Professional Documents
Culture Documents
Epidemiology
Epidemiology
1. What is epidemiology?
The word epidemiology consists of the Greek words epi = among, demos = people, and logos
= doctrine, and thus means the doctrine of what is among or happening to people.
Epidemiology is the study of the distribution and determinants of disease frequency in human
population (or in the group of population) These three closely interrelated components
encompass all epidemiological principles and methods. The measurement of disease
frequency involves quantification of the existence or occurrence of disease. The distribution of
the disease considers such questions as who is getting disease within a population, where and
when the disease is occurring. Knowledge of such distributions is essential to describe
patterns of disease as well as to formulate hypotheses concerning possible causal or
preventive factors. The determinants of disease, which can be derived from the knowledge of
frequency and distribution of the disease, are necessary to test the epidemiological
hypotheses. Epidemiological research is based on the systematic collection of observations
related to the phenomenon of interest in a defined population. These data then are subjected to
quantification, which includes the measurement of random variables, the estimation of
population parameters, and the statistical testing of hypotheses.
Epidemiologists study the occurrence of disease or other health-related conditions or events in
defined populations. Viewed as the study of the distribution and societal determinants of the
health status of populations, epidemiology is the basic-science foundation of public health.
Investigation of disease occurrence is not a new phenomenon. The development of
epidemiological theory and methods in recent decades, however, has opened new possibilities
and stimulated interest within many fields of application.
Sometimes the starting point for an epidemiological study is a certain characteristic or
exposure rather than a disease. There is a close connection between epidemiology and
preventive medicine. Prevention programs are rarely implemented for an entire population,
therefore, prevention programs may be planned to enable studies of the effect of' the
intervention on the disease frequency in the population by comparing disease rates among
those receiving the preventive program with rates among those who do not. In this way there is
usually an opportunity to evaluate preventive measures that have been undertaken, using
experimental epidemiology.
2
Epidemiological research has often provided information that has formed the basis for public
health decisions long before the basic mechanism of a particular disease was understood. In
recent years the value of information about disease distribution for planning the delivery of
health care has become more apparent. In several studies disease occurrence has been
related to health-care need, demand, and supply. There is also an increasing interest in
studying the effectiveness of the health-care system and/or of different treatments. The
common basis for these diff erent applications of epidemiology is the study of disease
occurrence and its relation to various characteristics of individuals or their environment.
3
2. Epidemiological observations
The quality of the data is commonly described with use of four terms:
accuracy
precision
reliability
validity
Accuracy is the degree to which a measurement represents the true value of the attribute being
measured. A measurement or observation can represent a true value without detail. As with
other measurements the accuracy of the study result depends on "validity" and "precision" of
the measure.
Validity is the extent to which the study measures what it is intended to measure; lack of validity
is referred to as "bias" or "systematic error." An important distinction must be made between
the terms internal validity and external validity. Internal validity concerns inferences about the
population of individuals of restricted interest from which a study sample has been drawn;
external validity concerns inferences about an external population beyond the study's restricted
interest.
Precision is the reproducibility of a study result, that is, the degree of resemblance among
study results, were the study to be repeated under similar circumstances: lack of precision is
referred to as "random error". A study that is based on information from too few subjects
allows for considerable random variation in the study result, since only a few extra cases
occurring in one or another category would substantially affect the results. Such a study would
have a low precision.
Reliability is a measure of how dependably an observation is exactly the same when repeated;
it refers to the measuring procedure rather than to the attribute being measured. Reliability is
not synonymous with repeatability or reproducibility; rather, it is a broader term that includes
the concept of consistency, which refers to how closely the findings in different samples or
populations conform to one another under different conditions or at different times.
4
3. Diagnosis of disease
Before a disease can be studied, it is necessary to define it. Often, this results in a case
definition, which may differ from a clinical definition. Studies in which different definitions are
used may lead to different conclusions.
In principle the variables that are the basis for diagnosis may depend upon:
subjective observations by the patient (symptoms)
subjective observations by the examiner (signs)
objective observations (tests)
Symptoms refers to manifestations that only the examined person (patient) may observe, e.g.,
insomnia, nausea, or fatigue. Of course, symptoms can be perceived and described differently
by different individuals and by the same individual in different situations.
Accuracy in recording symptoms is influenced by the instrument used to collect the data;
standardized interview and questionnaire methods have been developed to increase
reproducibility.
Signs refers to manifestations that may be observed by an examiner (usually a physician), for
example, weight loss, apathy or agitation. Ascertainment of signs is affected by the subjective
judgment of one (or several) examiner(s). The accuracy of such examinations is dependent on:
inter-observer variation (the degree of agreement among different examiners)
intra-observer variation (the degree of agreement between different examinations made by
one examiner).
In many cases reproducibility can be increased, and hence the importance of this source of
error in the diagnosis reduced, by standardization of the conditions under which the
observations are made. The examination routines can be standardized and sometimes
detailed and standardized criteria for classification of the observations can be formulated.
5
Tests will refer here to manifestations that can be read from an instrument and hence are less
dependent on subjective judgments by the person examined or the examiner. The
reproducibility of the tests result can be examined, by repeating. The difference in the results
may be accounted for by the differences in instruments, techniques, or performance.
The manifestations (symptoms, signs, and tests) that are considered to be typical for a certain
disease have been used to formulate diagnostic criteria, i.e., conditions that must be fulfilled
for the diagnosis to be made. The choice of diagnostic criteria therefore determines whether
an examined individual is to be classif ied as having, a certain disease. If stringent criteria are
used, there is only a small probability that people who do not have the disease are being
classified as having it, but a relatively large probability that some people who have the disease
will be classified as not having it. On the other hand, if less stringent criteria are used, the
opposite type of misclassification will tend to occur: most people with the disease will be
correctly classified as having it, but there is a relatively large probability that people without
disease will be classified as having it.
There is no completely accurate method to identify those individuals who have a certain
disease. Some observations, though, are considered more accurate than others when making
a diagnosis. Naturally, diagnostic accuracy varies from one disease to another, but also from
one group of individuals to another. The accuracy also depends on what kind of examination
has been performed and on the interpretation of the observations.
The diagnostic criteria can be examined in terms of four parameters:
sensitivity
specificity
positive predictive value
negative predictive value
"Sensitivity" refers to the probability that the test result is positive, given that the people tested
actually have the disease that the test is supposed to detect. "Specificity" to the probability that
the test result is negative, given that the people tested do not have the disease that the test is
supposed to detect. Positive predictive value refers to the probability that the people who
suffer the disease tested will have positive test result. Negative predictive value refers to the
6
probability that the actual disease state is negative given that the test result is negative (Table
1).
Table 1. Possible combinations of whether or not a disease is present and whether the result
of a diagnostic test is positive or negative.
Actual disease status (Dx)
Test results (T) Positive Negative Total
Positive a b a+b
Negative c d c+d
Total a+c b+d
Sensitivity = Probability (T
+
, Dx
+
) =
a/a+c
Specificity = Probability (T
-
, Dx
-
) = d/b+d
Predictive value
+
= Probability (Dx
+
,T
+
) = a/a+b
Predictive value
-
= Probability (Dx
-
,T
-
) = d/c+d
7
4. Epidemiological measures of disease
Measures of disease occurrence are central to all epidemiological activity. Such measures
can be formulated in a variety of ways:
absolute numbers and numbers related to size of the population
incidence and prevalence
Measures of prevalence describe what proportion of the population has the disease in question
at one specific point in time. Prevalence is the total number of individuals who have a
characteristic or disease at a particular point in time divided by the number who are at risk of
having that characteristic or disease at that designated point in time. Prevalence depends on
both the number of people who have had the disease or characteristic in the past and the
duration of the disease or characteristic.
Measures of incidence, on the other hand, describe the frequency of occurrence of new cases
during a time period. Incidence is the number of new cases of a disease in a defined
population within a specified time-period divided by the number who are at risk of having that
disease or characteristic at that designated time-period.
It is usef ul to think of each individual as being in one of two "states": diseased or disease-free.
In this framework the prevalence measure describes the proportion of the population that is in
the diseased state at a specific time. The incidence measure describes the rate of flow from
the disease-free state to the diseased state.
The relevant measure of disease occurrence is incidence. Measures of prevalence may be
relevant in connection with the planning of health services or in assessing the need for medical
care in a population. For example, studies of chronic diseases such as rheumatoid arthritis, in
which the point of transition from non-diseased to diseased occurs gradually, and the definition
of disease is arbitrary, generally employ prevalence measures, whereas studies on cancer or
myocardial infarction generally use incidence measures.
8
The prevalence measure is called simply the "prevalence" (P) (other terms in use include
prevalence rate and prevalence proportion). The prevalence is defined as:
N
d
is the number of individuals having the disease at a specific time
N
t
is the number of individuals in the population at that point in time
The prevalence is dimensionless and can never take values less than 0 or greater than 1.
Example: A sample including 1, 500 women age 70-74 years was selected from the population
of Helsinki City area. After examination, 120 were classified as having the diagnosis of
depression. The prevalence of depression was:
This means that 8% of the women in the age group 70-74 suffer from depression.
There are two commonly used incidence measures: cumulative incidence or incidence
proportion (CI) and incidence rate or incidence density (IR).
The definition of cumulative incidence is:
N
gd
is the number of individuals who get the disease during a certain period
P
N
N
d
t
=
CI
N
N
gd
tb
=
P = =
120
1500
0 08 ,
9
N
tb
is the number of individuals in the population at the beginning of the period
Both numerator and denominator include only those individuals who at the beginning of the
period are free from the disease and therefore at risk to get it. The cumulative incidence is,
therefore, the proportion of individuals in the disease-free state at the beginning of the period
that move to the disease state during the period. That is, the numerator is a subset of the
denominator. Alternatively, it can be viewed as the average risk for the individuals in the
population to get the disease during that period. Being a proportion, the cumulative incidence is
dimensionless and can only take numeric values in the range from 0 to 1.
Example: The Swedish sensus from 1960 showed there were 3.076 males age 20-64 who
were employed as plastic workers. According to the Swedish Cancer Environment Registry, 11
of those workers developed brain tumors during the period 1961-1973 (National Board of
Health and Welfare). The cumulative incidence during the 13-year period therefore is:
This means that 0.4% of plastic workers develops a brain tumor in a time span of 13 years.
In some studies different subgroups of the study population are considered at risk of getting
the disease during different periods of time and individuals are considered at risk for periods of
varying lengths. This variation in the risk period derives from the fact that different individuals
enter the study at dif ferent points in time or that some migrate during the observation period. In
such situations the cumulative incidence may not be directly calculable from the data. The
length of the observation period directly affects the cumulative incidence: the longer the period,
the greater cumulative incidence.
Incidence rate is a basic measure of disease occurrence, which is defined as
N
cd
is the number of cases of the disease that occur in a population during a period of time
CI = =
11
3076
0004
.
.
Poistettu: <sp>
10
n
rd
is the sum for each individual in the population of the length of time at risk of getting the
disease
The sum of the time periods in the denominator is often measured in years and is referred to
as "person years," "person time," or "risk time." For each individual in the population the time at
risk is the time during which that individuals is in the study population and remains free from the
disease, and theref ore at risk to get it. These time periods at risk are then summed for all
individuals. The rationale is that the total number of individuals who move from the disease -
free state to the disease state during any period of time is the product of three factors: size of
the population, the length of the time period, and the "force of morbidity" that operates on the
population. It is this "force of morbidity" that the incidence rate measures. Therefore the
incidence rate is obtained by dividing the number of cases by the product of the size of the
population and the length of the period, which is equivalent to summing the individual periods of
time for each individual in the population. By dividing the number of cases by the time at risk,
the length of the observation period is taken into account. Also, individuals entering or exiting
the population during the observation period because of initiation, competing mortality, or any
other reason are automatically accounted for. Thus, by including time at risk in the definition,
the incidence rate accounts for the major drawbacks encountered with the cumulative incidence
measure.
Example: In 1973 there were 29 cases of myocardial infarction in Stockholm among men age
40-44 years (Ahlbom 1978). The number of person-years was 41.532 for men in that age
group. The incidence rate per year is therefore:
This means that 7 male persons/10 000 (7 x 10
-4
) per one year, in the age group 40-44 will
suffer from a myocardial infarction.
I = =
29
41532
00007
,
.
11
In practical situations it is often not possible to calculate the time at risk for each individual. It
may not even be possible to exclude the period during which some of the individuals are no
longer at risk because they already have developed the disease. An approximation to the total
time at risk that is usually satisfactory can be obtained by multiplying the average of the
population size at the beginning and the end of the observation period by the length of the
period. The size of the population at the middle of the observation period may also be used for
this approximation. In this case the incidence rate can be calculated as:
Example: During a 5-year period 270 cases of duodenal ulcer occurred in the male population
of a city. The number of men in the city was 18. 500 at the beginning of the period and 21. 500
at the end. Which is the incidence rate?
Number of cases = 270
Average population size =
Observation period = 5 years
Total risk period = 20. 000 x 5 = 100. 000 person-years
Incidence rate per year:
( . . )
.
18500 21500
2
20000
+
=
IR = =
270
100 000
0 0027
.
.
12
The incidence rate is not a proportion like the two previous measures, since the numerator is
the number of cases and the denominator is the number of person-time units. The magnitude
of the incidence rate can never be less than zero but there is no upper limit for the numerical
value of the incidence rate.
The interrelation among the three measures
Prevalence is dependent on the incidence and the duration of disease. In a stable situation this
association may be expressed as follows, where D indicates average duration of the disease:
P/ (1 - P) = I x D
The denominator on the left side of the equation reflects the part of the population that is free
from the disease. It is included in the formula since only those people who are free from the
disease are at risk of getting it. For rare diseases, i.e., diseases where P is low, the following
approximation may be used:
P = I x D
The cumulative incidence depends on the incidence rate and the length of the period at risk. It
also is affected by mortality from diseases other than the disease studied, as stated
previously. If this mortality from other diseases is disregarded the following relation applies:
CI = 1 - exp(-I x t)
where t is the length of the period and "exp" indicates that the mathematical constant e 2.718
should be raised to the power given by the expression in parenthesis. For diseases with low
incidence rate or when the period is short, the following approximation may be used:
CI = I x t
Despite the close interrelationship of prevalence and incidence, each provides a somewhat
different type of information with diff ering utility. Prevalence measures are most useful for
13
health care providers, to assess the public health impact of a specific disease within a
community and to project medical care needs for affected individuals, such as numbers of
health care personnel or hospital beds that will be required.
The measures of disease occurrence discussed above may be calculated for a whole
population or calculated separately for parts of the population. In the first case the measures
are called "crude" measures and in the latter case "specific" measures. For example, if
incidence rates are calculated f or different age groups within a population, they are referred to
as age-specific incidence rates. Crude rates represent the actual experience of the population
and provide data for the allocation of health resources and public health planning. The
magnitude of a crude measure for a population depends not only on the magnitude of the
specific measures that apply to subgroups of tile population, but also on the way the population
is distributed over the diff erent subpopulations. Specific rates provide the most detailed
information about the pattern of the disease in a population.
14
5. Measures of association
In epidemiological studies the f requency of disease occurrence among individuals who have a
certain characteristics is generally compared with the corresponding frequency among those
who do not have that characteristic. The compared groups are usually referred to as exposed
and non exposed. This comparison constitutes the fundamental way of studying the
association between exposure and disease occurrence.
The concept of an association, or statistical dependence, between a factor and a disease is
fundamental to ascription of the factor as possibly causal. An exposure or attribute that
increases the probability of a disease is a risk factor. Epidemiologists calculate a measure of
association as a single summary parameter that estimates the association between the risk of
developing a disease and an exposure. These measures of association draw on two concepts:
the probability of an event or disease and the odds of an event or disease. The most
frequently used measures of association are the attributable risk (AR), relative risk (RR) and
the odds ratio (OR).
Attributable risk is the difference between the incidence of the disease in exposed group (I
e
)
and incidence of the disease in non-exposed group (I
0
)
AR = I
e
I
0
AR indicates on an absolute scale how much greater the frequency of disease is in one group
compared with the other.
Relative risk (RR) indicates the average risk of disease that is due to a given exposure in the
exposed group. It provides an estimate of the magnitude of an association between an
exposure and a disease; it is the ratio of the incidence of the disease in exposed group (I
e
)
divided by the corresponding incidence of the disease in non-exposed group (I
0
).
RR = I
e
/ I
0
A relative risk of 1.0 indicates that the incidence rates of disease in the exposed and non-
exposed groups are identical and thus that there is no association observed between the
15
exposure and the disease in the data. The value greater than 1.0 indicates a positive
association or an increased risk among those exposed to factor. A relative risk less than 1.0
means that there is an inverse association or a decreased risk among those exposed.
The relative risk (RR) can be calculated on the basis of historical or prospective cohort studies
as well as cross-sectional studies. In a case control study, where participants are selected on
the basis of disease status, it is usually not possible to calculate the rate of development of
disease given the presence or absence of exposure. In this case relative rate can be estimated
by calculating the ratio of the odds of exposure among the cases to that among the controls.
To facilitate the calculation of these measures, epidemiological data are presented in a two-by-
two table (Table II, Table III).
Table II. Two-by-two table showing the relationship between the presence of a disease and
exposure to a risk factor that may contribute to the disease
*
Disease No disease
exposure a b a+b
No exposure c d c+d
a+c b+d
*
The letter a represents people who have the disease and have been exposed to a risk factor
under study.
RR
a
a b
c
c d
=
+
+
16
Table III. Two-by-two table showing the relationship between the case-controls and exposure
to a risk factor that may contribute to the disease
*
Exposure No exposure
Cases a b
Controls c d
*
The letter a represents cases, which have been exposed to a risk factor under study.
RR is a measure of the strength of the association between an exposure and disease and
provides information that can be used to judge whether a valid observed association is likely to
be causal. In contrast, AR provides a measure of the public health impact of an exposure,
assuming that the association is one of cause and effect. The magnitude of the RR alone does
not predict the magnitude of the AR. The odds ratio and risk ratio are very similar in instances
of rare disease but are quite dissimilar when the prevalence begins to exceed 5 or 10 per cent.
Example: Of 595 patients who had received blood transfusion and 712 patients who had not,
75 and 16, respectively, developed hepatitis during a 2,5 year follow-up in a study designed to
evaluate the relative risk.
The study is a cohort study.
Healthy Sick
Exposed, n=595
Unexposed, n=712
OR
ad
cb
=
Poistettu: <sp>
Poistettu: <sp>
17
This means that there is a 6,5 time higher risk to develop hepatitis if you have had a blood
transfusion compared to if you have not had one.
Example: The incidence of myocardial inf arction is higher in Finland than in Sweden. In the
southern suburbs of Stockholm a study was conducted based on middle-aged men who
developed myocardial infarction in the years 1974-76 and on controls from the general male
population. Information was compiled on the country of origin and length of stay in Sweden.
Calculate the Odds ratio of myocardial infarction
a) for those living in Sweden for 0-19 years
b) for those living in Sweden for 20 years or more
Table IV. Distribution of cases and controls for those living in Sweden for 0-19 years
Exposed
yes no
Cases 22 324
Controls 31 832
CI
exp
. = =
75
595
013
CI
unexp
, = =
16
712
0 02
RR = =
013
0 02
6 5
,
,
,
OR
0 19
22 832
324 31
18
=
( )
( )
,
18
Table V. Distribution of cases and controls for those living in Sweden for more than 20 years
Exposed
yes no
Cases 10 324
Controls 24 832
These both measures should be interpreated so that there is almost a twofold risk for middle
aged Finnish men to suffer from myocardial infarction if they have lived in Sweden for less than
19 years compared to Swedish middle aged men. Furthermore, the risk for Finnish men is only
a little elevted if they lived in Sweden for more than 20 years.
OR
20
10 832
324 24
11
=
( )
( )
,
19
6. Design strategies in epidemiological research
The basic design strategies used in epidemiological research can be broadly categorized
according to whether such investigations focus on describing the distributions of disease
(descriptive epidemiology) or elucidating its determinants (analytical epidemiology).
Descriptive epidemiology is concerned with describing the general characteristics of the
distribution of a disease, particular in relation to person, place in time. The characteristics
measured in this type of study are: crude number of e.g. death, or of a specific disease,
number adjusted to the size of population at risk (prevalence and incidence), means of studied
variables (e.g. mean age, mean alcohol consumption and so on). Descriptive data provide
valuable information to enable health care providers and administrators to allocate resources
efficiently and to plan effective prevention or education programs. Due to limitation inherent in
their design, however, descriptive studies are primarily useful for the formulation of hypotheses
that can be tested subsequently using an analytical design.
There are three main types of descriptive studies:
correlational (ecological) study
case reports and case series
cross-sectional surveys
Correlational study uses data, which usually have been routinely collected, from entire
population to compare disease frequencies between different groups during the same period
of time or in the same population at different points of time. While such studies are useful for
the formulation of hypotheses, they cannot be used to test them because the number of
limitations inherent in their design. Correlational studies refer to whole populations rather than
to individuals, it is not possible to link an exposure to occurrence of disease in the same
person.
Example: Examination of the relationship between per capita consumption of cigarettes and
operative procedures for the treatment of non-unions in five counties in California in 1996.
Even if this study showed that counties with a higher rate of consumption of cigarettes had a
higher rate of operations for non-unions, the investigator still could not be sure that the
individuals who smoked in these counties truly had a higher rate of non-union. This conclusion,
20
if indeed it is erroneous, is known as an ecological fallacy because the correlation between two
ecological variables is often different from the corresponding individual correlation within the
same populations.
Case report is the most basic type of descriptive study of individuals, consisting of careful,
detailed report by one or more clinicians of the profile of a single patient. The individual case
report can be expanded to a case series, which describes characteristics of a number of
patients with a given disease. Some people would argue that these types of studies are not
epidemiological studies, since they focus on a few individuals. Rather they are clinical studies,
which may generate hypotheses to be tested in epidemiologic studies.
Cross-sectional survey (also known as prevalence studies or disease-frequency surveys) is a
study in which the status of an individual with respect to the presence or absence of both
exposure and disease is assessed at the same point in time. Since exposure and disease are
assessed at the same point in time, cross-sectional survey cannot always distinguish whether
the exposure preceded the development of the disease or whether presence of the disease
affected the individuals level of exposure. This type of studies useful for raising the question of
the presence of an association rather than for testing a hypothesis.
Example: Silverstein et al., in a cross-sectional study, determined the prevalence of carpal
tunnel syndrome among 652 workers in thirty-nine jobs at seven different industrial sites.
Specific hand-force and repetitiveness characteristics their were estimated for the different
jobs. The prevalence of carpal tunnel syndrome ranged front 0.6 percent among workers in
low-force, low-repetition jobs to 5,6 per cent among those in high-force, high-repetition jobs. In
order to infer that a statistical association between job-related exposure factors and carpal
tunnel syndrome is evidence of etiology, it should be demonstrated that the job exposure
occurred before the carpal tunnel syndrome. The temporal relationship between physical load
factors and the onset of carpal tunnel syndrome cannot be demonstrated in a cross-sectional
study examining the prevalence of carpal tunnel syndrome. Prevalence studies are performed
on survivor populations and thus may be affected by selection bias for example, individuals
who have more severe carpal tunnel syndrome may have left the workforce and thus may not
be accounted for in a cross-sectional survey.
21
In analytical study designs the comparison is explicit, since the investigator assembles groups
of individuals for the specific purpose of systematically determining whether or not the risk of
disease is different for individuals exposed or not exposed to a factor of interest. These
studies can be divided into two broad design strategies:
observational (non-experimental)
intervention (experimental)
The major difference between the two lies in the role played by the investigator. In
observational studies, the investigator simply observes the natural course of events, noting
who is exposed and non-exposed and who has and has not developed the outcome of interest.
In intervention studies, the investigator themselves allocate the exposure and then follow the
subjects for the subsequent development of the disease. For ethical and practical reasons,
however, epidemiological investigations are usually non-experimental; that is, they are based
on existing exposure conditions.
There are two basic types of observational analytical investigations:
cohort (also known as a "follow-up") study
case-control study
The goal of an observational study is to arrive at the same conclusions that would have been
derived in experimental trial. In general, the decision to use a particular design strategy is
based on features of the exposure and disease, the current state of the knowledge, and logistic
consideration such as available time and resources.
Cohort studies
The cohort study is the more straightforward approach. All subjects in the study population are
assigned to a category of exposure at the start of the time period under study. Study subjects
are defined into two groups: people who are exposed become the exposed cohort, people who
are not exposed become the non-exposed cohort. The study subjects are then followed up for
a defined observation period and all new cases of the disease under study are identif ied. All
potential subjects must be free of the disease being studied at the time that exposure status is
22
defined. The combination of the population and time period at risk (i.e., the person-time
observed) is sometimes referred to as the study base, indicating the base experience from
which the cases arise. The information obtained is used to estimate the incidence rate or
cumulative incidence among the exposed and unexposed groups. The comparison of disease
incidence among exposed and unexposed may be either absolute (absolute risk) or relative
(relative risk).
The unexposed group is intended to provide information about the disease incidence rate that
would be expected in the exposed group if the exposure under study did not affect the
occurrence of the disease. Therefore, the unexposed group should be selected in such a way
that it is similar to the exposed group with regard to other risk indicators for the disease under
study.
In principle there are three different approaches to defining the unexposed group:
1. Internal comparison: A single cohort is identified that contains a sufficient number of
exposed and unexposed subjects.
2. External comparison: An exposed cohort is identified and efforts are made to find
another cohort that is unexposed but is similar in other respects to the exposed
cohort.
3. Comparison with the "general" population: An exposed cohort is identified and
comparisons are made with the disease incidence in, for example, the total
population of a defined geographic region (considered as "unexposed").
Conf ounding may occur if a risk indicator other than the studied exposure is unequally
distributed between the groups. Confounding can to a certain extent be controlled in the data
analysis; for example, differences in the distributions of factors such as age and sex between
the groups can be corrected in the analysis. There are other problems to be considered,
however, when selecting the unexposed group.
The incidence of disease can vary considerably over time (year, season etc.), and by
geographic area, ethnic background, and socioeconomic status. Differences between the
exposed and unexposed groups with regard to these or other factors can influence the results
of the study. To avoid such problems it may be advisable to select exposed and unexposed
23
subjects with similar distributions for these factors and to follow them during the same
observation period.
If information from the total population is used to represent the "unexposed" group, then all
study subjects classified as exposed (and other people with the same exposure) are also
included among the "unexposed. Unless the proportion of exposed subjects in the total
population is low, this dilution of the "unexposed" population may result in an underestimation
of the relative morbidity in the exposed group.
The procedures used to identify new cases of the disease under study should be similar for the
exposed and unexposed groups. If, for example, certain screening procedures are used to a
different extent in the exposed group, case identification may be more complete in the exposed
group.
Selecting more than one unexposed group can reveal to what extent the study results are
influenced by the selection of an unexposed group. Separate analyses using these groups that
yield similar results indicate that the results probably were not influenced by the choice of the
unexposed group. Another approach that has sometimes been used to check the comparability
between the exposed and unexposed group is to include comparisons of the occurrence of
diseases that are not expected to be associated with the exposure under study.
Cohort studies are sometimes based on information about exposure and disease collected in
the past (so called "retrospective cohort studies"). By the other words, the investigation is
initiated at a point of time after both the exposure and disease have already occurred. The
accuracy of such a study depends upon the completeness of disease ascertainment in the
registry for the population and time period under study.
In prospective cohort studies, the groups of exposed and unexposed subjects have been
assembled at the beginning of the study, but the disease has not yet occurred, so that
investigator must conduct f ollow-up during an appropriate interval to ascertain the outcome of
interest.
Case-control studies
24
For most diseases, the occurrence of new cases is a relatively rare event. Thus, in a cohort
study, a large number of study subjects may have to be examined for exposure status and
followed up for a long time to obtain a sufficient number of cases. Such a study is often not
practical or feasible. The problem can be avoided by using exposure, information from a
sample of the study population rather than including the entire study population.
Case-control studies are based on this principle. As in cohort studies, information is obtained
on all cases that occur in the study population during a defined observation period. In addition,
a comparison group of "controls" are selected as a representative sample of the study
population; ideally the control group reflects the exposure distribution in the entire study
population. Exposure information is then obtained for cases and controls only, rather than for
all members of the population.
In a case-control study, the investigator selects individuals on the basis of whether or not they
have the disease and then determines their previous exposure. Exposure status is determined
retrospectively f or both cases and controls. Exposure status is compared between cases and
controls. If the exposure is more common among cases than controls, then the exposure is
associated with the disease. The association is expressed as the odds ratio. The control group
in a case-control study should be selected in such a way that it reflects the exposure
distribution in the study population, i.e., in the population that generated the cases. There are
two different approaches to control selection:
1. A random sample of the study population. This approach has the advantage that the
controls will be representative of the study population in a formal (statistical) sense and
therefore the selection process for controls does not introduce any systematic error. This
option obviously requires that the population is accessible for random sampling. Potential
disadvantages of random sampling for control selection are the possibility of a high non-
response rate among healthy population controls, and the possibility of differences in the
quality of exposure information between cases and healthy controls.
2. A sample of the study population that is not randomly selected. This is the only option
when the cases are identified in such a way that the study population
is not accessible for random sampling. For example, if the cases are patients
diagnosed with the disease being studied in a particular clinic, these patients
most likely do not represent all cases of the disease occurring in a population
from which a random sample may be drawn, and therefore a nonrandom
selection of controls is necessary. Sometimes this approach is used for other
25
reasons, such as reducing non-response among controls or improving, the
comparability of exposure information between cases and controls.
While such nonrandom selection schemes may result in greater comparability of cases
and controls, controls selected in this way may not reflect the exposure distribution in the
overall study population, thus possibly introducing, a systematic error.
When controls are selected non-randomly, it can be difficult to judge to what extent their
exposure distribution reflects the exposure distribution in the population that generated the
cases. When hospital controls are selected, a single diagnostic group may be used as a
source of controls, provided that neither the disease nor the probability for hospitalization is
associated with the studied exposure(s). It may be useful to select two or more diagnostic
groups sufficiently large to permit separate analyses to minimize the problem, especially when
such associations cannot be ruled out at the beginning of the study. Using a sample of all
patients (or hospitalization episodes) regardless of diagnosis is less advisable, since many
exposures (such as diet, tobacco and alcohol consumption) tend to be related to several
common diseases. When there is a choice between these two approaches in selecting
controls, random sampling is the preferred method.
In case-control studies, controls are sometimes selected by individual matching to the cases:
for every case, one or more controls are selected who are similar to that case in certain
respects. Cases and controls are matched on potential confounders, such as age, sex and
residence, for which information can be obtained before data collection begins.
The matching is performed with the purpose of reducing random error rather than confounding.
Example: Study of the epidemiology of acute herniation of lumbar intervertebral discs, (Kelsey
and Hardy). The investigators compared the characteristics of patients who had such
herniation with those of two control groups who were known not to have it and found that
driving a motor vehicle was associated with an increased risk of herniated discs. That study
provides an excellent demonstration of how the many concerns in case-control studies,
including selection of controls, analysis of data with various confounders, and consideration of
other biases, should be addressed.
26
Experimental studies or clinical trials may be viewed as a type of prospective cohort study,
because participants are identified on the basis of their exposure status and followed to
determine whether they develop the disease. The distinguishing feature of the intervention
design is that the exposure status of each participant is assigned by the investigator.
Intervention studies can be divided on:
true experimental (randomized controlled trial, RCT)
quasi-experimental
In randomized controlled trial the exposure status is determined by random allocation. It is the
optimal design to study cause-effect or efficacy of an intervention. Ideally the control group
should receive placebo treatment and both the patients and the researcher should be blinded
to the treatment status, that is no-one knows who is being treated and who is not until after the
outcome is measured. This is randomized double blind placebo controlled trial. When using a
placebo in the control group, the expected placebo ef fect in the intervention group is accounted
for when the two groups are compared. If there is a placebo-effect of the intervention, this
placebo effect will be the same in both groups, and the observed differences between the
groups are likely to be due to the real effect of the intervention. Unfortunately such studies are
not possible for practical, economical, or ethical reasons.
Quasi-experimental study is an experimental study where at least one criteria for RTC is
missing: either no random allocation of exposure or/and no control group.
Example (quasi-experimental design): Testing a new treatment in one group of ill patients and
assessing whether the subjects are better or worse after the treatment (before-after design).
No control group, in a sense the group serve as its own control in such a study. Or if the
treatment is tested in a group of patients, which is compared to another group of patients who
do not receive the treatment, but treatment allocation is not random.
Intervention studies are often considered as providing the most reliable evidence from
epidemiologic research. This is due to the unique strength of randomization as a means of
determining exposure status in a trial. When participants are allocated to a particular exposure
group at random, such a strategy achieves, on average, control of all other factors that may
affect disease risk. The ability to control both known and unknown influences makes the
27
randomized trial such a powerful epidemiologic strategy, especially for studying small or
moderate effect. There are also particular concerns of costs and feasibility for intervention
studies. Nevertheless, when well designed and conducted, intervention studies can indeed
provide the most direct epidemiologic evidence on which to judge whether an exposure causes
or prevents a disease.
Advantages and disadvantages of the cohort and the case-control designs
The cohort design might seem to be a more appealing choice than the case-control design. It
has a simple, logical structure that leads to measurements of disease incidence for the
exposed and unexposed groups, or each category of exposure if several are used. This design
permits absolute as well as relative comparisons of disease incidence among the exposed and
unexposed. The case-control design provides somewhat less information, in that usually only
the relative risk can be estimated. Absolute measures of incidence can also he made in case-
control studies, but only if the sampling fraction for the controls or the baseline incidence rate
for the total study population is known.
Case-control studies, like cohort studies, are based on follow-up of incident cases in a certain
study population during a def ined observation period. Case-control studies, however, use
exposure inf ormation from a sample of' the study population rather than from the whole study
population. Unless the disease incidence is very high, obtaining exposure information for a
sample of the study population will be much less expensive and can yield more information on
exposure, as fewer subjects need to be studied. The case-control design therefore makes
investigations based on large study populations more feasible, an important consideration
since large studies is usually needed to reduce random error.
The disadvantage of the case-control design is the difficulty in selecting a satisfactory control
group, with the consequent problem of introducing systematic error in the study by the selection
of controls. This problem is an issue only for non-random selection of controls. When the
control group is defined as a random sample of the study population, control selection is a
simple technical procedure and introduces no systematic error beyond what would be present
in a cohort study using the entire study population. Using information from sample (controls)
rather than from the entire study population does increase, random variability, but when the
size of the control group is adequate, this effect is small or negligible. The amount of random
28
error that can be removed by expanding the control group to include the entire study population
is often trivial, whereas the corresponding cost would be great.
In case-control studies involving contact with study subjects or their relatives, questions about
previous exposure are usually answered by the study subjects only after the cases have fallen
ill. The cases, therefore, may have spent more time thinking about past exposures and causes
of their disease, while the controls have no motivation to do so. This difference between
cases and controls in the accuracy and completeness of exposure information can introduce a
"recall bias" into the study. A similar bias may be seen in cohort studies in which, for some
reason, exposure information is obtained only after the cases have been identified. In Table IV
the advantages and disadvantages of the different study designs are presented.
29
Table IV. Advantages and disadvantages of the different study designs.
Study design Advantages Disadvantages
Ecological study is quick and cheap if the population based data have been routinely collected and is
readily available
is a useful first step to explore an association
not possible to e
- information is no
level
- temporal sequen
data quality m
routinely collecte
Cross sectional
study
- sually good control over selection of study subjects; random sampling easy;
representative sample
- yields prevalence
- relative risk quick and cheap
- can study several outcomes andexposures at the same time
- does usually
- large samp
exposures
- does not yie
- does not yie
Cohort study - logic temporal sequence from exposure to outcome (in most cases we can be
certain that the exposure camebefore the outcome)
- establishes the absolute risk
- exposure can be measured without bias because the outcome is not known at the
time whenexposure is measured
- can assess multiple outcomes
- potential confounders can be measured easier than in a case control study
- often not
outcome is
- time consum
- long delay f
results are a
- ethical conc
30
Case control study - valuable for studying rare conditions
- can be carried out ina short time
- relatively inexpensive
- relatively small number of study subjects required
- can study several exposure variables at the same time
- limited to the
- does usual
exposure an
- does not yie
- does not yie
- potential bia
- fromselecti
- in measuring
Randomized
controlled trial
- with randomization, comparability of groups is very likely for both known and
unknown confounders
- the temporal sequence canbe established
- provides the best evidence for a causal relationship/efficacy of treatments
- may be exp
- may be une
the interven
- to a greater
situation (p
fromthose
- standardize
practice
- this may ma
31
32
7. Statistical association and cause-effect relationships
Testing an epidemiological hypothesis first involves consideration of the concept of
association between a particular exposure and a disease. Association refers to the statistical
dependence between two variables, that is, the degree to which the rate of disease in persons
with a specific exposure is either higher or lower than the rate of disease among those without
that exposure. The presence of an association, however, in one of cause and effect. Making
judgments about causality from epidemiological data involves a chain of logic that addresses
two major areas:
whether for any individual study, the observed association between an exposure and
disease is valid
whether the totality of evidence taken from a number of sources supports judgment of
causality
The first - related to the question whether the particular study findings reflect the true
relationship between the exposure and the disease is a matter of determining the likelihood
that alternative explanations such as:
chance
bias
confounding
could account for the findings.
Causal Inference
Contemporary epidemiologists are concerned with determining the etiology of disease. A
cause is an event that, either alone or in conjunction with other elements produces a sequence
of other events that result in an effect. Patients, referring physicians, employers, and other
frequently ask Orthopedists: What is the cause of this disease? It is hopeless to attempt to
answer that question; however, we may be able to answer other questions, such as "What are
some of the causal contingencies in a particular disease?" and "Is there a known factor, a risk
factor for this disease?" Epidemiologists prefer to use the term risk factor instead of cause to
33
indicate an attribute or exposure that is related to an increased probability of a disease.
Additionally, for a factor to be considered a risk factor it must precede the occurrence of the
disease and the observed association must not be due to problems with the study design or
the analysis of the data.
The important, or primary, causal contingency may depend on one's viewpoint. Each observer
selects the causal contingency on the basis of his or her perspective. It is compelling to search
for a primary, or the most important, causal factor among many. From a practical standpoint, if
the goal is to prevent a disease it may be more beneficial to focus on a causal factor
somewhat remote from the disease. For instance, general improvement in living conditions and
economic development in underdeveloped countries can do more to reduce the incidence of
tuberculosis and its orthopedic manifestations than can any vaccination, chemotherapy, or
operative procedure.
Epidemiologists, more than other medical scientists, are concerned with philosophical theories
rather than the purely technical aspects of the experimental method because experiments play
a minor role in the analysis of naturally occurring phenomena. Causal inference is the logical
development of a theory based on observations and a series of arguments that attributes the
development of a disease to one or more risk factors.
Thus, testing hypotheses and discarding those that fail advances knowledge. The
epidemiologist starts with a hypothesis, collecting data about a disease that then are used
either to refute or to accept the hypothesis. The alternation between the generation of
hypotheses and the collection of data allows many hypotheses to be discarded without any
experiment being performed. Thus, experiments are reserved for the testing of hypotheses
derived observational studies. Judgments about causality in epidemiology depend on new
knowledge.
The probability that an association exists is the first criterion used in causal inference in
epidemiology. For an association to be considered causal, the cause must precede the effect
(the property of time-order) and there must be an asymmetrical direction such that the cause
leads to the effect.
Statistical Inference
34
Statistical inference is part of the basis of epidemiology because the observations studied by
epidemiologists are subject to random fluctuations. The testing of hypotheses is a statistical
procedure to determine the probability that the data that are collected are consistent with the
specific hypothesis under investigation. Make example of hypothesis formulation. The
opposite of H
0
is a hypothesis, H
A
, that states: .. Conventionally, the investigator seeks to
deny or nullify H
0
; thus, it is known as the null hypothesis. H
A
and H
0
are mutually exclusive and
exhaustive: that is, one or the other must be true, but they cannot both be true. Thus, if H
0
is
denied, then H
A
is affirmed. H
A
and H
0
refer to the entire population, even though data are
available only from one sample of patients.
The significance test is based on the calculation of a test statistic (for example, the t value, z
value, or chi-square value) and on some theoretical assumptions. One such assumption is that
the null hypothesis is true, and another is that sampling uncertainty is random. On the basis of
these assumptions, it is possible to calculate how likely or unlikely the outcome observed in the
sample would be. The test statistic, therefore, is a number that compares the observed and
expected values of the parameter being measured under the null hypothesis. The significance
test encompasses the rationale that there is some range of test-statistic values such that either
the assumptions of the test, including the null hypothesis, are true or a rare event has occurred
or one of the assumptions is untrue and, specifically, the null hypothesis is false. The point at
which a test-statistic value is rare enough to warrant rejection of the null hypothesis is
determined by convention but typically is set at the value that would occur no more than 5% or
1% of the time in repeated tests if the null hypothesis were true. This absolute value, which
must be exceeded in order for the null hypothesis to be rejected, is called the critical value. The
probability that the test-statistic value is too small to be consistent with H
0
being true is known
as the signif icance level, or ; it is conventionally set at either 0.05 or 0.01 and is commonly
called the p value. In summary, if the test statistic represents an occurrence of less than 5% (or
1%) of the time under random sampling if the null hypothesis were true, then the result is
considered significant and the null hypothesis is rejected in favor of H
A
.
Many tests of significance have been used, depending on the nature and distribution pattern
(discrete, continuous, categorical, and so on) of the data being analyzed. Often, an investigator
selects the appropriate test, calculates the test statistic (for example, the t value, z value, or
chi-square value) and the significance level (the p value) with use of the collected data, and
either rejects or fails to reject the null hypothesis.
35
When data are analyzed for all possible statistical associations without a previous specific
hypothesis, the likelihood that an investigator will find a significant association by chance alone
increases as the number of statistical tests that are performed increases.
There are two types of error associated with the significance test. It is possible to reject the null
hypothesis when a rare event has occurred even though the null hypothesis is actually true. This
is called a type-I error. In all cases in which a null hypothesis is true, a type-I error will occur
(100 x ) per cent of the time, where equals the significance level (usually 0.05 or 0.01). In a
type-11 error, the null hypothesis is false but the calculated test statistic is not significant and
therefore is determined to be consistent with the null hypothesis being true. The null hypothesis
is thus accepted in error. The relative frequency with which a type-II error occurs is symbolized
by . An experiment usually is designed to control the probability of 0 to be less than 0.20.
Power- is the complement of a type-II error, or the probability that the null hypothesis will be
rejected when it is indeed false and is equal to (1-). Power is a function of the level of
significance, the reliability of' the sample data (the degree of spread in the data or the standard
deviation), and the size of the experimental eff ect. A power analysis should be considered if no
significant difference can be found between the two groups being compared in a study. The
groups may not be enough to allow detection of a significant difference (the null hypothesis
being that there is a diff erence), and the power analysis will demonstrate the probability that the
null hypothesis has been correctly rejected. The probability of a type-I or type-II error is
inversely related for any determined experimental design and fixed sample size. Improving the
experimental design and increasing the sample size can increase power.
Estimation is another tool used in statistical inference concerned with calculation of the values
of specific population parameters. A point estimate (the sample mean) can be used to
estimate the population mean. A weakness in the point estimate is its failure to make a
probability statement regarding how close the estimate is to the population parameter. A
confidence interval estimate remedies this problem by providing an interval of plausible
estimates of the population mean as well as a best estimate of its precise value. The
confidence interval that is conventionally chosen is 95 or 99 per cent, similar to the
conventional choice of 0.05 or 0.01 for the level of significance. The 95% conf idence interval
means that, assuming the sample mean will follow an approximately normal distribution, 95%
of all sample means based on a given sample size will fall within +1.96 standard errors of the
36
population mean. This number is derived from tire mathematics, of normal curve distribution.
Similarly, 99% of all sample means based on a given sample size will fall within +2.576
standard errors of the population mean (the 99% confidence interval). A confidence interval,
therefore, is a range of values for a study variable specifying the probability (usually 95%) that
the true value of the variable is included within the range. The size of a confidence interval
gives some idea of the precision of the point estimate in a way that is not offered by a p value;
the narrower the confidence interval, the more precise the data. As the sample size increases,
the size of the confidence interval decreases. As the standard deviation increases, reflecting
the increased variability between individual observations, the size of the confidence interval
increases. As the desired level of conf idence increases (for example, from 95 to 99 per cent),
the size of the confidence interval increases.
Statistical reasoning is based on the precepts that natural processes can be described by
stochastic models and that the study of random collections of individuals will allow identification
of "systematic patterns of' scientific import. The truth or falsehood of a hypothesis cannot be
inferred from a significance test. The type-I and type-II error rates define a critical region for
the summary statistic, which represents a decision rule as to whether the null hypothesis is to
be rejected or not. Goodman () warned that a decision rule tells nothing about whether a
particular hypothesis is true; it says only that, if investigators behave according to such a rule,
in the long run they will reject a true hypothesis not more, for example, than once in 100 times
and they may have evidence that they will reject the hypothesis sufficiently often when it is
false. Poor data are often irrelevantly supported by sophisticated statistical techniques. A
significance test makes no assumption about the plausibility of the null hypothesis.
Investigators must focus on the substance of the issues being studied and must not become
sidetracked by the mechanics of data analysis. Judgment must take precedence over statistical
inference when large sources of error are not quantified in the statistical analysis. These
sources of error are collectively known as bias and deserve caref ul consideration in the design,
execution, and analysis of all studies.
Bias
Bias is defined as any trend in the collection, analysis, interpretation, publication, or review of
data that can lead to conclusions that are systematically different from the truth. () Bias can
lead to an incorrect estimation of the association between an exposure and the risk of a
disease. Such an association is considered real if all attempts to explain it away as due to bias
37
have failed. There are three broad categories of bias: confounding bias, selection bias, and
information bias.
Conf ounding is a distortion in an effect (such as relative risk) that results from an effect of
another variable (the confounder) associated with the disease and exposure being studied.
Conf ounding can lead to an overestimate or an underestimate of the true association between
a disease and an exposure and even can change the direction of the observed effect. For a
factor to be a confounder, it must, in and of itself, be a risk factor for the disease in tile
unexposed population and it must be associated with the exposure variable in the population
from which the cases were derived. In addition, it must not be an intermediate step in the
causal pathway between the exposure and the disease. For example, if a study were to he
designed to explore whether smoking is a risk factor for motor-vehicle accidents, consumption
of alcohol would be considered a confounder because it is both a risk factor for motor-vehicle
accidents and it is associated with smoking. During the introduction of a new operative
procedure, the good-risk patients may be selectively managed with tile procedure while the
poor-risk patients may receive the standard treatment. This is called a confounding by
indication bias. Age and gender are well recognized confounders.
An effect modifier is a factor that changes the magnitude of an effect measure (for example,
relative risk or odds ratio). Effect modification differs from confounding: the latter is a bias that
the investigator tries to prevent or remove from the data, whereas the former is a constant of
nature. For example, immunization status against serum hepatitis is an effect modifier for the
consequences of being stuck by a needle that was used on a person infected with hepatitis.
Immunization status is an effect modifier because people who are immunized are less likely to
contract hepatitis than those who are not.
Selection bias (also known as detection bias and unmasking bias) refers to a distortion in the
estimation of an effect due to systematic differences in characteristics between subjects who
are selected for a study and those who are not. Selection bias can result when a procedure
used to identify a disease varies with the exposure status. For example, this bias can be
introduced by an examiner who performs a clinical evaluation without being blinded to the
disease or exposure status of the subject. The results of the evaluation may differ if the
examiner expects the disease to be present in the patients and absent in the controls. A lack of
a response to a questionnaire or loss of patients to follow-up would not be serious problems if
they merely resulted in a reduction in the number of subjects available for study; however, they
38
may result in selection bias if the respondents and the non-respondents of tile patients being
followed and those who have been lost diff er with respect to some characteristic being studied.
Information bias (also known as observational bias and misclassification bias) refers to a
distortion in the estimation of an effect that results from error in the measurement of either an
exposure or a disease or from the misclassification of subjects with regard to at least one
variable. In describing inaccuracy of measurement, two types of misclassification bias can
occur: non-differential misclassification, when the inaccuracy is the same for the two study
groups (for example, the patients and controls in a case-control study), and differential (non-
random) misclassification, when the inaccuracy differs between groups (for example, when an
exposure measure, such as repetitive work tasks, is determined more accurately among the
patients than among the controls). Non-differential misclassification increases the similarity
between the exposed and unexposed groups. Any association between exposure and disease
will be underestimated, so the observed estimate of effect is said to be biased toward the null
value of 1.0, meaning that if exposure and disease had no association the expected relative
risk would be equal to 1.0. Differential misclassification leads to a biased risk estimate that
may be either away from (an overestimate) or toward (an underestimate) the null value.
Whereas confounding bias is generally correctable in the analysis stage of a study, selection
and information biases may not be correctable.
Study-Design Strategies for Determining the Relationships between Exposure and Disease
Case-control and cohort studies are tire two basic types of observational analytical study
designs that have is their objective the testing of hypotheses.
This study design is good for the investigation of rare diseases but is highly susceptible to
selection and recall bias. Recall bias occurs when a study relies on the patient's memory to
determine exposure status because a patient who has a disease is more likely to remember
possible exposures than a healthy person.
Example: Probably the most impressive prospective cohort study is the Framingham study,
which began around 1950. Coronary heart disease and its consequences were thought by the
medical profession to be inevitable changes of aging; however, clinical observations and
descriptive epidemiological studies suggested that preventable environmental factors played a
39
role. More than 5000 people who did not have coronary artery disease were enrolled in the
long-term study, originally planned to last for twenty years, and each participant was given a
comprehensive medical examination every two years. The findings of this study have
continued to emerge, delineating the risk factors for heart disease and associated
atherosclerotic and non-atherosclerotic disorders, and they have led to preventive and
treatment paradigms that have resulted in better health care.
Example (Retrospective (historical) cohort study): It would be unethical to test the hypothesis
that low levels of exposure to radiation shorten human life expectancy with use of a
prospective study; however, if a group of people (a cohort) that already has been exposed can
be identified, then even if the exposure was in the past a retrospective cohort study would be
feasible. Seltser and Sartwell undertook such a study, comparing members of the Radiological
Society of North America with members of other medical specialty societies. These authors
demonstrated that the death rate was highest among radiologists, intermediate among
internists, and lowest among ophthalmologists and otolaryngologists; improvements in
equipment and safety techniques showed a so-called disappearing effect in the latter part of
the study.
The advantages of a cohort study include a temporal sequence of exposure and disease that is
usually clear, a minimization of observational bias in determining exposure, the ability to
examine multiple effects of one exposure, and usefulness when exposure is rare. Cohort
studies are time-consuming, expensive, not suitable for the investigation of rare diseases, and
potentially biased with regard to loss of subjects to follow-up when subjects must be f ollowed
for many years. A case-control study can be tested within a cohort study. In this study design,
patients and controls are drawn from the population in a cohort study, as has been done
frequently in the Framingham study. Nested case-control studies obviate the issue of recall
bias because the data are collected prospectively.
The advantage of intervention study (clinical trials) design is that known and unknown
confounders are distributed, on the average, equally among the study groups. Susceptibility
bias may occur in a clinical trial if the two groups are dissimilar in terms of their initial state. A
randomized clinical trial precludes the possibility of susceptibility bias.
40
A double-blind randomized trial is the best way to test the hypothesis that Coumadin (warfarin)
prevents deep-vein thrombosis after total hip arthroplasty because randomization minimizes
confounding and blinding eliminates selection bias.
A common practice in the epidemiological literature is the documentation of unavoidable
departures from the ideal study design as well as a discussion of estimates of the magnitude
and direction of biases and deduction of the extent to which these errors threaten or do not
threaten validity. Gartland reviewed ten articles on the long-term f ollow-up of patients after
primary total hip arthroplasty to determine the strategy that had been used in the design of
each study. He concluded that all of the studies were deficient in design, were flawed by
confusing data, and contained results of doubtful validity. The quality of orthopaedic studies
and literature will improve if sufficient attention is paid to alternative explanations and there is
recognition of the central role that epidemiological principles play in advancing knowledge and
understanding.
References
Ahlbom A. & Norell, S. Introduction to modern epidemiology. Epidemiology Resources Inc.
1990.