Biostatistics 2 - Correlation and Risk

Similar documents
Biostatistics 3. Developed by Pfizer. March 2018

STATISTICS INFORMED DECISIONS USING DATA

CHAPTER ONE CORRELATION

Critical Appraisal of Evidence

How to assess the strength of relationships

Critical Appraisal of Evidence A Focus on Intervention/Treatment Studies

Results & Statistics: Description and Correlation. I. Scales of Measurement A Review

Section 3 Correlation and Regression - Teachers Notes

Further Mathematics 2018 CORE: Data analysis Chapter 3 Investigating associations between two variables

Construct Reliability and Validity Update Report

Providing High Value Cost-Conscious Care:

Section 3.2 Least-Squares Regression

Statistical Methods and Reasoning for the Clinical Sciences

Consider the following hypothetical

Undertaking statistical analysis of

1.4 - Linear Regression and MS Excel

STAT 405 BIOSTATISTICS (Fall 2016) Handout 9 Number Needed to Treat, Number Needed to Harm, and Attributable Risk

Making comparisons. Previous sessions looked at how to describe a single group of subjects However, we are often interested in comparing two groups

Day 11: Measures of Association and ANOVA

11/18/2013. Correlational Research. Correlational Designs. Why Use a Correlational Design? CORRELATIONAL RESEARCH STUDIES

W e have previously described the disease impact

Welcome to this third module in a three-part series focused on epidemiologic measures of association and impact.

7. Bivariate Graphing

Hypertension Management Controversies in the Elderly Patient

Appendix B Statistical Methods

Why nursing students should understand statistics. Objectives of lecture. Why Statistics? Not to put students off statistics!

Glossary of Practical Epidemiology Concepts

PPH Domain: Learning Objectives Synopsis: probability frequency Prevalence Incidence Relative risk - Question 1

Quantifying the Benefits and Harms of Interventions: Relative vs. Absolute Risk

Live WebEx meeting agenda

CHAPTER 3 DATA ANALYSIS: DESCRIBING DATA

IAPT: Regression. Regression analyses

3 CONCEPTUAL FOUNDATIONS OF STATISTICS

Results. Example 1: Table 2.1 The Effect of Additives on Daphnia Heart Rate. Time (min)

Clinical Epidemiology for the uninitiated

Chapter 7: Descriptive Statistics

How to describe bivariate data

Do the sample size assumptions for a trial. addressing the following question: Among couples with unexplained infertility does

Quick start guide for using subscale reports produced by Integrity

Correlational Research. Correlational Research. Stephen E. Brock, Ph.D., NCSP EDS 250. Descriptive Research 1. Correlational Research: Scatter Plots

MEASURING THE SIZE OF A TREATMENT EFFECT: RELATIVE RISK REDUCTION, ABSOLUTE RISK REDUCTION, AND NUMBER NEEDED TO TREAT

Statistical reports Regression, 2010

Overview. Goals of Interpretation. Methodology. Reasons to Read and Evaluate

This article is the second in a series in which I

Interpretation of Data and Statistical Fallacies

TREATING THE MANY TO BENEFIT THE FEW : USING EXAMPLES FROM HEALTHCARE TO MAKE STATISTICS EXCITING AND RELEVANT

ADMS Sampling Technique and Survey Studies

Psychology Research Process

Essential Skills for Evidence-based Practice: Statistics for Therapy Questions

EXECUTIVE SUMMARY DATA AND PROBLEM

Study Guide #2: MULTIPLE REGRESSION in education

Evidence Based Medicine

Confidence Intervals On Subsets May Be Misleading

Epidemiologic Study Designs. (RCTs)

GLOSSARY OF GENERAL TERMS

Market Research on Caffeinated Products

BIOSTATISTICAL METHODS AND RESEARCH DESIGNS. Xihong Lin Department of Biostatistics, University of Michigan, Ann Arbor, MI, USA

Name Psychophysical Methods Laboratory

Research Methods in Forest Sciences: Learning Diary. Yoko Lu December Research process

Biases in clinical research. Seungho Ryu, MD, PhD Kanguk Samsung Hospital, Sungkyunkwan University

Examining the Psychometric Properties of The McQuaig Occupational Test

Clincial Biostatistics. Regression

Find the slope of the line that goes through the given points. 1) (-9, -68) and (8, 51) 1)

Still important ideas

Measures of association: comparing disease frequencies. Outline. Repetition measures of disease occurrence. Gustaf Edgren, PhD Karolinska Institutet

LEGAL NOTICE. This plan MAY NOT be reproduced in anyway, nor copyright claimed for any part or in whole of the plan or contents.

Chapter 14: More Powerful Statistical Methods

4 Diagnostic Tests and Measures of Agreement

An Introduction to Epidemiology

Readings: Textbook readings: OpenStax - Chapters 1 13 (emphasis on Chapter 12) Online readings: Appendix D, E & F

The Fallacy of Taking Random Supplements


Patient decision aid: Type 2 diabetes blood pressure control

CEP Example Outline, as submitted by the author USE CHEMICAL ENGINEERING TO REVAMP YOURSELF!

Psychology Research Process

Reminders/Comments. Thanks for the quick feedback I ll try to put HW up on Saturday and I ll you

11-3. Learning Objectives

Research Methods 1 Handouts, Graham Hole,COGS - version 1.0, September 2000: Page 1:

Global Clinical Trials Innovation Summit Berlin October 2016

Describe what is meant by a placebo Contrast the double-blind procedure with the single-blind procedure Review the structure for organizing a memo

(CORRELATIONAL DESIGN AND COMPARATIVE DESIGN)

Two-Way Independent ANOVA

AmericanJournal ofpublic Health

Observational Studies and Experiments. Observational Studies

Firstly, there are no magic ratios, techniques or programmes that will create the Holy Grail macronutrient split. So stop looking.

Index. Springer International Publishing Switzerland 2017 T.J. Cleophas, A.H. Zwinderman, Modern Meta-Analysis, DOI /

Still important ideas

STP 231 Example FINAL

Describe what is meant by a placebo Contrast the double-blind procedure with the single-blind procedure Review the structure for organizing a memo

Probability and Statistics. Chapter 1

Intro to SPSS. Using SPSS through WebFAS

Business Statistics Probability

EXERCISE: HOW TO DO POWER CALCULATIONS IN OPTIMAL DESIGN SOFTWARE

Correlated to: ACT College Readiness Standards Science (High School)

2 Assumptions of simple linear regression

Review Statistics review 11: Assessing risk Viv Bewick 1, Liz Cheek 1 and Jonathan Ball 2

MS&E 226: Small Data

Lecture Week 3 Quality of Measurement Instruments; Introduction SPSS

the standard deviation (SD) is a measure of how much dispersion exists from the mean SD = square root (variance)

The North Carolina Health Data Explorer

Transcription:

BROUGHT TO YOU BY Biostatistics 2 - Correlation and Risk Developed by Pfizer January 2018 This learning module is intended for UK healthcare professionals only. PP-GEP-GBR-0957 Date of preparation Jan 2018.

1 Introduction This learning module is intended for UK healthcare professionals only. Medical knowledge is constantly changing. As new information becomes available, changes in treatment, procedures, equipment and the use of drugs become necessary. The authors and editors have, as far as it is possible, taken care to ensure that the information given in this module is accurate and up to date at the time it was created. However, users are strongly advised to confirm that the information, especially with regard to drug usage, complies with current legislation and standards of practice.this learning module is intended for UK healthcare professionals only.

Introduction This course is designed to allow you to implement and understand the statistics used in many aspects of biology, from clinical trials to non-investigator lead research and even the more recent phenomenon classified as real life data. This course is split into three modules: Biostatistics 1, 2 and 3. To allow all readers to have a sound understanding we will start from the very basics, but feel free to skip the chapters if you feel they are unneeded or browse through for a refresh.

Learning objectives The module presents the second of three courses. It will help you to understand how to assess relationship between two variables and how to interpret risk measures.

2 Correlation This learning module is intended for UK healthcare professionals only.

Drug Serum Level (mcg/ml) Drug Serum Level (mcg/ml) Correlation The relationship that occurs between two variables when a change in one variable is associated with a change in the other variable Variable 1 Variable 1 Variable 2 Variable 2 100 60 0 20 40 60 80 100 120 Creatinine Clearance (ml/min) A positive correlation is the extent the variable increase or decrease in parallel A negative correlation indicates the extent to which one variable increases as the other decreases

Scatter Plot Scatter plots visually represent correlation between two variables Data plotted on a scatter plot give a visual representation of how strongly or weakly correlated one variable is to another. The more closely the data points line up, the stronger the correlation, and vice versa

Pearson Coefficient Pearson Coefficient (or Pearson Product- Moment Correlation) is indicated by r The strength of the correlation increased towards -1 or 1 r value range -1 0 +1 r = -1 r =.7 Perfect Correlation r = 0 Graph B r = -.7 Graph C r = 1 Perfect Correlation Graph D Graph E

Renal Function (ml/min) Pearson Coefficient Positive or Direct Correlation Two variables move in the same direction Negative or Indirect Correlation Two variables move in opposite directions 100 80 60 40 20 40 Negative Correlation r = -0.7 Age (years) 60 80 100

Pearson Coefficient No Relationship r = 0 indicates that there is no relationship between the variables

Pearson Coefficient r Value Chart rr General Interpretation 0.8 to 1 Very strong relationship 0.6 to 0.8 Strong relationship 0.4 to 0.6 Moderate relationship 0.2 to 0.4 Weak relationship 0.0 to 0.2 Weak or no relationship This table can be used as an easy way to interpret the value of r For example, if the correlation between two variables is 0.4, you can safely conclude that the relationship is weak to moderate However this is subjective thus, using this rule-of-thumb table is great for a quick assessment of r, but we will go onto explain a more precise method

Interpreting Correlation Coefficient of Determination: r 2 The percentage of variance in one variable that is accounted for by the variance in the other variable For example; r = 0.7 r 2 = 0.49 49% of the variance in blood pressure can be explained by the variance in dietary salt intake

Interpreting Correlation Coefficient of Determination: r 2 Even with a strong correlation of 0.7, the majority (51%) of the variance goes unexplained 51% Unexplained Variance 19% Unexplained Variance r = 0 r 2 = 0 r = +/- 0.7 r 2 = 0.49 r = +/- 0.9 r 2 =.81 Here, with r 2 equal to 0, the two variables have nothing in common and are totally unrelated When r 2 equals 0.49, both variables share 49% of the variance between themselves It follows that when r 2 equals 0.81, 81% of the variance is shared and relatively little variance (19%) goes unexplained

Blood Pressure (mmhg) Linear Regression A line of best fit is applied to the data points which measures the effect of a single independent variable Once the best fit line has been calculated, the regression analysis assesses how well the individual data points adhere to the line 180/120 160/110 140/100 130/90 Correlation r =.7 This allows the researcher to predict the change that will occur in one variable based on a change in a second variable 120/80 0 2 4 6 8 10 Salt Intake (grams)

Blood Pressure (mmhg) Linear Regression In the graph shown, you can predict that a dietary salt intake of 8g per day will correlate to a blood pressure reading of 160/110 mm Hg. The closer the fit of the data to the regression line, the higher the correlation coefficient, and the more accurate the prediction will be 180/120 160/110 140/100 Correlation r =.7 You have probably already surmised that if the correlation coefficient is +1 or -1, the prediction will be perfect, that is, from the value of the first variable you can predict the exact value of the second variable 130/90 120/80 0 2 4 6 8 10 Salt Intake (grams)

3 Risk This learning module is intended for UK healthcare professionals only.

Understanding Risk There are manly different types of risk described in statistic (listed below), throughout this chapter we will describe what each type of risk means and how it can be used Risk Terminology Absolute Risk (AR) Absolute Risk Reduction (ARR) Relative Risk (RR) Relative Risk Reduction (RRR) Odds Ratio Hazard Ratio

Absolute Risk Absolute Risk (AR) = Incidence or Occurrence Rate Absolute risk is another way of expressing the incidence, or occurrence rate, of an event. Drug A Drug B (Control) Example: in a clinical study, 6% of patients experienced nausea with Drug A. In the control group, 9% of patients experienced nausea on Drug B. We can now say that the absolute risk of nausea is 6% and 9% with Drug A and Drug B, respectively 6% AR = 6% 9% AR = 9%

Absolute Risk Reduction Absolute Risk Reduction (ARR) = Absolute Difference ARR = 3% (AR Drug B AR Drug A) Drug A Drug B (Control) The absolute difference, or ARR, with Drug A is 3%, simply by subtracting the proportion (or percentage) of patients 6% who experienced nausea with Drug A 9% from the proportion (or percentage) of AR = 6% patients who experienced nausea with Drug B 9% 6% = 3% ARR = 3% AR = 9%

Number Needed to Treat (NNT) Drug A Drug B (Control) The numbers needed to treat helps to establish clinical significance of the ARR. NNT = 1/ARR thus for the previous example 1/0.03 = 33 patients 6% 9% AR = 6% AR = 9% NNT = 1/.03 = 33 patients

Number Needed to Treat (NNT) Absolute difference (ARR) of 3% may, or may not be statistically significant, but we also want to determine the possibly clinically significance. To help to attribute a clinical aspect to the data, the numbers needed to treat (NNT) is calculated, the number of patients that the physician needs to treat so that one patient is likely to experience a benefit, in our case, no nausea. The NNT is calculated by using the formula 1/ARR. In our case NNT would be 1/0.03, or 33 patients. Thus, the physician would have to make a clinical judgment as to whether or not it is beneficial to treat 33 patients with Drug A so that one patient can avoid nausea. 6% Drug A Drug B (Control) 9% AR = 6% AR = 9% NNT = 1/0.03 = 33 patients

Number Needed to Harm (NNH) Although we can establish the benefit we also want to look at the possibility of harm. Instead of absolute risk reduction, we state the results in terms of absolute risk increase, or ARI In turn, number needed to treat becomes number needed to harm, or NNH Absolute Risk Increase (ARI) Number Needed to Harm (NNH) NNH = 1/ARI Drug A 6% AR = 6% ARI = 3% or.03 Drug B (Control) 9% AR = 9% NNH = 1/.03 = 33 patients

Absolute Risk Increase (ARI) Number Needed to Harm (NNH) NNH = 1/ARI Drug A Drug B (Control) Let s go back to our example: suppose that the treatment group (Drug A) was associated with an absolute risk of nausea of 9% and the control group (Drug B) was associated with a nausea risk of only 6%. In other words, we have reversed the incidence rates, or absolute risk. Using a similar formula to NNT, NNH becomes 1/ARI or again, 1/.03 or 33 patients. 6% AR = 6% ARI = 3% or.03 9% AR = 9% NNH = 1/.03 = 33 patients In this case, the physician would have to treat 33 patients with Drug A for one more patient to experience nausea than if they were taking Drug B

Relative Risk (RR) Incidence in Drug A/Incidence in Drug B Another way to express risk is as relative risk, that is, the risk of patients receiving Drug A developing nausea, relative to that among patients in the control group receiving Drug B Drug A Drug B (Control) Example: the risk of nausea with Drug A is 6% compared to that of Drug B, 9%. Mathematically, this would be expressed as the incidence of nausea in the treatment group divided by the incidence of nausea in the control group, or 0.06/0.09, which equals 0.67, or 67%. Thus, we can say that the relative risk of nausea with Drug A is 0.67 or 67% of that seen with Drug B 6% 9% AR = 6% AR = 9% RR = 0.06/0.09 = 0.67 or 67%

Hazard Ratio (HR) Another common term you are likely to see in the literature is hazard ratio Hazard ratio is a term used in the context of survival over time and basically indicates the increased speed in which one group is likely to experience an event, such as death So if the hazard ratio is 0.5, then the relative risk of dying in one group is half that of the second group Relative risk of death: 1/15 Relative risk of death: 3/15 Hazard ratio of coal miners: 3

Hazard Ratio (HR) A survival analysis produces a hazard ratio rather than an odds ratio or relative risk The hazard ratio is roughly equivalent to the relative risk of death For instance, if a risk factor, such as working in a coal mine, is associated with a fatal form of lung cancer and has a hazard ratio of 3, then we can say that a coal miner is likely to die of lung cancer at three times the rate of those who are not coal miners. Relative risk of death: 1/15 Relative risk of death: 3/15 Hazard ratio of coal miners: 3

Understanding Risk If either RR or OR is 1, there is no increase in risk, because the numerator and denominator are the same No Increased in risk 2-Pack Smokers 1-Pack Smokers 3% 3% OR =.03/.03 = 1

RRR and ARR Relative Risk Reduction (RRR) vs. Absolute Risk Reduction (ARR) : Differences It is important to appreciate and understand the distinct differences between relative risk reduction and absolute risk reduction More often than not, the medical literature will report the relative risk reduction rather than the absolute risk reduction. The reason, quite simply, is that RRR is often more impressive than ARR and, thus, reinforces and, in some cases magnifies the true differences between study groups Although RRR is more impressive, it s the ARR that is more clinically useful, as illustrated earlier in our discussion of NNT and NNH RRR More often reported Often more impressive Less clinically useful ARR Less often reported Often less impressive More clinically useful

RRR and ARR Example: ARR (Drug A) = 3%, RRR (Drug A) = 33%, Absolute Difference (ARR) =3%, NNT= 33 patients The absolute risk reduction in nausea with Drug A was 3% compared to Drug B, although the relative risk reduction was 33% If the authors of the study were to report a relative risk reduction of 33% in nausea with the new treatment, Drug A, one would be inclined to immediately stop using the control drug B and begin using Drug A, all other things being equal However, a closer look at the data revealed an absolute difference of only 3%. And again, the physician would have to treat 33 patients with the new drug in order for one less patient to experience nausea vs Drug B probably not clinically worthwhile, especially if there was a difference in cost (more expensive) or dosing frequency, and so on RRR More often reported Often more impressive Less clinically useful ARR Less often reported Often less impressive More clinically useful

4 Testing your knowledge This learning module is intended for UK healthcare professionals only.

Question 1 Looking at the diagram, what type of correlation is this an example of? A) Strong B) Weak

Question 2 Absolute Risk Reduction (ARR) = Absolute Difference True False

Question 3 Select three characteristics of relative risk reduction (RRR) in comparison to absolute risk reduction (ARR) Often reported less Often reported more Less clinically useful More clinically useful Often more impressive Often less impressive

Registration Sign up Thank you for completing this module. Please click here to sign up for updates and new modules. 3 4

Contact Us For general inquiries or information about Pfizer medicines, you can contact Pfizer on 01304 616161. This learning module is intended for UK healthcare professionals only. PP-GEP-GBR-0957 Date of preparation Jan 2018.