals, and hypothesis tests;
• recognize key epidemiological study designs, including cross-sectional, case-control, cohort, and experimental studies, including their characteristics, strengths, and main limitations;
• interpret the main measures of frequency and epidemiological association;
• recognize the main sources of bias, confounding and effect modification;
• choose the appropriate statistical method or model in relation to the scientific question, study design, nature of the outcome and data structure;
• apply and interpret the main linear, logistic, Poisson and Cox regression models and Kaplan-Meier survival curves;
• distinguish between univariate and multivariate analyses and between crude and adjusted estimates, motivating the choice of the main adjustment variables;
• understand the conceptual distinction between statistical association, prediction, and causal interpretation;
• use R to perform simple statistical analyses on biomedical and epidemiological data and correctly interpret their outputs;
understand the key statistical challenges associated with omics and high-dimensional data, including multiple comparisons, False Discovery Rate, overfitting, and validation;
• critically interpret the results of the analyses in the context of the scientific question and the characteristics of the study design.