.com
Study Designs
n How Research Is Classified
n Terminology
n Important epidemiologic
concepts
Descriptive Statistics
n Measures of central
tendency
n Measures of spread
n Measures of frequency of
events
n Measures of Association
n Terms used to describe the
quality of measurements
n Measures of diagnostic test
accuracy
n Expressions used when
making inferences about
data
n Multivariable Regression
Methods
References
MedPage Tools
Guide to Biostatistics
Here is a compilation of important epidemiologic concepts and
common biostatistical terms used in medical research. You can
use it as a reference guide when reading articles published on
MedPage Today or download it to keep near the reading stand
where you keep your print journals. For more detailed infor-
mation on these topics, use the reference list at the end of this
presentation.
Study Designs in Clinical Research
Cohort
study
Cross-sectional
study
Case control
study
Yes
No
No
Exposure and
Outcome at
the same time
Did
researcher
assign
exposures?
Is there a
Comparison
group?
Observational
study
Exposure
n
Outcome
Exposure
l
Outcome
Direction of
the study?
Descriptive
Study
Analytical
Study
No
Experimental
study
Yes
Yes
Non-Randomised
controlled Trial
Randomised
controlled
Is
allocation
Random?
How research is classified
1
.com
Study Designs
n How Research Is Classified
n Terminology
n Important epidemiologic
concepts
Descriptive Statistics
n Measures of central
tendency
n Measures of spread
n Measures of frequency of
events
n Measures of Association
n Terms used to describe the
quality of measurements
n Measures of diagnostic test
accuracy
n Expressions used when
making inferences about
data
n Multivariable Regression
Methods
References
Terminology
Clinical Trial Experimental study in which the exposure status
(e.g. assigned to active drug versus placebo) is determined by
the investigator.
Randomized Controlled Trial A special type of clinical trial
in which assignment to an exposure is determined purely by
chance.
Cohort Study Observational study in which subjects with an
exposure of interest (e.g. hypertension) and subjects without
the exposure are identified and then followed forward in time to
determine outcomes (e.g. stroke).
Case-Control Study Observational study that first identifies a
group of subjects with a certain disease and a control group
without the disease, and then looks to back in time (e.g. chart
review) to find exposure to risk factors for the disease. This type
of study is well suited for rare diseases.
Cross-Sectional Study Observational study that is done to ex-
amine presence or absence of a disease or presence or absence
of an exposure at a particular time. Since exposure and outcome
are ascertained at the same time, it is often unclear if the expo-
sure preceded the outcome.
Case Report or Case Series Descriptive study that reports on
a single or a series of patients with a certain disease. This type
of study usually generates a hypothesis but cannot test a hy-
pothesis because it does not include an appropriate comparison
group.
Important Epidemiologic Concepts
Bias Any systematic error in the design or conduct of a study
that results in a mistaken estimate of an exposure’s effect on risk
of disease.
Selection Bias Bias introduced by the way in which participants
are chosen for a study. For example, in a case-control study using
different criteria to select cases (e.g. sick, hospitalized population)
versus controls (young, healthy outpatients) other than the pres-
ence of disease can lead the investigator to a false conclusion
about an exposure.
Confounding This occurs when an investigator falsely concludes
that a particular exposure is causally related to a disease without
adjusting for other factors that are known risk factors for the
disease and are associated with the exposure.
.com
Study Designs
n How Research Is Classified
n Terminology
n Important epidemiologic
concepts
Descriptive Statistics
n Measures of central
tendency
n Measures of spread
n Measures of frequency of
events
n Measures of Association
n Terms used to describe the
quality of measurements
n Measures of diagnostic test
accuracy
n Expressions used when
making inferences about
data
n Multivariable Regression
Methods
References
Descriptive Statistics
Measures of Central Tendency
Mean equals the sum of observations divided by the number of
observations.
Median equals the observation in the center when all observa-
tions are ordered from smallest to largest; when there is an even
number of observations the median is defined as the average of
the middle two values.
Mode equals the most frequently occurring value among all
observations.
Measures of Spread
Spread (or variability) describes the manner in which data are
scattered around a specific value (such as the mean). The most
commonly used measures of spread are:
Range is the difference between the largest observation and
the smallest.
Standard Deviation measures the spread of data around the
mean. One standard deviation includes 68% of the values in a
sample population and two standard deviations include 95% of
the values.
Standard Error of the Mean describes the amount of variability
in the measurement of the population mean from several differ-
ent samples. This is in contrast to the standard deviation which
measures the variability of individual observations in a sample.
Percentile equals the percentage of a distribution that is below
a specific value. As an example, a child is in the 80th percentile
for height if only 20% of children of the same age are taller than
he is.
Interquartile Range refers to the upper and lower bound-
ary defining the middle 50 percent of observations. The upper
boundary is the 75th percentile and the lower boundary is the
25th percentile.
Measures of Frequency of Events
Incidence The number of new events (e.g. death or a particular
disease) that occur during a specified period of time in a popula-
tion at risk for developing the events.
Incidence Rate A term related to incidence that reports the
number of new events that occur over the sum of time indi-
viduals in the population were at risk for having the event (e.g.
.com
Study Designs
n How Research Is Classified
n Terminology
n Important epidemiologic
concepts
Descriptive Statistics
n Measures of central
tendency
n Measures of spread
n Measures of frequency of
events
n Measures of Association
n Terms used to describe the
quality of measurements
n Measures of diagnostic test
accuracy
n Expressions used when
making inferences about
data
n Multivariable Regression
Methods
References
events/person-years).
Prevalence The number of persons in the population affected by
a disease at a specific time divided by the number of persons in
the population at the time.
Measures of Association
The types of measures used to define the association between
exposures and outcome depends upon the type of data. For
categorical variables, the relative risk and odds ratio are com-
monly used to describe the relationship between exposures and
outcome.
Relative risk and cohort studies The relative risk (or risk ratio)
is defined as the ratio of the incidence of disease in the exposed
group divided by the corresponding incidence of disease in the
unexposed group (Figure 2). Relative risk can be calculated in co-
hort studies such as the Framingham Heart Study where subjects
with certain exposures (e.g. hypertension, hyperlipidemia) were
followed prospectively for cardiovascular outcomes. The inci-
dence of cardiac events in subjects with and without exposures
was then used to calculate relative risk and determine whether
exposures were cardiac risk factors.
Odds ratio and case-control studies The odds ratio is defined
as the odds of exposure in the group with disease divided by the
odds of exposure in the control group (Figure 1). As described
above, subjects are selected on the basis of disease status in
case-control studies, therefore it is not possible to calculate the
rate of development of disease given presence or absence of
exposure. So, the odds ratio is often used to approximate the
A/(A+B)
C/(C+D)
Relative Risk
A B
C D
+
Disease
Test
Yes No
A/C A×D
B/D B×C
=
Odds Ratio
A B
C D
Disease
Yes No
Cohort Study
Case Control Study
Figure 1: In a case-control study, the odds ratio can be used to approximate the
relative risk under the assumption that the disease is rare.
.com
Study Designs
n How Research Is Classified
n Terminology
n Important epidemiologic
concepts
Descriptive Statistics
n Measures of central
tendency
n Measures of spread
n Measures of frequency of
events
n Measures of Association
n Terms used to describe the
quality of measurements
n Measures of diagnostic test
accuracy
n Expressions used when
making inferences about
data
n Multivariable Regression
Methods
References
relative risk in case-control studies. For example, a case-control
study was done to evaluate the relationship between artificial
sweeteners and bladder cancer. The odds of artificial sweetener
use in the cases and controls were used to calculate an odds
ratio and determine whether sweeteners were associated with
bladder cancer. Under the assumption that the disease under
consideration is rare (e.g. bladder cancer), the odds ratio gives a
stable, unbiased estimate of the relative risk (Figure 1). The odds
ratio from a case-control study nested within a defined cohort
also approximates the relative risk even when the rare disease
assumption is not held.
If the disease is rare, A<<B and C<<D. So, A/(A + B) is approximated
by A/B and C/(C + D) approximated by C/D. In this situation, the
relative risk equals (A/B)/(C/D) which, rearranged, equals the odds
ratio A×D/B×C
Absolute risk The relative risk and odds ratio provide a measure
of risk compared with a standard. However, it is sometimes desir-
able to know the absolute risk. For example, a 40% increase in
risk of heart disease because of a particular exposure does not
provide insight into the likelihood that exposure in an individual
patient will lead to heart disease.
The Attributable risk or Risk difference is a measure of abso-
lute risk. It represents the excess risk of disease in those exposed
taking into account the background rate of disease. The attribut-
able risk is defined as the difference between the incidence rates
in the exposed and non-exposed groups.
A related term, the Population Attributable Risk is used to de-
scribe the excess rate of disease in the total study population of
exposed and non-exposed individuals that is attributable to the
exposure. This measure is calculated by multiplying the Attributable
risk by the proportion of exposed individuals in the population.
Number needed to treat (NNT) The number of patients who
would need to be treated to prevent one adverse outcome is
often used to present the results of randomized trials. NNT is the
reciprocal of the absolute risk reduction (the absolute adverse
event rate for placebo minus the absolute adverse event rate for
treated patients). This approach can be used in studies of vari-
ous interventions including both treatment and prevention. The
estimate for NNT is subject to considerable error and is generally
presented with 95% confidence intervals so that it can be prop-
erly interpreted.
Terms Used To Describe The Quality Of Measurements
Reliability The concept of reliability or reproducibility is related
to the amount of error in any measurement (e.g. blood pressure
.com
Study Designs
n How Research Is Classified
n Terminology
n Important epidemiologic
concepts
Descriptive Statistics
n Measures of central
tendency
n Measures of spread
n Measures of frequency of
events
n Measures of Association
n Terms used to describe the
quality of measurements
n Measures of diagnostic test
accuracy
n Expressions used when
making inferences about
data
n Multivariable Regression
Methods
References
measurement). A more formal definition of reliability is variability
between subjects divided by inter-subject variability plus mea-
surement error. Thus, reliability is greater when measurement er-
ror is minimal. There are several types of reliability including: inter
and intra-observer reliability and test-retest reliability.
Percent agreement and the kappa statistic are often used to re-
port reliability. The kappa statistic takes into account agreement
that would be seen by chance alone while percent agreement
does not. Generally, a kappa greater than 0.75 represents excel-
lent agreement beyond chance, a kappa below 0.40 represents
poor agreement and a kappa of 0.40-0.75 represents intermedi-
ate to good agreement.
Validity refers to the extent to which a test or surrogate is mea-
suring what we think it is measuring. There are several types of
validity that can be measured including content validity (the
extent to which the measure reflects the dimensions of a particu-
lar problem), construct validity (the extent to which a measure
conforms to an external established phenomenon), and criterion
validity (the extent to which a measure correlates with a gold
standard or can predict an observable phenomenon). These
types of validity are often applied to questionnaires in which the
truth is not physically verifiable.
Measures Of Diagnostic Test Accuracy
Sensitivity is defined as the ability of the test to identify cor-
rectly those who have the disease. It is the number of subjects
with a positive test who have disease divided by all subjects who
have the disease. A test with high sensitivity has few false nega-
tive results.
Specificity is defined as the ability of the test to identify cor-
rectly those who do not have the disease. It is the number of
subjects who have a negative test and do not have the disease
divided by the number of subjects who do not have the disease.
A test with high specificity has few false positive results.
Sensitivity and specificity are test characteristics that are most
useful when assessing a test used to screen a free-living popula-
tion. These test characteristics are also interdependent: an in-
crease in sensitivity is accompanied by a decrease in specificity
and visa versa. This is illustrated best by continuous tests where
the cut-off for a positive test result can be varied. For example,
consider the use of the white blood cell (WBC) count as a test to
diagnose bacterial infection. If one sets a high cut-off for a posi-
tive test (e.g. WBC> 25,000) then the test will have a low sensitiv-
ity and high specificity compared to the test characteristics if the
cut-off is lower (e.g. WBC>10,000).
.com
Study Designs
n How Research Is Classified
n Terminology
n Important epidemiologic
concepts
Descriptive Statistics
n Measures of central
tendency
n Measures of spread
n Measures of frequency of
events
n Measures of Association
n Terms used to describe the
quality of measurements
n Measures of diagnostic test
accuracy
n Expressions used when
making inferences about
data
n Multivariable Regression
Methods
References
Predictive values are important for assessing how useful a test
will be in the clinical setting at the individual patient level. The
positive predictive value is the probability of disease in a pa-
tient with a positive test. Conversely, the negative predictive
value is the probability that the patient does not have disease if
he has a negative test result.
Predictive values depend on the prevalence of a disease in a
population. A test with a given sensitivity and specificity can
have different predictive values in different patient populations.
If the test is used in a population with a high prevalence, it will
have a high positive predictive value and the same test will have
a low positive predictive value when used in a population with
low disease prevalence. For example, a positive stool test for oc-
cult blood is much more likely to be predictive of colon cancer in
a population of elderly people compared with twenty year olds.
Likelihood ratios Calculating likelihood ratios is another meth-
od of assessing the accuracy of a test in the clinical setting. Likeli-
hood ratios also offer the advantage of being independent of
disease prevalence.
The likelihood ratio indicates how much a given diagnostic test
result will raise or lower the odds of having a disease relative to
the prior probability of disease. Each diagnostic test is character-
ized by two likelihood ratios: a positive likelihood ratio that tells
us the odds of disease if the test result is positive and a negative
likelihood ratio that tells us the odds of disease if the test result is
negative:
LR+ = Sensitivity / (1- Specificity)
LR- = (1- Sensitivity) / Specificity
A likelihood ratio greater than 1 increases the odds that the per-
Sensitivity
A/(A+C)
Specificity
D/(B+D)
Positive Predictive Value
A/(A+B)
Negative Predictive Value
D/(C+D)
A B
C D
+
-
Disease
Test
Yes No
Figure 2: Calculating sensitivity, specificity, and predictive values
.com
Study Designs
n How Research Is Classified
n Terminology
n Important epidemiologic
concepts
Descriptive Statistics
n Measures of central
tendency
n Measures of spread
n Measures of frequency of
events
n Measures of Association
n Terms used to describe the
quality of measurements
n Measures of diagnostic test
accuracy
n Expressions used when
making inferences about
data
n Multivariable Regression
Methods
References
son has the target disease, and the higher the LR the greater this
increase in odds. Conversely, a likelihood ratio less than 1 dimin-
ishes the odds that the patient has the target disease.
Expressions Used When Making Inferences About Data
Confidence Intervals The results of any study sample are an
estimate of the true value in the entire population. The true value
may actually be greater or less than what is observed. A confi-
dence interval gives a range of values within which there is a high
probability (95% by convention) that the true population value
can be found. The confidence interval takes into consideration the
number of observations and the standard deviation in the sample
population. The confidence interval narrows as the number of
observations increases or standard deviation decreases.
Errors In hypothesis testing, there are two types of errors:
Type I error (alpha) is the probability of incorrectly concluding
there is a statistically significant difference in the population
when none exists. This type of error is also called alpha and is
the number after a P-value. A P<0.05 means that there is a less
than 5% chance that the difference could have occurred by
chance.
Type II error (beta) is the probability of incorrectly concluding
that there is no statistically significant difference in a popula-
tion when one exists.
Power is a measure of the ability of a study to detect a true dif-
ference. It is measured as 1- type II error rate or 1-beta. Every
researcher should perform a power calculation prior to carrying
out a study to determine the number of observations needed
to detect a desired degree of difference. Ideally this difference
should equal the smallest difference that would still be consid-
ered to be clinically important. However, the smaller the dif-
ference, the greater the number of observations needed. For
example, it takes fewer patients to observe a 50% reduction in
mortality from a new therapy than a 5% reduction.
Multivariable Regression Methods
In medical research, one is often interested in studying the inde-
pendent effect of multiple risk factors on outcome. For example,
we may want to know the independent effect of age, gender and
smoking status on the risk of having a myocardial infarction. Fur-
thermore, we may want to know if smoking raises the risk equally
in men and women. Multivariable regression methods allow us
to answer these types of questions by simultaneously account-
ing for multiple variables. The type of regression model used
depends on the type of outcome data being evaluated.
.com
Study Designs
n How Research Is Classified
n Terminology
n Important epidemiologic
concepts
Descriptive Statistics
n Measures of central
tendency
n Measures of spread
n Measures of frequency of
events
n Measures of Association
n Terms used to describe the
quality of measurements
n Measures of diagnostic test
accuracy
n Expressions used when
making inferences about
data
n Multivariable Regression
Methods
References
Multiple linear regression is used when the outcome data is
a continuous variable such as weight. For example, one could
estimate the effect of a diet on weight after adjusting for the ef-
fect of confounders such as smoking status. Another use of this
method is to predict a linear variable based on known variables.
Logistic regression is used when the outcome data is binary
such as cure or no cure. Logistic regression can be used to esti-
mate the effect of an exposure on a binary outcome after adjust-
ing for confounders. Logistic regression can also be used to find
factors that discriminate two groups or to find prognostic indica-
tors for a binary outcome. This method can also be applied to
case-control studies.
Survival Analysis
In survival analysis, one is commonly interested in the time until
some event such as the time from treatment of disease to death.
In the study population, only some subjects will have the event of
interest (e.g. death, stroke), others will have alternate events or no
events. The duration of follow-up will also vary among subjects
and it is important to account for the different follow-up times.
The Kaplan-Meier analysis and a regression method, the Cox pro-
portional hazards analysis are two methods of survival analysis that
account for inter-subject variation in events and follow-up time.
Kaplan-Meier analysis measures the ratio of surviving subjects
(or those without an event) divided by the total number of sub-
jects at risk for the event. Every time a subject has an event, the
ratio is recalculated. These ratios are then used to generate a
curve to graphically depict the probability of survival (Figure 3).
Percent Survival
Follow-up Time
Drug Group
100
0
Placebo Group
Figure 3: Kaplan-Meier Survival Curves
.com
Study Designs
n How Research Is Classified
n Terminology
n Important epidemiologic
concepts
Descriptive Statistics
n Measures of central
tendency
n Measures of spread
n Measures of frequency of
events
n Measures of Association
n Terms used to describe the
quality of measurements
n Measures of diagnostic test
accuracy
n Expressions used when
making inferences about
data
n Multivariable Regression
Methods
References
In studies with an intervention arm and a control arm, one can
generate two Kaplan-Meier curves. If the curves are close to-
gether or cross, a statistically significant difference is unlikely to
exist. Statistical tests such as the log-rank test can be used to
confirm the presence of a significant difference.
Cox proportional hazards analysis is similar to the logistic
regression method described above with the added advantage
that it accounts for time to a binary event in the outcome vari-
able. Thus, one can account for variation in follow-up time among
subjects. Like the other regression methods described above, it
can be used to study the effect of an exposure on outcome after
adjusting for confounders. Cox analysis can also be used to find
prognostic indicators for survival in a given disease.
The hazard ratio that results from this analysis can be interpreted
as a relative risk (risk ratio). For example, a hazard ratio of 5
means that the exposed group has five times the risk of having
the event compared to the unexposed group.
Rubeen K. Israni, M.D.
Fellow, Renal-Electrolyte and Hypertension Division,
University of Pennsylvania School of Medicine.
.com
Study Designs
How Research Is Classied
Terminology
Important epidemiologic
concepts
Descriptive Statistics
Measures of central
tendency
Measures of spread
Measures of frequency of
events
Measures of Association
Terms used to describe the
quality of measurements
Measures of diagnostic test
accuracy
Expressions used when
making inferences about
data
Multivariable Regression
Methods
References
References
1. Grimes DA, Schultz KF: An overview of clinical research: the lay of the land.
Lancet 359:57-61, 2002
2. Grimes DA, Schultz KF: Bias and causal associations in observational
research. Lancet 359:248-252, 2002.
3. Gordis L: Epidemiology, 3rd Edition, Philadelphia, Elsevier Saunders, 2004.
4. Rosner B: Fundamentals of Biostatistics, 4th Edition, Daxbury Press, 1995.
5. Grimes DA, Schultz KF: Cohort studies: marching towards outcomes. Lancet
359: 341-345, 2002.
6. Schultz KF, Grimes DA: Case-control studies: research in reverse. Lancet
359:431-434, 2002.
7. Streiner DL, Norman GR: Health Measu
rement Scales: A Practical Guide to
their Development and Use, 2nd Edition, New York, Oxford University Press,
2000.
8. Jaeschke R, Guyatt GH, Sackett DL: Users’ guides to the medical literature. III.
How to use an article about a diagnostic test. B. What are the results and
will they help me in caring for my patients? The Evidence-Based Medicine
Working Group. Jama 271:703-707, 1994.
9. Guyatt G, Jaeshke R, Heddle N, et al. Basic statistics for clinicians: 2.
Interpreting study results: confidence intervals. Cmaj 152:169-173, 1995.
10. Katz MH: Multivariable analysis: a primer for readers of medical research.
Ann Intern Med 138:644-650, 2003.
11. Campbell MJ: Statistics at Square Two, 4th Edition, London, BMJ Publishing
Group, 2004.
Copyright © 2007 MedPage Today, LLC. All Rights Reserved.
Source: http://www.medpagetoday.com/