EPIDEMIOLOGY 385
stratum, so that the trial is described as stratifi ed. An exam-
ple may be the stage of disease (if distinct stages are discern-
ible), while another might be for an ethnic group. Finally,
when the numbers of strata have been decided, the patients
within each ultimate subgroup should be randomly allocated
to the treatment groups. The more strata that are used, the
more patients will be required to enable a proper balance
between treatments.FOLLOW-UPWhen a clinical trial of treatments for a cancer is conducted
within the territory of a cancer registry, and, as almost invari-
ably happens, not all eligible patients have been included in
the trial, the data from the registry concerning those patients
excluded (for whatever reason) can help to indicate whether
the trial represents a reasonably unbiased selection of the
totality of patients with the specifi c tumor in the period cov-
ered. If the registry also usually undertakes follow-up of its
cases, it can also provide this service for the clinical trial,
although the time intervals may be different, and there may
also be certain monitoring requirements for trial patients to
be added.ETHICAL CONSIDERATIONSSome clinical trials have been set up for a fi xed period of
time, or with a certain number of patients, suffi cient to eval-
uate a difference of a previously determined size between
the treated and control groups. For these, therefore, there
is an endpoint at which the outcome is assessed. Often, the
desired magnitude of the difference was not attained, and so
the trial was in that respect inclusive. Sometimes the reverse
was the case, and those on the new treatment fared so much
better that it could be considered unethical to proceed with
the trial, but instead, the control group should be given the
new treatment also. The ethical question raised is a diffi cult
one, since it has sometimes happened that the favorable dif-
ference has not continued into a longer follow-up period.
There are many other aspects to questions of ethics in the
fi eld of clinical trials, which are not appropriately consid-
ered here other than to note that they exist.SURVIVAL: LIFE-TABLE FORMOften the results of a clinical trial will be displayed in what
is described as life-table form. This can be constructed at
specifi c intervals of time (e.g., 3 months, 6 months, then
annually) or can be related directly to the times of events,
such as the death of a patient or his withdrawal from the
trial. This latter event could be because that period rep-
resents his whole experience of the trial from entry to the
current date: this situation is sometimes described as a
“censored” survival time, since it will be true that his sur-
vival is, at its least, that time, and is likely to be greater iffollow-up can continue. If withdrawal is made on grounds
related to the conditions on the trial itself, such as a side
effect of treatment, it will clearly have an effect on the ulti-
mate interpretation of the trial as a whole. These patients
may be removed from the trial, but their number and the
cause of removal must be reported.
When the life table is constructed in relation to the occur-
rence of events, the patients in a specifi c group (e.g., treat-
ment, or control) are arranged in order of their survival time
(including censored times). If there are n patients in one such
group, they will of course start at time zero ( t 0 0), as 100%
alive. When the fi rst event occurs, at time t 1 , suppose it is that
one patient has died. Then the probability of dying at that
time can be regarded as 1/ n ( q 1 ), and the probability ( p 1 ) of
surviving as its complement ( p 1 1 q 1 ). Suppose the next
event is the death of two patients at time t 2 ; the number of
patients at risk at the time of the event is ( n 1), and so q 2.
The survival probability to time t 2 is therefore the product of
p 1 and p 2 s 2. Now, at time t 3 , one patient arrives at his cen-
sored survival period. The probability of dying then is q 3 0
(since he is still alive) and thus p 3 1, and s 3 s 2 , the same
proportion. At the time q 4 is 5( n 4), since four had been
removed from the numbers continuing up to the maximum
follow-up time. The resultant curve is in the form of a series
of steps, at varying intervals (the t ’s), and represents the sur-
vival curve for that group.RETROSPECTIVE AND PROSPECTIVE METHODSThe case-control study discussed earlier can be regarded as
a clinical trial, but in the fi eld of etiology rather than in ther-
apy. There is the same call for close similarity between case
and control as between the parallel cases of the therapeutic
trial. The etiological study, however, is essentially retrospec-
tive, beginning with cases of a disease, and appropriate con-
trols, and comparing their past experiences in an attempt to
discover relevant differences that could be etiological fac-
tors. The therapeutic trial, on the contrary, is prospective,
in the sense that, having set up the two-treatment group, the
outcome of the trial is awaited and assessed. Another type
of etiological study is known as a “cohort” study because its
basis is a group or cohort of people defi ned in a certain way
and exposed to some putative hazard (an etiological factor of
a resultant disease or disability). In the realm of occupational
health, for instance, we may need to investigate whether the
work of a particular factor or substance thereof might include
some carcinogenic aspects.
We have already referred to the phenomenon of latency
as one of the diffi culties encountered when investigating car-
cinogenic hazards. This is the term used for describing the
long period elapsing between the fi rst exposure to the hazard
and the development of a recognizable tumor. In other words,
this is the time from the fi rst exposure to the time of diag-
nosis of the disease and can be applied to any disease. The
duration of the latency period varies from a few years (short
period of time), possibly even less, up to decades, like 20,
30, and over 40 years. Some of the shortest latency periodsC005_011_r03.indd 385C005_011_r03.indd 385 11/18/2005 10:25:44 AM11/18/2005 10:25:44 AM