TRRM - 2005/6 1

Study protocol

Jim teWaterNaude

UCT Dept Public Health

TRRM - 2005/6 2

Research

The process of asking and answering a question

TRRM - 2005/6 3

Protocol

The reason for having for a study plan or protocol is efficiency

You want to be able to do the intended study quickly, cheaply, easily and ethically, and with no major oh-my-gosh-I-didn’t-think-of-that moments.

Here follow 3 examples of protocol outlines

Element Purpose

1. Research questions What questions will the study address?

2. Significance (background) Why are these questions important?

3. Design Time frame Epidemiologic approach

How is the study structured?

4. Subjects Selection criteria Sampling design

Who are the subjects and how will they be selected?

5. Variables Predictor variables Confounding variables Outcome variables

What measurements will be made?

6. Statistical issues Hypotheses Sample size Analytic approach

How large is the study and how will it be analyzed?

1. Study protocol outline(This is the preferred style, as found in Hulley)

TRRM - 2005/6 5

Hulley explains the process in familiar terms:

Anatomy of research(what it’s made of)

• Research question • Significance • Design • Subjects – Population – Sample size • Variables – Predictor – Outcome

Physiology of research (how it works)

Using measurements in a sample to draw inferences about phenomena in a population

2. Major headings in a protocol(MRC approach)

TitleInvestigators, institutional affiliations, & qualificationsSummary/Abstract1. Introduction

– Literature review; Motivation for the study (problem); Purpose; Specific objectives; Implementation objectives

2. Methods– Definition of terms; Study design; Study population and sampling;

Measurements; Pilot studies3. Logistics and time schedule

– Responsibilities of investigators and of staff; Time schedule4. Data management and analysis5. Resources

– Available resources; Budget and budget motivation6. Ethical and legal considerations7. Reporting of results8. Appendices

TRRM - 2005/6 7

3. Components of a research proposal(Shi)

A. Title pageB. Table of contentsC. AbstractD. Project Description

1. Introduction2. Problem statement and significance3. Goals and objectives4. Methods and procedures5. Evaluation6. Dissemination

E. ReferencesF. Budget and JustificationG. Human SubjectsH. Appendices

TRRM - 2005/6 8

Main protocol elements

The above protocol outlines have these as main elements:

Population Question Relevance Study design and variables Timeframe

To understand these, we need to revisit some epidemiology…

TRRM - 2005/6 9

Epidemiology

The study of the distribution and determinants of health in humans

It is the science of the occurrence of disease in human populations

TRRM - 2005/6 10

… distribution and determinants…

Distribution ‘What’ (the disease), plus ‘when’ + ‘where’ + ‘who’

or the disease described by TIME + PLACE + POPULATION

DeterminantsThe ‘Why’

= the ‘causes’, or risk factors

TRRM - 2005/6 11

DistributionWhat, when, where, who

• Descriptive studies

Determinants‘Why’

What is associated with/ caused by• Analytic studies• Interventive studies

… distribution and determinants…

TRRM - 2005/6 12

Population

“a group sharing certain common characteristics”

Do not have to be people (most often are) Can be records, institutions, farms, events Does need to be clearly defined and

specified

TRRM - 2005/6 13

QuestionQuestions arise out of involvement in the field

The research question occurs in 3 layers:1. The conceptual research question is what

the investigator wants to know about the world

2. The operational research question is what is deemed achievable though the deliberations of drawing up the study plan or protocol

3. The actual research question is what actually gets answered during the conduct of the study.

TRRM - 2005/6 14

The research cycle(after Hulley)

Actual question

Findings in the study

Operational question

Truth in the study

Conceptual question

Truth in the universe

Design Implement

InferInfer

TRRM - 2005/6 15

The research cycle

Actual question

Findings in the study

Operational question

Truth in the study

Conceptual question

Truth in the universe

Design Implement

InferInfer

Targetpopulation

Phenomenaof interest

Intendedsample

Intendedvariables

Actualsubjects

Actualmeasurements

TRRM - 2005/6 16

Asking your question

++++++++++++++++++++++++++++++++++++++

Answering your question

Research – the process of asking and answering a question

TRRM - 2005/6 17

Asking your question

Background

Methods

+++++++++++++++++++++++++++++++++++++++++

Results

Discussion

Answering your question

TRRM - 2005/6 18

Asking your question

General

Specific

+++++++++++++++++++++++++++++++++++++++++

Specific

General

Answering your question

TRRM - 2005/6 19

Asking your question

Conceptual

Operational

+++++++++++++++++++++++++++++++++++++++++

Operational

Conceptual

Answering your question

TRRM - 2005/6 20

Asking your question

BackgroundLiterature review

Objectives

MethodsDesign

Measurements

++++++++++++++++++++++++++++++

TRRM - 2005/6 21

++++++++++++++++++++++++++++++

ResultsLimitations

Findings

DiscussionWhat this means

Conclusion

Answering your question

22

Research questions using existing data

1. Choose a database2. Become thoroughly familiar with all variables and

how they were measured3. Identify variables whose association may be of

interest4. Review the literature and consult experts to

determine relevant research questions5. Formulate specific hypotheses and settle on the

statistical methods

Table 13.1 in Designing Clinical Research

Will you use existing, or generate your own data?

TRRM - 2005/6 23

“… but it doesn’t mean anything”(one of the children’s lament from Sound of Music)

Population Question Relevance (it has to mean something)

Study design and variables Timeframe

TRRM - 2005/6 24

FINER - Criteria for a Good Research Question Table 2.1 in Hulley

FeasibleAdequate number of subjectsAdequate technical expertiseAffordable in time and moneyManageable in scope

InterestingTo the investigator

NovelConfirms or refutes previous findingsExtends previous findingsProvides new findings

Ethical Relevant

To scientific knowledgeTo clinical and health policyTo future research directions

Relevance

TRRM - 2005/6 25

Study– design and variables

There are 2 broad study design classes and

2 broad variable classes.

Which is the best study design?

“The question being asked determines the appropriate research architecture, strategy, and tactics to be used”

This is quoted from the accompanying editorial:

Sackett DL, Wennberg JE. Choosing the best research design for each question. Editorial. BMJ 1997;315:1636

TRRM - 2005/6 26

Study designs- arranged in ascending order of credibility

Case reports Case series Correlational studies Cross-sectional studies Case-control studies Cohort studies Controlled trials

TRRM - 2005/6 27

Descriptive and analytic In descriptive epidemiology, we describe the

distribution of an exposure or outcome, without overtly seeking to explain the distribution by looking for associations

(distribution of health in humans) In analytical epidemiology, we examine

associations, often with the aim of identifying possible causes for an outcome

(determinants of health in humans)

2 broad study design classes

TRRM - 2005/6 28

Observational and interventional In observational epidemiology we examine the

distribution or determinants of an outcome without any attempt to influence them

(examples are smoking, drinking, sexual behaviours)

In interventional epidemiology we test a hypothesis by modifying an exposure within the study population and examining the effect on the outcome

(examples are vaccine or drug trials)

Analytic study designs

TRRM - 2005/6 29

Exposure and outcome variables Exposures are also called risk factors,

factors, predictor variables, independent variables.

These may or may not be the cause of the outcome – we determine this through research

Outcomes are also called effects, dependent variables, diseases, events, or health-related states that we are interested in

2 broad variable classes

TRRM - 2005/6 30

Time frame

“Rule of halves”– Spend half on Asking

• ½ on Background• ½ on Methods

– Spend half on Answering• ½ on Results• ½ on Discussion

½ + ½ + ½ + ½ = 1

TRRM - 2005/6 31

Epidemiological terms not covered

Quantitative and Qualitative Validity and Variability Causality/ Causal inference Bias Confounding The 2x2 table Occurrence and effect measures

TRRM - 2005/6 32

Protocol – more comments

The scientific thinking that goes into the protocol attempts to control the errors that commonly occur – these errors are either random (due to chance) or systematic (due to bias)

Developing a protocol is an iterative process of drafting and redrafting, visiting and revisiting, and is best approached initially by divergent rather than convergent thinking

The perfect study has yet to be done, as there are many trade-offs in the process of asking and answering your question

TRRM - 2005/6 33

An example of applying PQRST- the Dop system

Population Productive farms in the Stellenbosch district Question What is the current prevalence of the Dop system (in 1995)? Relevance Never previously documented. Presumed health effects Study design and variables Cross-sectional descriptive. Frequency & amount of dop Timeframe ~ 4 months in all. 3-4 weeks for data collection, collected by

nurses on their mobile clinic visits to the farms

TRRM - 2005/6 34

Apply PQRST:

TRRM - 2005/6 35

1st Task at the lecture- task taken from Chapter 1 of the textbook

For each of the following 4 abstracts:1. State the research question in a single sentence

that specifies the predictor and the outcome variables, as well as the population sampled

2. State the study design. Think also about the main inference that can be drawn from the study, to whom it can be generalised, and what the potential errors in drawing and applying these inferences are

(Answers are given at the very end of this presentation)

TRRM - 2005/6 36

a. Giving vitamin D to patients with vitamin D deficiency can improve strength. To find out whether the ordinary weakness of aging could be treated with vitamin D, we selected 38 men and women 70 years of age and greater from a hypertension treatment clinic and randomly assigned them to receive either vitamin D3 or identical placebo. Muscle strength of the quadriceps, measured with an isokinetic dynamometer after 6 months of treatment, was similar in the two groups.

The research question:The study design:

TRRM - 2005/6 37

b. To assess whether the sedative effects of psychotropic drugs might cause hip fractures, we studied 1021 men and women with hip fractures and 5606 without hip fractures among elderly Medicaid enrollees. Persons treated with short-acting tranquillizers had no increased risk of hip fracture. By contrast, there was an increased risk associated with current use of tranquillizers having half-lives of more than 24 hours (odds ratio, 1.8; 95 percent confidence interval, 1.3 to 2.4).

The research question:The study design:

TRRM - 2005/6 38

c. Knowledge about AIDS was studied among 893 teenaged boys and 633 girls drawn from 12 secondary schools in Zimbabwe. Ninety-three percent of the children thought that it was an infection caused by having sexual relations, and 10% believed that it could be contracted from toilet seats.

The research question:The study design:

TRRM - 2005/6 39

d. We examined the use of estrogen replacement therapy in relation to breast cancer in postmenopausal women. During 367 187 person-years of follow-up, there were 722 new cases of breast cancer. The risk of breast cancer was significantly elevated among current estrogen users (relative risk, 1.36; 95% confidence interval, 1.11 to 1.67), but not among former users.

The research question:The study design:

TRRM - 2005/6 40

2nd Task at the lecture

Write down your own Research Question in one sentence

TRRM - 2005/6 41

Task for September

Get the book

Get a research mentor in your department

Decide on your area of research. Will you use existing or generate new data?

Compose your one sentence Research Question

Flesh this out into a half-page study plan or protocol, using a structure that suits you. Start your literate review.

Submit this to jim@cormack.uct.ac.za for feedback

TRRM - 2005/6 42

Task for December

Complete your 5-page protocol(format outline in the next slide)

Submit this to jim@cormack.uct.ac.za for feedback

TRRM - 2005/6 43

5-page protocol

Title Abstract Specific aims Significance (limit to 1/2 page) Design Overview (time frame and nature of control) Study subjects (selection criteria, plans for recruiting) Measurements (predictors, confounders, outcomes) Statistical issues (hypotheses and sample size) Pretests, quality control, data management TimetableAddendum: Ethical considerations (not part of the 5

pages)

TRRM - 2005/6 44

(but do please get to the start line)

Many thanks

UCT Dept Public Health

“It’s not how you start, it’s how you finish”9 times winner Bruce Fordyce, describing the Comrades Marathon

TRRM - 2005/6 45

a. Giving vitamin D to patients with vitamin D deficiency can improve strength. To find out whether the ordinary weakness of aging could be treated with vitamin D, we selected 38 men and women 70 years of age and greater from a hypertension treatment clinic and randomly assigned them to receive either vitamin D3 or identical placebo. Muscle strength of the quadriceps, measured with an isokinetic dynamometer after 6 months of treatment, was similar in the two groups.

1. Does treatment with vitamin D increase leg muscle strength in healthy people 70 years of age or greater?

Audience comment: Healthy should read Hypertensive

2. Randomised blinded trial

TRRM - 2005/6 46

b. To assess whether the sedative effects of psychotropic drugs might cause hip fractures, we studied 1021 men and women with hip fractures and 5606 without hip fractures among elderly Medicaid enrollees. Persons treated with short-acting tranquillizers had no increased risk of hip fracture. By contrast, there was an increased risk associated with current use of tranquillizers having half-lives of more than 24 hours (odds ratio, 1.8; 95 percent confidence interval, 1.3 to 2.4).

1. Do psychotropic medications increase risk for hip fracture among elderly men and women?

2. Case-control studyAudience comment: Surely this could be a cross-sectional study?Response: Yes, but the wording seems to favour case-control,

because they mention the hip fracture group as if they had been initially selected before the controls were selected, reflecting the conduct of a case-control study. Another clue is the use of odds ratio, which is synonymous with case-control studies

TRRM - 2005/6 47

c. Knowledge about AIDS was studied among 893 teenaged boys and 633 girls drawn from 12 secondary schools in Zimbabwe. Ninety-three percent of the children thought that it was an infection caused by having sexual relations, and 10% believed that it could be contracted from toilet seats.

1. What is the state of knowledge about AIDS among schoolchildren in Zimbabwe?

2. Cross sectional descriptive study

TRRM - 2005/6 48

d. We examined the use of estrogen replacement therapy in relation to breast cancer in postmenopausal women. During 367 187 person-years of follow-up, there were 722 new cases of breast cancer. The risk of breast cancer was significantly elevated among current estrogen users (relative risk, 1.36; 95% confidence interval, 1.11 to 1.67), but not among former users.

1. Does estrogen replacement therapy in postmenopausal women increase risk for breast cancer?

2. Cohort study