++
Epidemiologic methods can be used for a number of distinct purposes.
In the following sections, these areas of application are specified,
with corresponding illustrations drawn from the literature on AIDS.
++

Perhaps the most basic question that can be asked about a disease
is “What is the frequency with which the disease occurs?” To
answer this question, it is necessary to know the number of persons who
acquire the disease (cases) over a specified period of time, and
the size of the unaffected population. Measures of frequency of
occurrence of a disease, described in
Chapter 2: Epidemiologic Measures, are used to characterize
the patterns of the occurrence of the disease, described in
Chapter 3: Patterns of Occurrence, and the medical surveillance of the disease, discussed in
Chapter 4: Medical Surveillance. Typically, the criteria used to define the occurrence of a disease
depend on current knowledge about the disease; such criteria may
become more refined as the causes of a disease are delineated and
new diagnostic tests are introduced. For example, in 1982, the CDC
created an initial, relatively simple surveillance definition for
AIDS: “A disease, at least moderately indicative of a defect
in cell-mediated immunity, occurring in a person with no known cause
for diminished resistance to that disease.”
++
A more specific definition became possible once the causative
agent, HIV, was identified and tests for the detection of antibodies
to the virus were developed. In 1987, the CDC surveillance definition
was expanded to incorporate clinical conditions that are indicative
of AIDS. A 1993 revision further expanded the surveillance definition
to include three additional indicator conditions (pulmonary tuberculosis,
recurrent pneumonia, or invasive cervical cancer), or the presence
of a severely depressed CD4+ T-lymphocyte count.
In 2000, the CDC integrated monitoring of both HIV infection and
AIDS.
++
Such changes in diagnostic criteria can have a profound effect
on the apparent frequency of a disease. The expanded definition
of AIDS introduced in 1987 led to an increase in the number of reported
AIDS patients by about 50% during the next 2 years. The
1993 revision more than doubled the number of persons who met the
surveillance definition. Most of the latter increase was attributable
to persons made eligible on the basis of reduced CD4+ T-lymphocyte
counts and HIV infection. Accordingly, analysis of trends in disease
occurrence over time must account for the possible effects of any
temporal changes in diagnostic criteria.
++
The identification of patients with a disease can occur through
various mechanisms, most commonly by physician and laboratory reporting.
In the United States, a number of diseases, including AIDS, must
be reported to public health authorities. Monitoring the patterns
of occurrence of a disease within a population is referred to as surveillance. There are many potential
benefits from the collection of surveillance data. This type of
information (1) can help to identify the new outbreak of an illness,
such as AIDS, (2) can provide clues, by considering the population
groups that are most affected by the illness, to possible causes
of the condition, (3) can be used to suggest strategies to control
or prevent the spread of disease, (4) can be used to measure the
impact of disease prevention and control efforts, and finally, (5)
can provide information on the burden of illness, data that are
necessary for determining health and medical service needs.
++
The course of the AIDSpandemic in the United States is depicted
in Figure 1–1. To diminish the impact of changes of the
surveillance definition over time, a single definition, the 1993
CDC version, was used throughout. From 1985 through 1992 there was
an unrelenting rise in the number of newly reported cases. From
1993 through 1998 there was a progressive fall in the number of newly
reported cases. Between 1999 and 2001 the level of AIDS cases was
unchanged, with a slight increase in 2002. It should be noted that
the information in Figure 1–1 relates to the number of newly diagnosed cases per
year. Changes in the counts of new cases can be affected by a number
of factors, including among others, changes in the following:
++
++
- 1.Frequency with which the disease
occurs
- 2.Definition of the disease
- 3.Size of the population from which
the cases develop
- 4. Completeness of the reporting
of the cases.
++
With respect to point 2, the same surveillance definition was
used consistently for all years in Figure 1–1, minimizing
any confounding influence of a change in definition of the disease
over time. With regard to point 3, growth in the size of the population
of the United States could not explain more than a trivial amount
of the rise in the cases seen between 1985 and 1992. The national
population grew at only about 1% per year, whereas the
average annual increase in reported persons with AIDS exceeded 30%.
Moreover, the declines in AIDS cases observed between 1993 and 1998
occurred while the population of the United States continued to
grow. Concerning point 4, the overall completeness of reporting
of AIDS cases is estimated to be about 85% in the United States.
Although there is some internal variation by geographic subregion
and patient population, it is unlikely that these patterns could
have given rise to more than a small part of the trend observed
in Figure 1–1. Since items 2 through 4 do not appear to
account for the marked changes in the annual numbers of reported
AIDS cases, it is reasonable to conclude that the observed trends
reflect true changes in the occurrence of the disease.
++
For surveillance purposes, the size of the source population
from which cases arise usually is estimated from census data. The
frequency of disease occurrence is then expressed as the number
of new cases developing within a specified time among a standard
number of unaffected individuals. For example, during 2002 over
42,000 cases of AIDS were reported in the United States; the U.S. population
in 2002 was about 288,000,000. Dividing the number of reported cases
by the size of the population yields 0.00015 cases per person during
that year. For ease and consistency of expression of such figures,
epidemiologists typically express such frequencies of disease occurrence
for a population of a specified size, say 100,000 persons. By multiplying
0.00015 by 100,000, the number 15 is obtained. That is to say, within
a standard population of 100,000 persons in the United States, 15
persons would have been reported as developing AIDS during 2002. This
measure of the rapidity of disease occurrence is referred to as
an incidence rate. More information
on incidence rates is presented in Chapter 2: Epidemiologic Measures.
++
To characterize patterns of disease occurrence, incidence rates
may be determined for groups defined by geographic area. For example, in
Figure 1–2 annual incidence rates for AIDS are presented
by place of residence in the United States. During 1997, the incidence
rate for the District of Columbia was the highest observed, with
162.4 reported cases for every 100,000 residents. At the other extreme,
North Dakota experienced the lowest annual incidence rate (0.5 cases
per 100,000 residents). In other words, AIDS occurred in the District
of Columbia about 325 times (162.4/0.5 = 325)
more frequently than in North Dakota. Why are persons in the District
of Columbia so frequently diagnosed with AIDS; conversely, why are
persons in North Dakota so infrequently affected?
++
++
Answers to such questions typically do not derive from surveillance
information alone. Surveillance data usually are limited to general
characteristics of affected persons, such as their age, race, sex,
and place of residence. Although variations in incidence rates according
to these demographic features can lead to the identification of
high-risk groups, explanations for these patterns generally require
more in-depth investigation into personal characteristics, behaviors,
and environments.
++

To study personal and environmental characteristics, epidemiologists
often rely on interviews, review of records, and laboratory examinations.
Through such sources of information, a profile of characteristics
that accompany the disease can be generated. Associations between
these characteristics and the occurrence of disease can arise by
coincidence, by noncausal linkages to other features, or by cause-and-effect
relationships.
++
The epidemiologist is primarily interested in the last category,
ie, determinants of disease development, also known as risk factors. Identification of risk
factors can provide a better understanding of the pathways leading
to disease acquisition, and consequently, better strategies for
prevention.
++
Again, returning to the AIDS example, early epidemiologic studies
played an important role in determining the cause of this disease.
Within the first 5 months after recognition of this syndrome, the
CDC had received reports on 70 patients with AIDS in four urban
centers. Of these individuals, 50 homosexual male patients with
AIDS were interviewed; also interviewed were 120 unaffected homosexual
male comparison subjects. Persons who are affected with a disease
are referred to by epidemiologists as cases, and
unaffected comparison persons are called controls. Comparison
of the responses from cases and controls revealed that the AIDS
patients had a higher number of sexual partners. This type of investigation
is referred to as a case–control
study; the basic design of such a study is illustrated in Figure
1–3.
++
++
In essence, this study is an attempt to look backward in time
to identify characteristics that may have contributed to the development
of the disease. The increased number of sexual partners—as well
as a greater frequency of syphilis among cases—suggested
that AIDS resulted from a sexually transmitted infectious agent,
later discovered to be the HIV virus. Case–control studies
are described in Chapter 9: Case–Control Studies.
++
Comparison of historical exposures reported by cases and controls
can provide suggestive evidence of a cause-and-effect relationship.
This type of information, however, may be distorted or biased by the fact that the ability
of cases and controls to recall earlier exposures differs. Such
bias could be avoided by using a cohort
study design, in which exposure is assessed among unaffected
persons, and subjects are then observed for subsequent development
of illness. To collect such data, a cohort of 2507 homosexual men without
antibodies to HIV (seronegative) was questioned about their sexual
practices, and then followed for development of antibodies to HIV
(seroconversion). Within 6 months, 95 men (3.8%) seroconverted,
and the likelihood or risk of developing
HIV antibodies was found to be related to receptive anal intercourse.
The basic design of this cohort study is illustrated schematically
in Figure 1–4. Cohort studies are discussed in Chapter 8: Cohort Studies.
++
++
The proportion of new AIDS cases related to sexual transmission
between homosexual men has declined over time in the United States.
In 2002, slightly more than half of all reported AIDS cases occurred
among men who have sex with men. Use of injected drugs accounted
for almost one-fourth of cases, and heterosexual contact was responsible
for almost one-sixth of all cases. The diminished role of homosexual
contact as a contributing factor in the occurrence of AIDS in the
United States reflected avoidance of high-risk behaviors, although
there is evidence that high-risk behaviors continue in some communities.
++
HIV transmission through heterosexual intercourse continues to
increase in the United States and is the leading mode of transmission
worldwide. The practice of safe sex can prevent the transmission
of AIDS among heterosexuals, as demonstrated clearly in a cohort
study by de Vincenzi (1994). That European study included heterosexual
couples in which only one partner was HIV seropositive at the outset.
Couples were followed for an average of almost 2 years to determine the
relationship between certain sexual practices and the risk of HIV
transmission to the uninfected partner. Condom use was found to
be an effective barrier to HIV transmission. Among 124 couples who
consistently used condoms there were no episodes of seroconversion
of uninfected partners. However, among 121 couples whose condom
use was inconsistent, there were 12 instances of seroconversions
of initially uninfected partners.
++

The purpose of diagnostic testing is to obtain objective evidence
of the presence or absence of a particular condition. This evidence
can be obtained to detect disease at its earliest stages among asymptomatic
persons in the general population, a process referred to as
screening. In other circumstances,
diagnostic tests are used to confirm a diagnosis among persons with existing
signs or symptoms of illness. Ideally, a diagnostic test would correctly
distinguish affected persons from unaffected persons; unfortunately,
as is true of most diagnostic tests, assays for HIV infection are
not perfect.
++
Occasionally, a positive test result will incorrectly suggest
that infection is present in an unaffected person. This type of
outcome is referred to as a false positive, because
the positive test result was in error. Obviously, a false-positive
finding for HIV infection could be devastating to the tested individual,
so every effort must be made to keep such errors to a minimum. A
test with a very low percentage of false-positive results is said
to have high specificity (Figure 1–5).
++
++
Another type of error occurs when a test incorrectly suggests
that infection is not present (negative test result) in an affected
person. This type of outcome is referred to as a false negative, because the negative
test result was in error. A false-negative finding for HIV infection
could provide inappropriate reassurance to an infected person, thereby
delaying the start of treatment and possibly increasing the risk
of spread to other persons. A test with a very low percentage of false-negative
results is described as having high sensitivity (Figure
1–6). More detail on measures of test accuracy is presented
in Chapter 6: Diagnostic Testing.
++
++
A number of different tests for the presence of HIV infection
are available. The screening approach used most widely is to attempt
to detect antibodies to the virus. This strategy is based on two
assumptions: (1) HIV-infected persons have detectable antibodies,
and (2) persons with detectable antibodies to the virus are infected
with HIV. In practice, these assumptions appear to be reasonably
valid among patients beyond the first few months of infection. The
time required to mount an antibody response sufficient for detection
(seroconversion) varies across patients, but the vast majority seroconvert
in less than 6 months following initial HIV infection.
++
The performance of an enzyme-linked immunosorbent assay (ELISA)
test for antibodies to HIV was first reported in 1985. Among 74
patients who met the CDC clinical surveillance definition for AIDS
and had unequivocal ELISA test results, 72 (97%) had detectable
antibodies. In other words, a false-negative outcome was observed
for only 2 patients (3%). Among 261 healthy blood donors
with unequivocal ELISA test results, 257 (98%) had no detectable
antibodies (ie, a false-positive outcome was found for 4 persons [2%]).
Thus, the ELISA test was judged to be both sensitive and specific,
and it has become the most widely employed screening test for HIV infection.
++
A number of different ELISA kits are commercially available.
When false-negative ELISA results occur among high-risk individuals,
the most likely explanation is that the test was performed prior
to the development of detectable antibody levels in the immediate
postinfection period. False-positive ELISA test results have been
observed among patients with medical conditions unrelated to HIV,
such as autoimmune disorders, hematologicmalignancies, and infections with
viruses other than HIV, among patients recently vaccinated against
hepatitis B or influenza, and among patients who have received immune
globulin. Technical or human errors in performing the ELISA test
also can produce false-positive results.
++
Considering the potential for error, it is recommended that a
positive ELISA test be repeated in duplicate. If either of the follow-up
tests is positive, a supplementary test should be performed. The
most widely used confirmatory test is the Western blot. This type
of test is not recommended for screening purposes, because Western
blot can produce a substantial proportion of equivocal results among
persons who are negative to all other HIV tests.
++
The presence of infection with HIV also can be detected through
other approaches, such as the detection of viral genetic material
in plasma.
+++
Determining
the Natural History
++
After being informed of a new diagnosis, patients most frequently
ask “What will happen to me?” This question cannot
be answered with absolute certainty because of variations in outcome
across individual patients. Usually the best guidance for predictions
is the experience of other patients who are similar to the patient
in question. Even when the ultimate outcome can be predicted with some
confidence, the actual sequence of events can vary widely among
patients.
++
Consider, for example, a patient newly diagnosed as being HIV
positive. In this instance, with the advent of new treatments it
is reasonable to question whether the full syndrome of AIDS will develop,
and if it does how long it will take to occur. In attempting to
address these questions, the physician might consult published research
on the progression of HIV-related illness. Usually these data are
collected on large groups of patients. By noting the timing of critical
events for each patient (eg, date of determination of HIV infection,
development of clinical symptoms of illness, demonstrable changes
in immune function, diagnosis of AIDS, and subsequent clinical events), the
progression of the disease can be divided into phases.
++

When these events are summarized for many patients, precise and
accurate estimates of the typical sequence of events—the
natural history of the illness—can
be constructed. Some authors restrict the use of the descriptor “natural” to
situations in which medical treatment is unavailable or ineffective.
Others use the term more broadly, to indicate the typical course
of an illness, regardless of whether it can be treated effectively.
++
There are several ways to characterize the natural history of
an illness. One straightforward measure is the case
fatality, which represents the percentage of patients with
a disease who die within a specified observation period. For example,
among the 11,740 reported adolescent and adult patients diagnosed
with AIDS in 1985 in the United States, 10,946 are known to have
died before 1998. In other words, the case fatality was
++
++
The approach to determining the case fatality is illustrated
schematically in Figure 1–7.
++
++
Another method of characterizing the natural history of a disease
is to estimate the typical duration from diagnosis to death (survival time). As an illustration,
a study was conducted in a rural part of Uganda, a country with
a high prevalence of HIV infection. In this setting, in which economic
and other conditions limited treatment to simple and affordable
drugs, the natural history of HIV infection was characterized. The
study involved persons who were seropositive for HIV and a comparison
group of seronegative individuals. All subjects were identified
in 1990 and were evaluated clinically every 3 months until death or
the end of calendar year 1995, whichever came first. For the initial
3 years after seroconversion, there was no difference in survival
between the HIV-positive persons and the persons without infection.
However, by 5 years following seroconversion only 83% of
the HIV-positive persons were still alive, compared with 94% of
the seronegative persons.
++
A number of factors can affect the apparent natural history of
HIV-related illnesses. HIV infection may exist for a prolonged period
of time prior to the development of symptoms that lead to a clinical
diagnosis of AIDS. Recognition of the presence of infection during
this preclinical phase clearly depends on the availability of an
effective screening test, the sensitivity of the test to detect early
infection, and the extent to which the screening test is applied
in the population. The expectation, therefore, is that in the earliest
years of the AIDS epidemic, prior to the development and widespread
application of screening tests for HIV, the diagnosis was made at
comparatively advanced stages of infection, when symptoms already
were evident.
++
Comparisons of the survival experience of patients with AIDS
both regionally and internationally also might be distorted by differences
in the extent to which screening for HIV and CD4+ T-lymphocyte
counts are employed in the different locations.
++
Changes over time in the criteria used to diagnose AIDS could
also alter the apparent survival experience of patients with this
disease. For example, analysis of patients with HIV registered in Italy
between July 1987 and December 1991 revealed that when the 1987
CDC case definition for AIDS was applied, half of the patients survived
for 24 months more or longer. The length of survival that is met
or exceeded by 50% of the study population is referred
to as the median survival time. When
the broadened 1993 CDC case definition was retrospectively applied
to this same population, not only did a larger number of patients
meet the definition, but the median survival time was found to exceed
57 months. In other words, the population of patients who met the
1993 case definition tended to have a more favorable outcome than
the subset that met the earlier definition.
++
In estimating the natural history of HIV infection, an increasingly
important issue is the impact of more effective treatments on the
progression of illness. The benefits of improved clinical treatment
are not confined to persons in advanced stages of disease. As shown
schematically in Figure 1–8, the introduction of effective
therapy can delay the onset of AIDS after infection with HIV, as well
as extend the duration of survival after a diagnosis of AIDS. Several
studies in industrialized countries have demonstrated that the rate
of progression from HIV infection to a diagnosis of AIDS was reduced
by about 75% after the introduction of highly active antiretroviral
therapy (HAART). Similarly, the rate of progression from AIDS to
death declined by about two-thirds after the introduction of HAART.
++
+++
Searching for
Prognostic Factors
++

Analysis of survival can be employed to identify groups of patients
with unusually favorable (or unfavorable) clinical outcomes. Characteristics
that relate to the likelihood of survival are referred to as
prognostic factors. The approach to
identifying prognostic factors can be illustrated by a study conducted
by Mellors and colleagues (1997). Using data collected from the
Multicenter AIDSCohort Study of homosexual men in the United States,
the investigators evaluated factors related to the progression from initial
infection with HIV to two clinical end points: (1) the development
of AIDS and (2) AIDS-related death. A total of 1604 men were enrolled
in the study, which included a follow-up period on average of almost
10 years. Over this time period, 998 of the participants developed
AIDS and 855 died of AIDS. The design of this study is depicted
schematically in
Figure 1–9.
Figure 1–9 shows
that the study design is similar to that of the cohort study (
Figure
1–4), except that the focus is on predicting survival rather
than on determining risk factors for the onset of disease.
++
++
In the study by Mellors and associates, a number of potential
prognostic factors were assessed, including, among others, oral
candidiasis or fever, markers of immune stimulation, various lymphocyte
counts including CD4+ T-lymphocytes, and an assay
of the plasma concentration of HIV-1 RNA (ribonucleic acid—the genetic
material of the virus). The HIV-1 RNA assay provides a precise measurement
of the load of the virus circulating in a patient’s blood.
As might be expected, some association was seen between the initial
levels of the individual prognostic factors. For example, patients
with higher viral loads at the start of the study were more likely
to have fever or oral candidiasis and reduced levels of CD4+ T-lymphocytes.
Viral load also was seen as the single best predictor of the subsequent
decline over time in CD4+ T-lymphocyte levels,
as well as in the progression to AIDS and death. Specifically, when
study subjects were grouped into five ordered categories based on
plasma viral load, the 6-year probability of AIDS-related death
ranged from 1% among those with the lightest load, to 70% among those
with the heaviest load. By combining information on HIV-1 RNA concentrations
with CD4+ T-lymphocyte levels, even more effective
determination of the likelihood of disease progression could be
made (Table 1–2).
++
+++
Testing New
Treatments
++

In the United States, all new medications must be tested and
proved effective before they can be introduced into routine clinical
care. The standard approach used to evaluate treatment effectiveness
is the
randomized controlled clinical trial. The
term “controlled” means that patients (experimental
subjects) who receive the new medication are compared with patients
(control subjects) who receive either an inactive substance (placebo)
or a standard treatment if one exists. “Randomized” refers
to a method of assignment of subjects to either the experimental
or control group that is determined by chance rather than patient
preference or physician selection. This type of allocation system
is desirable because it tends to result in study groups that are
comparable with respect to important prognostic factors. Randomized
controlled clinical trials are discussed in
Chapter 7: Clinical Trials.
++
The principles of randomized controlled clinical trials can be
demonstrated by a study that has contributed to a revolution in
the treatment of HIV-infected persons. That study, published by Hammer
and colleagues in 1997, compared a standard therapy with a new experimental
treatment regimen. The standard therapy employed two drugs (zidovudine
and lamivudine), both of which are inhibitors of the HIV reverse
transcriptase. By interfering with the conversion of viral genetic material
to a form that can be incorporated into the host, these drugs limit
the replication of HIV within host cells. The experimental treatment
involved these two drugs plus another one (indinavir), which is
an inhibitor of the HIV protease. Protease inhibitors interfere
with the process of assembling viral components after replication
of HIV genetic material. The experimental therapy, therefore, involved
a simultaneous attack on two separate and distinct steps in the
process of HIV reproduction. Prior studies had demonstrated that
this combined therapy was capable of reducing viral plasma load
and raising CD4+ T-lymphocyte levels. Since favorable
responses were seen in these prognostic factors, it was reasonable
to anticipate that this combination therapy might diminish the rate
of progression of HIV-related disease.
++
Hammer and colleagues undertook a randomized controlled clinical
trial in which a standard two-drug reverse transcriptase regimen
was compared with a three-drug combined reverse transcriptase/protease
inhibition experimental treatment. The basic design of the trial
is depicted in Figure 1–10. Participants were recruited
from 40 different clinical centers throughout the United States.
The subjects were required to have documentation of HIV infection
and a CD4+ T-lymphocyte level diminished below
a predetermined level. To minimize effects of prior therapy, eligible
subjects were limited to those who had not been treated previously
with a protease inhibitor. A total of 1156 patients were randomized
between January 1996 and January 1997, with 579 assigned to the
standard therapy group and the remaining 577 assigned to the experimental therapy
group. The clinical characteristics of the two groups were similar
at the onset of treatment. After an average of about 38 weeks of
observation, the trial was terminated because of a dramatic difference
in risk of disease progression between the two groups. Within the
experimental treatment group of the study, 33 patients (6%)
progressed to AIDS or died. In contrast, within the standard treatment
group, 63 patients (11%) progressed to AIDS or died.
++
++
The results of this trial and other similar studies clearly demonstrated
the short-term therapeutic benefit of combined treatment with reverse
transcriptase inhibitors and protease inhibitors. The ethical imperative
to terminate this study early because of the substantial advantage
of combination therapy left unresolved the question of whether this
effect is sustainable over longer periods of time. Even without
data on the long-term benefits, the striking results of the studies
of combined reverse transcriptase and protease inhibition changed
the whole approach to clinical management of HIV infection. As the
search for even more effective treatments continues, randomized
controlled clinical trials will serve as the definitive approach
to establishing therapeutic superiority.