Table of Contents

Cover

Introduction

Part 1: Multiple-choice questions

Part 2: Structured Questions

Part 3: Critical appraisal Randomised controlled trial: template

Randomised controlled trial: Paper 1
Observational study: template
Observational study: Paper 2

Part 4: Data analysis

Dataset 1 analysed by Stata v11 (StataCorp LP, Texas, USA)
D1.1 Introduction
D1.2 Aims
D1.3 Repeatability
D1.4 Getting a feel for the data
D1.5 Baseline data: comparability of bracket groups
D1.6 Maximum pain intensity analysis
D1.7 AUC of pain intensity analysis
D1.8 Random effects analysis of longitudinal data
D1.9 Summary
D1.10 Note
Dataset 2 analysed using IBM SPSS Statistics v20
D2.1 Introduction
D2.2 Aims
D2.3 Relationship between UPSIT-40 and SS-16
D2.4 Univariable analyses
D2.5 Logistic regression analyses
D2.6 Using the receiver operator characteristic (ROC) curves
D2.7 Conclusion

Part 5: Solutions

Solutions to multiple-choice questions
Model answers for structured questions
Randomised controlled trial: critical appraisal of Paper 1
Observational study: critical appraisal of Paper 2

Appendices: Appendix I: list of multiple-choice questions with relevant chapter numbers from Medical Statistics at a Glance (3rd edn) and associated topics

Handling data
Sampling and estimation
Study design
Hypothesis testing
Basic techniques for analysing data
Additional techniques

Appendix II: list of structured questions with relevant chapter numbers from Medical Statistics at a Glance (3rd edn) and associated topics

Appendix III: chapter numbers from Medical Statistics at a Glance (3rd edn) with relevant multiple-choice questions and structured questions

End User License Agreement

List of Tables

Part 1: Multiple-choice questions

Table M62.1 Summary data for patients in observational study

Part 2: Structured Questions

Table S1.1 Portion of data set on cardiovascular disease risk
Table S3.1 Characteristics of subjects; data shown are mean (standard deviation (SD)) unless otherwise stated. Copyright © 2010, Serdi and Springer Verlag France.
Table S4.1 Summary of main characteristics of the over-65 and under-65 groups. Reproduced with permission of Elsevier.
Table S7.1 Vitreal glutamate concentrations (µmol/L)
Table S7.2 Computer output from the analysis of the vitreal glutamate concentration data in rhesus monkeys
Table S8.1 Baseline characteristics that were significantly different between the ‘neuropathic’ and ‘non-neuropathic’ PDQ groups. Reproduced with permission of Wolters Kluwer Health.
Table S10.1 The shear bond strength (megapascals) of orthodontic brackets for different drinks
Table S10.2 One-way ANOVA table for the shear bond strength data
Table S10.3 Scheffé’s test results for the shear bond strength data (I and J are labels to identify the groups that are being compared)
Table S12.1 Improvement in preparedness over time: percentage response to the statement. © 2007 Cave, et al., licensee BioMed Central Ltd.
Table S12.2 Changes in preparedness when new courses were introduced: proportion (%) strongly agreeing or agreeing that their medical school prepared them well. © 2007 Cave, et al., licensee BioMed Central Ltd.
Table S14.1 Estimated partial regression coefficients from a multiple regression analysis investigating the relationship between upper airway length and the Respiratory Disturbance Index. Reproduced with permission of Elsevier.
Table S15.1 Estimated odds ratios derived from a linear logistic regression analysis to investigate the relationship between CKD Stage III and potential risk factors
Table S16.1 Rate of first fracture (any or hip) by quintiles of average cumulative intake of dietary calcium in the cohort, as well as mean (standard deviation (SD)) age at entry in the cohort. Reproduced with permission of BMJ Publishing Group Ltd.
Table S17.1 Number of falls and rides, incidence rate ratios (IRRs) for steeplechase racing. Reproduced with permission of Elsevier.
Table S18.1 Adjusted in-hospital mortality risk ratios (RR) and confidence intervals (CI) after congenital heart surgery, 2004–2008. Reproduced with permission of Elsevier.
Table S20.1 Results of random effects models that assess the association between Body Mass Index (BMI) and duration of sleep at ages 3, 4 and 5 years. Reproduced with permission of BMJ Publishing Group Ltd.
Table S24.1 AUROC values and measures of performance for the five scoring systems, for assessing 28-, 84- and 180-day mortality. Reproduced with permission of Informa.
Table S25.1 Number (%) of patients with Classes 1, 2 and 3 liver failure, classified according to FIB-4 and APRI scores. Reproduced with permission of Oxford University Press.
Table S28.1 Postoperative complications by valve location and valve type (values shown are n (%) unless otherwise stated). Reproduced with permission of Elsevier.
Table S29.1 30-day, 90-day and 1-year survival rates (%) among individuals undergoing heart and lung transplants

Part 4: Data analysis

Table D2.1 Individual items (shown in the order in which they were presented) in the UPSIT-40 and SS-16 smell tests
Table D2.2 Contingency table showing the number of subjects with and without PD according to the true diagnosis and the decision rule SS ≤ 10
Table D2.3 Sensitivity, specificity and likelihood ratio for PD for different SS score thresholds with PD indicated if SS ≤ cut-off

Part 5: Solutions

Table S1.2 Summary of the type of variable, summary statistics, error checks and graphical measures that would be used to display each variable
Table S8.2 Observed (expected) frequencies in each cell. Reproduced with permission of Wolters Kluwer Health.
Table S9.1 Summary statistics for BIC, expressed as a percentage. Reproduced with permission of John Wiley & Sons.
Table S11.1 Contingency table for association between 30-day outcome and endovascular repair (EVAR) versus open ruptured abdominal aortic aneurysm (rAAA) repair. Reproduced with permission of Elsevier.
Table S12.3 Observed number of doctors who did (or did not) agree or strongly agree to the statement. © 2007 Cave, et al., licensee BioMed Central Ltd.
Table S17.2 Number of falls and rides, fall rate and IRRs for jump type and race distance. Reproduced with permission of Elsevier.
Table S21.1 Contingency table showing observed and expected frequencies of veterans requiring sedation and veterans requiring no sedation in the air and water groups
Table S23.1 Contingency table showing numbers of mild Alzheimer’s patients and healthy controls with a TYM ≤ or > 42

List of Illustrations

Part 2: Structured Questions

Figure S5.1 Flow of children through trial from initial assessment of eligibility to evaluation of primary endpoint.
Figure S26.1 Bland and Altman diagram assessing the repeatability of ambulatory monitoring in patients with multiple sclerosis. Reproduced with permission of Elsevier.
Figure S27.1 Forest plot from a meta-analysis of studies assessing the effect of periodontal treatment during pregnancy on preterm birth. Reproduced with permission of BMJ Publishing Group Ltd.
Figure S27.2 Funnel plot from a meta-analysis of studies assessing the effect of periodontal treatment during pregnancy on preterm birth. Reproduced with permission of BMJ Publishing Group Ltd.
Figure S28.1 Kaplan–Meier survival estimates stratified by type of valve received (biological or mechanical). (a) Isolated aortic valve replacements and (b) isolated mitral valve replacements. Reproduced with permission of Elsevier.

Part 4: Data analysis

Figure D1.1 Bland and Altman diagram for Little’s Total Irregularity Index.
Figure D1.2 Pain on VAS scale for each patient at each 12-hour interval: (a) Tru Straight and (b) Damon 3.
Figure D1.3 Mean pain at each time (95% CI) for Tru Straight (top line) and Damon 3 (bottom line). Time 1 = Day 1 before dinner, Time 2 = Day 2 before breakfast and each subsequent time is 12 hours later until Time 15 = Day 8 before dinner.
Figure D1.4 Box plot showing the distribution of maximum pain intensity in the two bracket groups.
Figure D1.5 Matrix scatter diagram showing the relationship between maximum pain intensity and each of Little’s irregularity indices.
Figure D1.6 Multiple regression model with maximum pain intensity as the outcome variables: (a) predicted values plotted against the residuals; (b) scatter matrix showing the relationship between the residuals and Little’s Total Irregularity Index and (c) histogram of the residuals.
Figure D1.7 Box plot showing the distribution of AUC for pain in the two bracket groups.
Figure D2.1 Scatter diagram showing the relationship between UPSIT-40 and SS-16 scores.
Figure D2.2 Box plot showing the distributions of (a) UPSIT-40 and (b) SS-16.
Figure D2.3 Predicted probability of PD from the logistic regression model plotted against SS score.
Figure D2.4 Log odds of PD plotted against the midpoint of the each quintile subgroup of SS to check for linearity.
Figure D2.5 DFBeta plotted against predicted probability of PD.
Figure D2.6 ROC curves for SS score (solid line) and UPSIT score (dashed line) with dotted reference line.

Part 5: Solutions

Figure S9.1 Box plot showing distributions of BIC% in the immediate loading and conventional loading groups.
Figure S10.1 Box plot showing distribution of SBS values in the six groups.

Introduction

This workbook is a companion volume to the third edition of Medical Statistics at a Glance. Although primarily directed at undergraduate medical students preparing for statistics examinations, we believe that the workbook will also be of use to others working in the biomedical disciplines who simply want to brush up on their analytical and interpretation skills (e.g. other medical researchers, postgraduates in the biomedical disciplines and pharmaceutical industry personnel). Our aim for this workbook is therefore for it to act as a revision aid, equip readers with the skills necessary to read and interpret the published literature and give them the confidence to tackle their own statistical analyses. Although designed as an accompanying text to Medical Statistics at a Glance, it is not indelibly linked to it and can be used as a stand-alone text or in conjunction with any reputable text on statistics.

We believe that the optimal way to learn statistics is to put the theory into practice by undertaking an analysis of a data set, but recognise that this may not always be practical. Instead, the use of carefully constructed exercises in a variety of formats can help to test and fully evaluate the reader’s understanding of the material (and identify any gaps that remain). As the At a Glance textbook presents information in a concise manner, there is limited space in it for worked examples and no room for exercises. Our workbook amends this insufficiency by providing an extensive set of questions, as well as templates for critical appraisal and descriptions of the statistical analyses of two data sets. Where possible, we have based questions on published studies in the medical and dental fields, and references are provided so that the reader may consult the original source material if interested.

The Structure of the Workbook

This workbook is divided into six parts:

Part 1
This section of the workbook contains multiple-choice questions (MCQs) that are generally brief, each testing the reader’s knowledge of a single theoretical concept or aspect of study interpretation. Only one of the five possible answers provided is correct: an explanation is given in Part 5 for each correct and incorrect answer. The ordering of the MCQs generally follows that of the chapters in the third edition of Medical Statistics at a Glance. To aid readers who may wish to focus on specific topics in the At a Glance textbook, we provide a list of MCQs and the related chapters in Appendix I.

Part 2
This section of the workbook contains structured questions that are longer than the MCQs and provide a more in-depth exploration of the reader’s knowledge of several statistical concepts. The questions may include elements that test a reader’s understanding of the theory, as well as his or her ability to interpret study findings and, in some instances, to perform basic statistical calculations. The questions are similar to those that we have set in the past for exams: detailed model answers are provided in Part 5. As the longer structured questions may relate to information contained in several diverse chapters of the textbook, these do not follow any particular order, but in Appendix II, to aid readers who may wish to focus on specific topics in the At a Glance textbook, we provide a list of the structured questions and their related chapters.

Part 3
The ability to critically appraise the published literature is an essential skill that is required by anyone in the medical and dental professions (or, indeed, anyone involved in research more generally) and, consequently, is an important objective of a statistics course. Many aspects of statistics must be considered when evaluating the evidence provided in a research article, for example biases that might arise from inappropriate designs, sample size, outcome measures, the choice of statistical analysis, the presentation of the data and the conclusions drawn. Whilst Medical Statistics at a Glance presents a brief introduction to critical appraisal in Chapter 40 (Evidence-based medicine), Part 3 of our workbook supplements this by providing structured templates that can be used when reviewing and/or assessing the published literature. We suggest that the reader use these templates to critically appraise two published articles: a randomised controlled trial and an observational study. Our own evaluation of these articles is to be found in Part 5. Whilst we cannot hope to cover all possible topics within these two appraisals, we hope that they will at least provide a basic structure for appraisal that readers may find helpful.

Part 4
In our experience, one of the most common complaints from our students and junior research colleagues is that they just do not know where to start when analysing a substantial data set. To address this need, we have included in Part 4 a detailed description of the analyses of two data sets, the latter being available on the accompanying website (www.medstatsaag.com) as Excel files. Each analysis starts with a description of the clinical problem, and then takes the reader through the various steps that would be undertaken when performing the analysis, from the initial exploratory and descriptive analyses to the final sensitivity analyses that assess the robustness of study findings. We believe that this is an innovative approach and hope that readers will find it useful.

Part 5
This section of the workbook contains solutions to the MCQs in Part 1, model answers for the structured questions in Part 2, and our own critical appraisals of the randomised controlled trial (Paper 1) and the observational study (Paper 2) in Part 3. The pages in Part 5 are shaded so that the reader is easily able to navigate to the solutions and model answers.

Appendices
In Appendix I, we provide an ordered list of the MCQs and show which chapters they relate to, with an indication of the material included in each question. Appendix II is similar but identifies the associated chapters for the structured questions. For those readers who require exercises that relate to specific chapters of the At a Glance textbook, we provide a list of the chapters and indicate which multiple-choice and structured questions are relevant to them in Appendix III.

Further Information

In addition to the workbook, we remind readers that the companion website to Medical Statistics at a Glance (www.medstatsaag.com) also contains an extensive set of interactive exercises, with references to many published papers that may be of interest.

Acknowledgements

Special thanks are due to Drs Laura Silveira-Moriyama and Angus Pringle who very kindly lent us their data sets for the analyses in Part 4 of the workbook. We are most appreciative of the extremely helpful comments and suggestions that they made during the development of the analyses, but we take full responsibility for any errors or misconceptions in the final presentations. We are also indebted to the authors and publishers of the two papers that we used for critical appraisal for allowing us to reproduce the articles, thereby providing useful exercises for our readers, and apologise if any of our criticisms cause offence. We acknowledge the generosity of the many authors and publishers who have kindly assented to our adapting or reproducing material for the multiple-choice and structured questions, and are grateful to the publishing team at Wiley-Blackwell both for suggesting that we write this workbook and for their ideas and support along its route to publication. Our acknowledgements would not be complete without thanking our students over the years from whom we have learnt the art of teaching, and Mike, Gerald, Nina, Andrew and Karen for their forbearance, encouragement and good humour during our absorption with this manuscript.