Post on 27-Jan-2016
description
Lesson 1Lesson 1Lesson 1Lesson 1The Cycle of ResearchThe Cycle of Researchyy
Definitions of TermsDefinitions of TermsStudy DesignStudy Designy gy g
Measurement ScalesMeasurement Scales
Understanding the ResearchUnderstanding the ResearchUnderstanding the ResearchUnderstanding the Research
The description of any medical or health research study The description of any medical or health research study p y yp y yshould provide answers to the following:should provide answers to the following: How were the data collected?How were the data collected?
What was the sampling method?What was the sampling method?gg What is the study design?What is the study design?
In addition to understanding how the data were In addition to understanding how the data were collected, the measurement scale of the data needs to collected, the measurement scale of the data needs to ,,be consideredbe considered
Both the type of data collected (measurement scale) and Both the type of data collected (measurement scale) and how the data were collected determine the appropriatehow the data were collected determine the appropriatehow the data were collected determine the appropriate how the data were collected determine the appropriate descriptive statistics, graphs, and analysis methods to descriptive statistics, graphs, and analysis methods to use. use.
2PubH 6414 Lesson 1
Cycle of ResearchCycle of ResearchCycle of ResearchCycle of Research
3PubH 6414 Lesson 1
Samples and PopulationsSamples and PopulationsSamples and PopulationsSamples and Populations PopulationPopulation:: complete set of individuals that complete set of individuals that
share some common characteristic(s)share some common characteristic(s)share some common characteristic(s). share some common characteristic(s). the group we are interested in learning aboutthe group we are interested in learning about..
Sample:Sample: A subset of the defined population. A subset of the defined population. the group that is actually studied and from which data the group that is actually studied and from which data
are collected. are collected.
4PubH 6414 Lesson 1
Why study samples?Why study samples?Why study samples?Why study samples? If the goal is to make some conclusion(s) about If the goal is to make some conclusion(s) about
the population why not just collect data on thethe population why not just collect data on thethe population, why not just collect data on the the population, why not just collect data on the entire population?entire population?
ItIt b t i d/ tib t i d/ ti i t ll ti t ll t It It can be too expensive and/or timecan be too expensive and/or time--consuming to collect consuming to collect data for the data for the populationpopulation
St d f th l ti bSt d f th l ti b i ibli ibl Study of the population may be Study of the population may be impossibleimpossible
Measurements Measurements from a sample may be more accurate from a sample may be more accurate p yp ythan measurements from the populationthan measurements from the population
5PubH 6414 Lesson 1
Cycle of ResearchCycle of ResearchCycle of ResearchCycle of Research
6PubH 6414 Lesson 1
Samples and PopulationsSamples and PopulationsSamples and PopulationsSamples and Populations Statistic:Statistic: A value obtained from a sample. A value obtained from a sample.
i i ht f 1 000 b l d f thi i ht f 1 000 b l d f th i.e. average weight of 1,000 boys sampled from the i.e. average weight of 1,000 boys sampled from the U.S. U.S.
Parameter:Parameter: A A population variable, whose value population variable, whose value i ki kis unknown. is unknown. i.e. average weight of ALL 10 year old boys in the i.e. average weight of ALL 10 year old boys in the
U.S. U.S.
7PubH 6414 Lesson 1
Random SamplingRandom SamplingRandom SamplingRandom Sampling
Generalization of studyGeneralization of study results is only valid if theresults is only valid if the Generalization of study Generalization of study results is only valid if the results is only valid if the sample is sample is representativerepresentative of the larger population of the larger population that it is drawn from.that it is drawn from.
Random samplesRandom samples tend to be representative of tend to be representative of the population from which the sample is drawn.the population from which the sample is drawn.
Random samples are also called Random samples are also called Probability Probability samplessamples because the probability of each subject because the probability of each subject ( i ) b i i l d d i h l i k( i ) b i i l d d i h l i k(or unit) being included in the sample is known. (or unit) being included in the sample is known.
8PubH 6414 Lesson 1
Simple Random Sample (SRS)Simple Random Sample (SRS)Simple Random Sample (SRS)Simple Random Sample (SRS)
Basic idea: Select a sample from the population Basic idea: Select a sample from the population p p pp p pso that:so that: Each subject in the population has equal probability of Each subject in the population has equal probability of
being selectedbeing selectedbeing selectedbeing selected Selection of one subject does not influence the Selection of one subject does not influence the
probability of selecting another subjectprobability of selecting another subject Formal definition: A sample of size Formal definition: A sample of size nn selected by selected by
a procedure that gives every sample of size a procedure that gives every sample of size nnthe same probability of being selectedthe same probability of being selectedthe same probability of being selectedthe same probability of being selected
SRS removes selection bias and results in a SRS removes selection bias and results in a representative samplerepresentative sample
9PubH 6414 Lesson 1
representative samplerepresentative sample
Selection BiasSelection BiasSelection BiasSelection Bias
Selection Bias: “A systematic tendency to favor Selection Bias: “A systematic tendency to favor y yy ythe inclusion in a sample of selected subjects the inclusion in a sample of selected subjects with particular characteristics while excluding with particular characteristics while excluding those with other characteristics.”those with other characteristics.”
Pocket Dictionary of StatisticsPocket Dictionary of Statistics
Garbage in Garbage in Garbage outGarbage out
Bias (in general): a systematic deviation from the Bias (in general): a systematic deviation from the
10PubH 6414 Lesson 1
truthtruth
How to obtain a Simple Random How to obtain a Simple Random Sample (SRS):Sample (SRS):
Obtaining a simple random sample requires that allObtaining a simple random sample requires that allObtaining a simple random sample requires that all Obtaining a simple random sample requires that all possible subjects are identified. possible subjects are identified.
Any of the following methods will result in a random Any of the following methods will result in a random y gy gsample of size sample of size nn
Place an ID for each subject in a hat and Place an ID for each subject in a hat and jjrandomly select a sample of size randomly select a sample of size nn
Use a table of random numbers to select a Use a table of random numbers to select a sample of size sample of size nn from the list from the list
Use a computer program to select a random Use a computer program to select a random
11PubH 6414 Lesson 1
sample of size sample of size nn from the listfrom the list
Sampling with and without Sampling with and without replacementreplacement
Sampling with replacement:Sampling with replacement: Sampling with replacement: Sampling with replacement: After a subject is selected, return the ID to the listAfter a subject is selected, return the ID to the list Possible for a single subject to be selected twicePossible for a single subject to be selected twiceg jg j
Sampling without replacementSampling without replacement Once selected the subject is not available for future Once selected the subject is not available for future
selectionselection Most often, sampling without replacement is usedMost often, sampling without replacement is used
F l l ti th diff b tF l l ti th diff b t For large populations, the difference between For large populations, the difference between these is negligiblethese is negligible
12PubH 6414 Lesson 1
Stratified Random SamplingStratified Random Samplingp gp g For stratified random sampling the population is For stratified random sampling the population is
divided into groups with similar characteristics. divided into groups with similar characteristics. These groups are called ‘strata’. A simple These groups are called ‘strata’. A simple random sample is selected from each stratarandom sample is selected from each strataPossible stratification variables arePossible stratification variables are Possible stratification variables arePossible stratification variables are GenderGender ClinicClinicClinicClinic Age groupsAge groups
This is a useful method when the subgroups This is a useful method when the subgroups vary in the study measurementsvary in the study measurements
Requires that information about the stratification Requires that information about the stratification variable(s) is available for all subjectsvariable(s) is available for all subjects
13PubH 6414 Lesson 1
variable(s) is available for all subjectsvariable(s) is available for all subjects
Consider the sampling MethodConsider the sampling MethodConsider the sampling MethodConsider the sampling Method
When reading the medical literature consider theWhen reading the medical literature consider the When reading the medical literature consider the When reading the medical literature consider the following:following: Is the study sample representative of the population Is the study sample representative of the population
of interest?of interest? What was the sampling method? If a random sample What was the sampling method? If a random sample
was used the statistical inference is validwas used the statistical inference is validwas used, the statistical inference is validwas used, the statistical inference is valid If complete population data were collected, no If complete population data were collected, no
inference is necessary. Descriptive statistics inference is necessary. Descriptive statistics y py pcompletely represent the population when complete completely represent the population when complete population data are available.population data are available.
14PubH 6414 Lesson 1
Cycle of ResearchCycle of ResearchCycle of ResearchCycle of Research
15PubH 6414 Lesson 1
Study DesignStudy DesignStudy DesignStudy Design
Twelve 18Twelve 18--24 year old males in a jogging24 year old males in a jogging Twelve 18Twelve 18 24 year old males in a jogging 24 year old males in a jogging club participated in a study of brain club participated in a study of brain gycogengycogen among physically fit individualsamong physically fit individualsgycogengycogen among physically fit individuals. among physically fit individuals. The The populationpopulation of interest is (potentially) of interest is (potentially)
1818 24 year24 year oldold malesmales1818--24 year24 year--old old males.males.The The samplesample is this group of twelve is this group of twelve
jjjoggers.joggers.The The sample design sample design is how these twelve is how these twelve
16PubH 6414 Lesson 1
where selected and studied. where selected and studied.
Cycle of ResearchCycle of ResearchCycle of ResearchCycle of Research
17PubH 6414 Lesson 1
CalculationCalculationCalculationCalculation
Th t l l i h d ltiTh t l l i h d lti The twelve males were weighed, resulting The twelve males were weighed, resulting in weights in pounds of :in weights in pounds of :
{129,134,136,140,141,142,144,155,158,162,{129,134,136,140,141,142,144,155,158,162,165,191}165,191}The The statistics statistics how obtained were the how obtained were the
average and standard deviation of the average and standard deviation of the
18PubH 6414 Lesson 1
ggweights. weights.
Cycle of ResearchCycle of ResearchCycle of ResearchCycle of Research
19PubH 6414 Lesson 1
EstimationEstimationEstimationEstimation
The sample statistics are our estimates of The sample statistics are our estimates of the population values:the population values:the population values:the population values:
Average: 149.75 poundsAverage: 149.75 pounds Standard Deviation: 17.35 poundsStandard Deviation: 17.35 pounds
20PubH 6414 Lesson 1
Cycle of ResearchCycle of ResearchCycle of ResearchCycle of Research
21PubH 6414 Lesson 1
InferenceInferenceInferenceInference
B i ith ti l h lth d tB i ith ti l h lth d t By comparison with national health data By comparison with national health data from the CDC, we conclude that young from the CDC, we conclude that young thl ti l h l i htthl ti l h l i htathletic males have lower average weights athletic males have lower average weights
than young males overall, and thus their than young males overall, and thus their b i l lt t bb i l lt t bbrain glycogen results may not be brain glycogen results may not be generalizable to young males overall.generalizable to young males overall.
22PubH 6414 Lesson 1
SamplePopulation
study cyclestudy cyclestudy cyclestudy cycle
StatisticParameter
Consider This StudyConsider This StudyConsider This StudyConsider This Study Two thousand AfricanTwo thousand African--American women, ages 35 to 55 years, American women, ages 35 to 55 years,
were surveyed to determine their adherence to recommendedwere surveyed to determine their adherence to recommendedwere surveyed to determine their adherence to recommended were surveyed to determine their adherence to recommended breast and cervical cancer screening schedules. At the time of breast and cervical cancer screening schedules. At the time of the survey, each woman lived in the inner city of a southern the survey, each woman lived in the inner city of a southern metropolitan areametropolitan areametropolitan area. metropolitan area.
Status Household income<$5000
Education>high school
Not married or
Employed
Had Papsmear
had mammography<$5000 school or
separatedsmear hy
Yes 27% 68% 68% 80% 48% 35%No 73% 32% 32% 20% 52% 65%No 73% 32% 32% 20% 52% 65%
The investigators concluded that African-American women in this age group who live in the inner city do not adhere tothis age group who live in the inner city do not adhere to suggested screening schedules.
Question 1: Question 1:
“___” corresponds to the SamplePopulationpprocedure highlighted in
this picture?this picture? StatisticParameter
1.1. The investigators made the conclusion The investigators made the conclusion whether Africanwhether African--American women in American women in this age group who live in the inner city this age group who live in the inner city adhere to suggested screening adhere to suggested screening schedules.schedules.
ff2.2. The health and socialThe health and social--economic data of economic data of those women were recorded.those women were recorded.
3.3. 2000 African2000 African--American women, ages American women, ages 35 55 d li i h i i35 55 d li i h i i35 to 55 years and lives in the inner city 35 to 55 years and lives in the inner city , were surveyed in this study., were surveyed in this study.
Question 2:Question 2:select the correct picture for the
following procedure.following procedure.“Twenty-seven percent of the women had a household income under $5000.”
1. 2.
3. 4.
Question 3:Question 3:
“Forty-eight percent had their most y g precent Pap smear” is ___?
1.1. ParameterParameter22 Statistic used toStatistic used to2.2. Statistic, used to Statistic, used to
estimate estimate parameterparameterparameterparameter
3.3. Statistic, also the Statistic, also the ttparameterparameter
Question 4:Question 4:l h i di hselect the picture corresponding to the
following statement.g“The investigators concluded that African-American women in this age group who live in the inner city do not adhere to suggested screening schedules ”adhere to suggested screening schedules.”1. 2.
4.3.
More on “How More on “How was the was the Data CollectedData Collected?”?”
Study DesignStudy DesignStudy DesignStudy Design
Classification of Study DesignsClassification of Study DesignsClassification of Study DesignsClassification of Study DesignsStudy designs can be classified as either Observational or Study designs can be classified as either Observational or
ExperimentalExperimentalExperimentalExperimentalObservational Studies: Observational Studies: subjects are observed subjects are observed withoutwithout any any
interventionintervention by the investigatorby the investigator Case seriesCase series CrossCross--Sectional Sectional CaseCase ControlControl CaseCase--Control Control Cohort Cohort Experimental Studies: Experimental Studies: a comparative study a comparative study involving an involving an
interventionintervention by the investigatorby the investigator Controlled studiesControlled studies Uncontrolled studiesUncontrolled studies
30PubH 6414 Lesson 1
Uncontrolled studiesUncontrolled studies
Case SeriesCase SeriesCase SeriesCase Series
A case series is observationalA case series is observational Subjects are observed and the outcomes or Subjects are observed and the outcomes or
characteristics of the observation are describedcharacteristics of the observation are described Case series typically have a small number of Case series typically have a small number of subjectssubjects
These are not planned studies so they are not These are not planned studies so they are not always considered ‘studies’always considered ‘studies’
Observations from a case series may lead to Observations from a case series may lead to research questions for more rigorously designed research questions for more rigorously designed studiesstudies
31PubH 6414 Lesson 1
CaseCase--Series ExampleSeries ExampleCaseCase Series ExampleSeries Example
In 1981, Gottlieb et al. reported a rare form of In 1981, Gottlieb et al. reported a rare form of , p, ppneumonia and unusual infections in 5 young pneumonia and unusual infections in 5 young menmen
G ttli b t lG ttli b t l P ti i iiP ti i ii i di d Gottlieb, et al. Gottlieb, et al. Pneumocystis cariniiPneumocystis carinii pneumonia and pneumonia and mucosal candidisasismucosal candidisasis in previously healthy in previously healthy homosexual men: Evidence of a new acquired cellular homosexual men: Evidence of a new acquired cellular i d fi ii d fi i N E l d J l f M di iN E l d J l f M di iimmunodeficiency.immunodeficiency. New England Journal of Medicine, New England Journal of Medicine, 1981; 305(24): 14251981; 305(24): 1425--14311431
Other similar case series were reportedOther similar case series were reportedOther similar case series were reportedOther similar case series were reported By 1982 this condition was named: Acquired By 1982 this condition was named: Acquired
Immune Deficiency Syndrome (AIDS)Immune Deficiency Syndrome (AIDS)
32PubH 6414 Lesson 1
CrossCross--Sectional DesignSectional DesignCrossCross Sectional DesignSectional Design
CrossCross--sectional studies collect data from a sectional studies collect data from a sample of subjects at one point in time. sample of subjects at one point in time. Provides a ‘snapshot’ of the characteristics of interest Provides a ‘snapshot’ of the characteristics of interest Often large number of subjectsOften large number of subjects Often large number of subjectsOften large number of subjects
Collected Collected data can be used to explore data can be used to explore relationships between variables at a point in timerelationships between variables at a point in time Possible outcome variables: Disease, Death, EventPossible outcome variables: Disease, Death, Event Explanatory variables: Exposure to some Risk FactorExplanatory variables: Exposure to some Risk FactorExplanatory variables: Exposure to some Risk FactorExplanatory variables: Exposure to some Risk Factor Demographic characteristics: age gender, ethnicity, Demographic characteristics: age gender, ethnicity,
etc.etc.
33PubH 6414 Lesson 1
CrossCross--sectional Design: Conssectional Design: ConsCrossCross sectional Design: Conssectional Design: Cons
Cross sectional studies are not useful for Cross sectional studies are not useful for identifying the ‘causeidentifying the ‘cause--effect’ relationship effect’ relationship between variablesbetween variables
There There is potential for bias in data collected by is potential for bias in data collected by surveyssurveyssurveyssurveys Volunteer biasVolunteer bias
NonNon--response biasresponse bias
34PubH 6414 Lesson 1
CaseCase--Control DesignControl DesignCaseCase Control DesignControl Design CaseCase--Control studies are useful for exploring the Control studies are useful for exploring the
relationship between disease (or some other outcome)relationship between disease (or some other outcome)relationship between disease (or some other outcome) relationship between disease (or some other outcome) and possible risk factorsand possible risk factors
Subjects are selected based on the outcomeSubjects are selected based on the outcome‘‘CC ’’ l d h th dil d h th di ‘‘CasesCases’ ’ –– already have the disease already have the disease
‘‘ControlsControls’ ’ –– individuals from a similar population who do not have individuals from a similar population who do not have the disease the disease
D t iD t i tt t t d i k f t ( ) ft t d i k f t ( ) f Determine Determine pastpast exposure to suspected risk factor(s) for exposure to suspected risk factor(s) for the disease or condition fromthe disease or condition from Interviews with subjects or family membersInterviews with subjects or family members Medical recordsMedical records
These are often called Retrospective studies because These are often called Retrospective studies because the outcome already occurred prior to inclusion in the the outcome already occurred prior to inclusion in the
35PubH 6414 Lesson 1
y py pstudy. study.
CaseCase--Control Design ExampleControl Design ExampleCaseCase Control Design ExampleControl Design Example
Inskip, PD et al. Cellular telephone use and brain Inskip, PD et al. Cellular telephone use and brain tt N E l d J l f M di iN E l d J l f M di i (2001)(2001)tumors. tumors. New England Journal of MedicineNew England Journal of Medicine (2001), (2001), 344(2), 79344(2), 79--8686
Cases and Controls were identifiedCases and Controls were identified 782 cases with various types of brain tumors 782 cases with various types of brain tumors 799 controls with nonmalignant conditions799 controls with nonmalignant conditions
Exposure to cell phone use was evaluated by subject Exposure to cell phone use was evaluated by subject i t ii t iinterview interview
Study conclusion: “There was no indication of higher Study conclusion: “There was no indication of higher brain tumor risk among persons who had used handbrain tumor risk among persons who had used hand--held cellular phones compared to those who had notheld cellular phones compared to those who had notheld cellular phones compared to those who had not held cellular phones compared to those who had not used them.” used them.”
htt // / t i /f t h t/b ihtt // / t i /f t h t/b i tt
36PubH 6414 Lesson 1
http://www.cancer.gov/cancertopics/factsheet/brainhttp://www.cancer.gov/cancertopics/factsheet/brain--tumorstumors--cellcell--phonesphones
CaseCase--Control DesignControl Design
Drawbacks of caseDrawbacks of case--control studies: control studies: Misclassification Bias: It isn’t always easy to determineMisclassification Bias: It isn’t always easy to determine Misclassification Bias: It isn t always easy to determine Misclassification Bias: It isn t always easy to determine
disease statusdisease status Recall Bias: Subjects don’t always accurately remember Recall Bias: Subjects don’t always accurately remember
hihiexposure historyexposure history Cannot be used to estimate the prevalence of a condition.Cannot be used to estimate the prevalence of a condition. SelectionSelection of appropriate controls can be difficultof appropriate controls can be difficult Selection Selection of appropriate controls can be difficult of appropriate controls can be difficult
Advantages of caseAdvantages of case--control studies: control studies: Can obtain an adequate number of cases for studies ofCan obtain an adequate number of cases for studies of Can obtain an adequate number of cases for studies of Can obtain an adequate number of cases for studies of
rare diseasesrare diseases Economical and quick compared to prospective studiesEconomical and quick compared to prospective studies
37PubH 6414 Lesson 1
Cohort DesignCohort DesignCohort DesignCohort Design Cohort studies follow a group of subjects (cohort) Cohort studies follow a group of subjects (cohort)
over timeover time to determine disease or mortality statusto determine disease or mortality statusover timeover time to determine disease or mortality status.to determine disease or mortality status. Sometimes called followSometimes called follow--up or longitudinal studiesup or longitudinal studies
Members of the cohort can be classified as Members of the cohort can be classified as ‘exposed’ or ‘not exposed’ to one or more risk ‘exposed’ or ‘not exposed’ to one or more risk factors at the beginning of the study. For example:factors at the beginning of the study. For example:
k dk d kk smokers and nonsmokers and non--smokers smokers overweight and normal weightoverweight and normal weight
The study follows the cohort over time toThe study follows the cohort over time to The study follows the cohort over time to The study follows the cohort over time to investigate the association between exposures investigate the association between exposures and outcomes.and outcomes.
38PubH 6414 Lesson 1
Cohort Study DesignCohort Study DesignCohort Study DesignCohort Study Design MostMost cohort studies are cohort studies are prospectiveprospective
Ri k f t i d t i d i thRi k f t i d t i d i th Risk factor exposure is determined in the Risk factor exposure is determined in the present and information about the outcome is present and information about the outcome is collected at a future timecollected at a future timeco ected at a utu e t eco ected at a utu e t e
Cohort studies can also be Cohort studies can also be retrospectiveretrospective Exposure and outcome have already occurredExposure and outcome have already occurred Exposure and outcome have already occurredExposure and outcome have already occurred Requires having complete and accurate Requires having complete and accurate
records to “follow” subjects from exposure to records to “follow” subjects from exposure to outcomeoutcome
These are also called ‘historical cohort studies’These are also called ‘historical cohort studies’
39PubH 6414 Lesson 1
Cohort Study ExampleCohort Study ExampleCohort Study ExampleCohort Study Example
NIH: Division of Epidemiology, Statistics & NIH: Division of Epidemiology, Statistics & P ti R h (DESPR) N ti l ChildP ti R h (DESPR) N ti l ChildPrevention Research (DESPR): National Childrens Prevention Research (DESPR): National Childrens StudyStudy
The The National Children's StudyNational Children's Study is the largest and longest is the largest and longest study of environmental effects on children’s health ever study of environmental effects on children’s health ever conducted in the United States. The Study will follow conducted in the United States. The Study will follow 100 000 children from before birth to age 21 to examine100 000 children from before birth to age 21 to examine100,000 children from before birth to age 21 to examine 100,000 children from before birth to age 21 to examine the effects of environmental influences on their health the effects of environmental influences on their health and development. and development.
For more information about the Study and its progress, For more information about the Study and its progress, visit the Study Web site at visit the Study Web site at http://www nationalchildrensstudy govhttp://www nationalchildrensstudy gov
40PubH 6414 Lesson 1
http://www.nationalchildrensstudy.govhttp://www.nationalchildrensstudy.gov..
Cohort DesignCohort DesignCohort DesignCohort Design Drawbacks of cohort studies: Drawbacks of cohort studies:
TimeTime consumingconsuming andand costlycostly TimeTime--consuming consuming and and costlycostly..
Not suitable for very rare Not suitable for very rare outcomesoutcomes
Historical cohort design requires complete and Historical cohort design requires complete and accurate medical recordsaccurate medical recordsaccurate medical recordsaccurate medical records..
Long followLong follow--up time required can result in up time required can result in loss to loss to followfollow--up (or drop out) up (or drop out) of some subjects in the studyof some subjects in the study
41PubH 6414 Lesson 1
Experimental StudiesExperimental Studies Experimental Studies involve an Experimental Studies involve an interventionintervention by by
the investigator. Some examples of interventions the investigator. Some examples of interventions areare::
New drugNew drug New surgical procedureNew surgical procedure
Ed cation / co nseling programEd cation / co nseling program Education / counseling programEducation / counseling program Exercise Exercise programprogram
42PubH 6414 Lesson 1
Experimental StudiesExperimental Studies Clinical Clinical Trials Trials are experimental studies of a are experimental studies of a
medical treatment or procedure that involve medical treatment or procedure that involve human subjects. human subjects.
Two types of clinical trials: those with control Two types of clinical trials: those with control subjects and those without control subjects. subjects and those without control subjects.
43PubH 6414 Lesson 1
Controlled Clinical TrialsControlled Clinical TrialsControlled Clinical TrialsControlled Clinical Trials
Control subjects in a clinical trial provide aControl subjects in a clinical trial provide a Control subjects in a clinical trial provide a Control subjects in a clinical trial provide a comparison group for the subjects who comparison group for the subjects who receive the interventionreceive the intervention
Control subjects receive one of the Control subjects receive one of the followingfollowinggg No interventionNo intervention Placebo Placebo Standard treatment (if one exists)Standard treatment (if one exists) Intervention at a later timeIntervention at a later time
44PubH 6414 Lesson 1
Types of Controlled TrialsTypes of Controlled Trials Concurrently assign study subjects to two groups: an Concurrently assign study subjects to two groups: an
intervention group and a control group and follow both over intervention group and a control group and follow both over the same period of time to observe the same period of time to observe outcomesoutcomespp
SelfSelf--controlled study design: Each study subject is controlled study design: Each study subject is observed before and after the intervention so each subjectobserved before and after the intervention so each subjectobserved before and after the intervention so each subject observed before and after the intervention so each subject is their own controlis their own control..
CrossCross--over study design: subjects are assigned to one of over study design: subjects are assigned to one of two groupstwo groups Intervention first followed by the placeboIntervention first followed by the placebo Intervention first followed by the placeboIntervention first followed by the placebo Placebo first followed by the interventionPlacebo first followed by the intervention
45PubH 6414 Lesson 1
Randomized Controlled Clinical Trials Randomized Controlled Clinical Trials (RCCT)(RCCT)(RCCT)(RCCT)
In a RCCT Subjects are In a RCCT Subjects are randomlyrandomly assigned to the assigned to the jj yy ggintervention group or the control group intervention group or the control group Each subject has equal probability of being in control or Each subject has equal probability of being in control or
treatment grouptreatment group Reduce selection bias and produce groups that have Reduce selection bias and produce groups that have
comparable baseline characteristics.comparable baseline characteristics. These trials provide the strongest evidence for a These trials provide the strongest evidence for a p gp g
causecause--effect relationship between the intervention and effect relationship between the intervention and the outcome. the outcome.
Drawbacks of randomized controlled clinical trialsDrawbacks of randomized controlled clinical trials Drawbacks of randomized controlled clinical trials Drawbacks of randomized controlled clinical trials TimeTime--consuming consuming ExpensiveExpensive
46PubH 6414 Lesson 1
Randomized Controlled Clinical Randomized Controlled Clinical TTTrial ExampleTrial Example
MRFIT: Multiple Risk Factor Intervention TrialMRFIT: Multiple Risk Factor Intervention Trialpp 12,866 men with RF for CHD randomly assigned to12,866 men with RF for CHD randomly assigned to
Intervention: counseling to reduce risk factorsIntervention: counseling to reduce risk factors Control: usual physician careControl: usual physician care Control: usual physician careControl: usual physician care
Endpoint: death from CHDEndpoint: death from CHD 66--8 years of follow8 years of follow--upup66 8 years of follow8 years of follow upup Study resultsStudy results
Risk factor levels declined in both groups but slightly Risk factor levels declined in both groups but slightly more in the special intervention groupmore in the special intervention groupmore in the special intervention group.more in the special intervention group.
Mortality from CHD and from all causes was not Mortality from CHD and from all causes was not significantly different between the two groups. (JAMA, significantly different between the two groups. (JAMA, 1982; 248:14651982; 248:1465--77)77)
47PubH 6414 Lesson 1
1982; 248:14651982; 248:1465--77) 77)
Blinded studiesBlinded studiesBlinded studiesBlinded studies
SingleSingle--blindblindgg Subjects do not know whether they are in treatment Subjects do not know whether they are in treatment
or control groupor control group DoubleDouble--blindblind DoubleDouble blindblind
Subjects Subjects andand investigators are not aware of treatment investigators are not aware of treatment assignmentassignment
TripleTriple blindblind TripleTriple--blindblind Subjects, investigators Subjects, investigators andand statistician do not know statistician do not know
treatment assignmenttreatment assignment Goals of blinding: to promote objective Goals of blinding: to promote objective
assessment of study results and reduce biasassessment of study results and reduce bias Placebos are used to keep the study blindedPlacebos are used to keep the study blinded
48PubH 6414 Lesson 1
Placebos are used to keep the study blindedPlacebos are used to keep the study blinded
Random AssignmentRandom AssignmentRandom AssignmentRandom AssignmentObservational studies use random Observational studies use random samplingsampling but but
controlled clinical trials use randomcontrolled clinical trials use random assignmentassignmentcontrolled clinical trials use random controlled clinical trials use random assignmentassignment In a randomized clinical trial, subjects are In a randomized clinical trial, subjects are
selectedselected based on inclusion and exclusion based on inclusion and exclusion criteriacriteria The selection process is not necessarily randomThe selection process is not necessarily random
However selected subjects areHowever selected subjects are randomlyrandomly However, selected subjects are However, selected subjects are randomly randomly assigned to either a treatment or control groupassigned to either a treatment or control group Goal is to have comparable groups in all Goal is to have comparable groups in all
characteristicscharacteristics exceptexcept the treatmentthe treatmentcharacteristics characteristics exceptexcept the treatmentthe treatment With large enough sample size, random assignment With large enough sample size, random assignment
usually results in comparable groupsusually results in comparable groups
49PubH 6414 Lesson 1
Bias in Clinical TrialsBias in Clinical TrialsBias in Clinical TrialsBias in Clinical Trials
The following describe possible sources of bias in The following describe possible sources of bias in g pg pa Clinical Triala Clinical Trial
Procedure BiasProcedure Bias Treatment group receives more attention than control Treatment group receives more attention than control
groupgroup Recall biasRecall bias Recall biasRecall bias
Subjects in one group are more likely to remember Subjects in one group are more likely to remember events than subjects in another groupevents than subjects in another group
Compliance biasCompliance bias Patients comply with one treatment more than Patients comply with one treatment more than
anotheranother
50PubH 6414 Lesson 1
anotheranother
Consider when reading medical Consider when reading medical literature:literature:
For an observational studyFor an observational study For an observational studyFor an observational study Is the study sample representative of the population Is the study sample representative of the population
of interest?of interest? What was the sampling method? If random sampling What was the sampling method? If random sampling
was used, statistical inference is valid.was used, statistical inference is valid. What are potential sources of bias?What are potential sources of bias? What are potential sources of bias?What are potential sources of bias?
For a clinical trialFor a clinical trial What are inclusion / exclusion criteria?What are inclusion / exclusion criteria? What are inclusion / exclusion criteria?What are inclusion / exclusion criteria? How were random assignments made?How were random assignments made? What are potential sources of bias?What are potential sources of bias?
51PubH 6414 Lesson 1
What are potential sources of bias?What are potential sources of bias?
Problem 1Problem 1Problem 1Problem 1
TwentyTwenty--five male volunteers, ages 18 to 24five male volunteers, ages 18 to 24TwentyTwenty five male volunteers, ages 18 to 24 five male volunteers, ages 18 to 24 years, participated in a study to compare years, participated in a study to compare effects of small doses of melatonin and a effects of small doses of melatonin and a l b i d l b d il b i d l b d iplacebo to induce sleep at bedtime. placebo to induce sleep at bedtime.
Subjects who took melatonin fell asleep in 5 Subjects who took melatonin fell asleep in 5 6 i t l t h i d th6 i t l t h i d thor 6 minutes; volunteers who received the or 6 minutes; volunteers who received the
placebo took 15 minutes or more to fall placebo took 15 minutes or more to fall asleep The researchers concluded thatasleep The researchers concluded thatasleep. The researchers concluded that asleep. The researchers concluded that melatonin in small doses may be an effective melatonin in small doses may be an effective sleep aid. sleep aid.
Eberly and Telke 52PubH 6450
1. What is the sample?p
1.1. TwentyTwenty--five male five male volunteersvolunteers
2.2. Male who ages 18 Male who ages 18 ggto 24 yearsto 24 years
33 All malesAll males3.3. All malesAll males
53PubH 6450 Eberly and Telke
2. The interested population is ____?p p
1.1. TwentyTwenty--five male five male volunteersvolunteers
2.2. Male who ages 18 Male who ages 18 ggto 24 yearsto 24 years
33 All malesAll males3.3. All malesAll males
Eberly and Telke 54PubH 6450
3. What is the outcome of interest?
11 The age ofThe age of1.1. The age of The age of volunteersvolunteers
22 The time to fallThe time to fall2.2. The time to fall The time to fall asleepasleepTh d fTh d f3.3. The doses of The doses of melatoninmelatonin
4.4. The doses of a The doses of a placeboplacebo
Eberly and Telke 55PubH 6450
4. What is the study design?y g
1.1. CrossCross--sectionalsectional2.2. CaseCase--controlcontrol33 CohortCohort3.3. CohortCohort4.4. Clinical trialClinical trial
Eberly and Telke 56PubH 6450
Problem 2Problem 2Problem 2Problem 2 Oncologists at the National Cancer Center Institute Oncologists at the National Cancer Center Institute
studied the effects of interleukinstudied the effects of interleukin--2 (IL2 (IL--2) an immune2) an immunestudied the effects of interleukinstudied the effects of interleukin--2 (IL2 (IL--2), an immune 2), an immune system protein, in patients with spreading melanoma or system protein, in patients with spreading melanoma or unresponsive, spreading kidney cell cancer. Their goal unresponsive, spreading kidney cell cancer. Their goal was to determine if the use of ILwas to determine if the use of IL--2 spurs the patient’s2 spurs the patient’swas to determine if the use of ILwas to determine if the use of IL 2 spurs the patient s 2 spurs the patient s immune system to fight the cancer and halt its spread. immune system to fight the cancer and halt its spread.
300 patients are randomized to treatment or placebo, 20 300 patients are randomized to treatment or placebo, 20 in the treatment group experienced total remission andin the treatment group experienced total remission andin the treatment group experienced total remission and in the treatment group experienced total remission and remained in remission for at least 8 years. 5 in the remained in remission for at least 8 years. 5 in the placebo group experienced total remission and remained placebo group experienced total remission and remained in remission for at least 8 years.. The investigators in remission for at least 8 years.. The investigators y gy gconcluded the use of ILconcluded the use of IL--2 marks a significant advance in 2 marks a significant advance in the treatment of these two types of cancer. the treatment of these two types of cancer.
Eberly and Telke 57PubH 6450
1. The 300 patients in this study are the ?the___?
1.1. PopulationPopulationS lS l2.2. SampleSample
3.3. Selected to Selected to treatment or treatment or placebo groups by placebo groups by someonesomeone
Eberly and Telke 58PubH 6450
2. This study is ____?y
1.1. Retrospective and Retrospective and observationalobservational
2.2. Retrospective and Retrospective and experimentalexperimental
3.3. Prospective and Prospective and experimentalexperimentalexperimentalexperimental
4.4. Prospective and Prospective and observationalobservational
Eberly and Telke 59
observationalobservational
PubH 6450
3. What is the study design?y g
1.1. CrossCross--sectionalsectional2.2. CaseCase--controlcontrol33 CohortCohort3.3. CohortCohort4.4. Clinical trialClinical trial
Eberly and Telke 60PubH 6450
Problem 3Problem 3Problem 3Problem 3
A study was conducted relating to the epidemiology A study was conducted relating to the epidemiology y g p gyy g p gyof breast cancer and the possible involvement of of breast cancer and the possible involvement of dietary fats, vitamins and other nutrients. dietary fats, vitamins and other nutrients.
It included 2024 breast cancer patients admitted toIt included 2024 breast cancer patients admitted to It included 2024 breast cancer patients admitted to It included 2024 breast cancer patients admitted to Roswell Park Memorial Institute, Erie County, New Roswell Park Memorial Institute, Erie County, New York, from 1958 to 1965. Another group of 1463 York, from 1958 to 1965. Another group of 1463 people was chosen from patients having no people was chosen from patients having no neoplasmsneoplasms and no pathology of gastrointestinal or and no pathology of gastrointestinal or reproductive systems The primary factors beingreproductive systems The primary factors beingreproductive systems. The primary factors being reproductive systems. The primary factors being investigated were vitamins A and E (measured in investigated were vitamins A and E (measured in international units per month). international units per month).
Eberly and Telke 61PubH 6450
1. What is the exposure of this study?p
11 Vitamins A and EVitamins A and E1.1. Vitamins A and EVitamins A and E2.2. Whether patients Whether patients
get breast cancerget breast cancerget breast cancerget breast cancer3.3. Whether patients Whether patients
hh llhave have neoplasmsneoplasms or or pathology of pathology of
t i t ti lt i t ti lgastrointestinal or gastrointestinal or reproductive reproductive
ttsystemssystemsEberly and Telke 62PubH 6450
2. This study is ____?y
1.1. Retrospective and Retrospective and observationalobservational
2.2. Retrospective and Retrospective and experimentalexperimental
3.3. Prospective and Prospective and observationalobservationalobservationalobservational
4.4. Prospective and Prospective and experimentalexperimental
Eberly and Telke 63PubH 6450
experimentalexperimental
3. What is the study design?y g
1.1. CrossCross--sectionalsectional2.2. CaseCase--controlcontrol33 CohortCohort3.3. CohortCohort4.4. Clinical trialClinical trial
Eberly and Telke 64PubH 6450
What type of data wereWhat type of data wereWhat type of data were What type of data were collected?collected?
Measurement ScalesMeasurement Scales
Measurement ScalesMeasurement ScalesMeasurement ScalesMeasurement ScalesThe two broad categories of data are Categorical The two broad categories of data are Categorical
and Numericaland Numericaland Numericaland Numerical Categorical dataCategorical data
Also called ‘Qualitative’Also called ‘Qualitative’ Examples: gender, blood type, race, disease status, Examples: gender, blood type, race, disease status,
etc.etc. Data consist of the counts in each categoryData consist of the counts in each categoryData consist of the counts in each category Data consist of the counts in each category
Numerical dataNumerical data Also called ‘Quantitative’Also called ‘Quantitative’
Diff b t th b h iDiff b t th b h i Differences between the numbers have meaning on a Differences between the numbers have meaning on a numerical scalenumerical scale
Examples: weight, height, survival time, blood Examples: weight, height, survival time, blood pressurepressure
66PubH 6414 Lesson 1
pressurepressure
Types of Categorical DataTypes of Categorical DataTypes of Categorical DataTypes of Categorical Data Nominal Nominal –– two or more categories with no natural two or more categories with no natural
ddorder order Examples: Gender, Race, Blood group Examples: Gender, Race, Blood group
O di lO di l t t i th t ht t i th t h OrdinalOrdinal –– two or more categories that have a two or more categories that have a natural ordernatural order Ex Apgar scores tumor stage Social classEx Apgar scores tumor stage Social class Ex. Apgar scores, tumor stage, Social classEx. Apgar scores, tumor stage, Social class Numerical assignments are relative (> or <) and do not Numerical assignments are relative (> or <) and do not
represent any interval relationship between categories represent any interval relationship between categories BinaryBinary –– a special case of either Ordinal or a special case of either Ordinal or
Nominal dataNominal data
67PubH 6414 Lesson 1
Binary data is categorical data with only two categoriesBinary data is categorical data with only two categories
Contingency TablesContingency TablesContingency TablesContingency TablesData for two categorical characteristics are often Data for two categorical characteristics are often
displayed in contingency tables. The cell counts are displayed in contingency tables. The cell counts are p y g yp y g ythe number of subjects with each combination of the number of subjects with each combination of the characteristics.the characteristics.
Lung CancerLung Cancer No Lung No Lung CancerCancer
SmokerSmoker 8585 1515
ExEx--smokersmoker 1515 1010ExEx smokersmoker 1515 1010
NonNon--smokersmoker 00 7575
68PubH 6414 Lesson 1
Numerical ScalesNumerical ScalesNumerical ScalesNumerical Scales
Continuous scaleContinuous scale –– data that can take on any data that can take on any real value on the measurement scalereal value on the measurement scale Blood pressure, temperature, age, weight, height etc. Blood pressure, temperature, age, weight, height etc.
can all be measured on a continuumcan all be measured on a continuumcan all be measured on a continuumcan all be measured on a continuum Discrete scalesDiscrete scales –– data that are only integersdata that are only integers
Number of children in a family number of births in aNumber of children in a family number of births in a Number of children in a family, number of births in a Number of children in a family, number of births in a year, number of accidents in a month etc. year, number of accidents in a month etc.
Arithmetic operations are meaningful forArithmetic operations are meaningful forArithmetic operations are meaningful for Arithmetic operations are meaningful for numerical data numerical data
69PubH 6414 Lesson 1
Diagram of Measurement ScalesDiagram of Measurement ScalesDiagram of Measurement ScalesDiagram of Measurement Scales
Categorical Numerical
Ordinal Nominal Continuous DiscreteOrdinal Nominal Continuous Discrete
scale not ordered any value integerson the scale
70PubH 6414 Lesson 1
Q i 1Question 1:Is this variable qualitative or quantitative?
Infant birth weight (grams)Infant birth weight (grams)
Q lit tiQ lit ti1.1. QualitativeQualitative2.2. QuantitativeQuantitative
Q estion 2:Question 2:Is this variable categorical or continuous?
Household income Household income (b l l l b l)(b l l l b l)(below normal, normal, above normal)(below normal, normal, above normal)
1.1. CategoricalCategorical22 ContinuousContinuous2.2. ContinuousContinuous
Q ti n 3Question 3:Is this variable qualitative or quantitative?
Number of children in the familyNumber of children in the family
1.1. QualitativeQualitative2.2. Quantitative (Discrete) Quantitative (Discrete) 3.3. Quantitative (Continuous)Quantitative (Continuous)( )( )4.4. None of these.None of these.
Question 4:Is this variable quantitative?
Social Security number
11 TrueTrue1.1. TrueTrue2.2. FalseFalse
Question 5:Question 5:Rate your instructor’s teaching ability.
1=Poor, 2=Fair, 3=Good, 4=Very Good, 5=Excellent
Is this variable categorical or continuous?Is this variable categorical or continuous?
11 CategoricalCategorical1.1. CategoricalCategorical2.2. ContinuousContinuous3.3. EitherEither4.4. NeitherNeither
Question 6Question 6Rate your instructor’s teaching ability.
1=Poor, 2=Fair, 3=Good, 4=Very Good, 5=Excellent
Is this variable Nominal Ordinal or Binary?Is this variable Nominal, Ordinal or Binary?
11 NominalNominal1.1. NominalNominal2.2. OrdinalOrdinal3.3. BinaryBinary4.4. None of these.None of these.
Question 7OUTCOME WHEN TOSSING AOUTCOME WHEN TOSSING A
COIN.Heads or tailsHeads or tails
Is this variable Nominal OrdinalIs this variable Nominal, Ordinal or Binary?
11 NominalNominal1.1. NominalNominal2.2. OrdinalOrdinal3.3. BinaryBinary4.4. None of these.None of these.
Question 8Number of heads when tossing a
coin 30 times.coin 30 times.
I hi i bl Bi ?Is this variable Binary?
11 NoNo1.1. NoNo2.2. YesYes
Question 8Number of heads when tossing a
coin 30 times.coin 30 times.
Wh i hi i bl ?What is this variable?
11 Quantitative (Discrete)Quantitative (Discrete)1.1. Quantitative (Discrete)Quantitative (Discrete)2.2. Quantitative (Continuous)Quantitative (Continuous)3.3. NominalNominal4.4. Categorical BinaryCategorical Binaryg yg y
Relationships between Numerical Relationships between Numerical and Categorical Dataand Categorical Data
Numerical data can be converted to categorical Numerical data can be converted to categorical ggdatadata For example, age is a numerical continuous variable For example, age is a numerical continuous variable
but it can be summarized as ‘age groups’: 0but it can be summarized as ‘age groups’: 0--5, 65, 6--10, 10, g g pg g p ,, ,,1111--19, 2019, 20--39, 4039, 40--59, 6059, 60--79, 80+79, 80+
When represented as ‘age groups’, age is a When represented as ‘age groups’, age is a categorical (ordinal) variablecategorical (ordinal) variableg ( )g ( )
Ordinal data can be coded numerically Ordinal data can be coded numerically The numerical coding of ordinal data does not mean The numerical coding of ordinal data does not mean
that the ordinal data are now numerical howeverthat the ordinal data are now numerical howeverthat the ordinal data are now numerical, howeverthat the ordinal data are now numerical, however The codes represent the The codes represent the order order of the categories, not of the categories, not
the interval relationship between categoriesthe interval relationship between categories
80PubH 6414 Lesson 1
Next topics: Summary Statistics and Next topics: Summary Statistics and graphsgraphs
For Numerical DataFor Numerical Data Statistics: Mean, standard deviation, median, Statistics: Mean, standard deviation, median,
range, percentilerange, percentile Tables and graphs: frequency distribution, Tables and graphs: frequency distribution,
histograms, boxhistograms, box--plots, scatterplots, scatter--plotsplotsF C t i l D tF C t i l D t For Categorical DataFor Categorical Data Statistics: proportion, percent, rates, relative Statistics: proportion, percent, rates, relative
i k dd ii k dd irisk, odds ratiorisk, odds ratio Tables and graphs: bar charts, pie charts, Tables and graphs: bar charts, pie charts,
ti t blti t bl81PubH 6414 Lesson 1
contingency tablescontingency tables
Readings and AssignmentsReadings and AssignmentsReadings and AssignmentsReadings and Assignments
Readings:Readings: Chapter 2: Study Designs in Medical ResearchChapter 2: Study Designs in Medical Research Chapter 3: Scales of Measurement pgs 26Chapter 3: Scales of Measurement pgs 26--2727 Chapter 4: Pages 68Chapter 4: Pages 68--7272 Chapter 4: Pages 68Chapter 4: Pages 68--7272 Optional: Chapter 13 pgs 332 Optional: Chapter 13 pgs 332 -- 335335
Practice Exercises for Lesson 1Practice Exercises for Lesson 1 Try Intro to Excel Module on your own:Try Intro to Excel Module on your own:http://www.biostat.umn.edu/~http://www.biostat.umn.edu/~susant/FALL09PH6414LAB.htmlsusant/FALL09PH6414LAB.html
St t H k 1 P bl 1 d 2St t H k 1 P bl 1 d 2 Start Homework 1: Problems 1 and 2Start Homework 1: Problems 1 and 2 New vocabulary: look up any unfamiliar terms in New vocabulary: look up any unfamiliar terms in
thethe Pocket Dictionary of StatisticsPocket Dictionary of Statistics or the textor the text
82PubH 6414 Lesson 1
the the Pocket Dictionary of Statistics Pocket Dictionary of Statistics or the text or the text glossaryglossary