and still has a specificity greater that 0.95 after 25
repetitions of the study. The 2
2s
rule, which requires
two consecutive study results to differ by more than
2 standard deviations from the preceding average
value, and the 4
1s
rule, which requires four consecu-
tive study results to differ from the preceding
average value by more than 1 standard deviation,
have specificities comparable to the 1
3s
rule.
Although these monitoring rules have similar speci-
ficities, they differ somewhat in terms of sensitivity
with the 4
1s
rule showing the best performance and
the 2
2s
rule the worst (Parvin 1991).
Other statistical approaches that have been
explored as a basis for the design of clinical
monitoring rules include time series modeling using
ARIMA (Crabtree
et al.
1990) and the CUSUM test
(Piccoli
et al.
1987).
MONITORING DISEASE ACTIVITY
Disease activity is monitored primarily with the
intention of detecting an increase in activity so that
therapy may be instituted or intensified as early as
possible and the clinical consequences of the worsen-
ing of the disease minimized. Laboratory studies
that may be used as markers of disease activity
include direct measures of disease activity, such as
plasma tumor markers in certain cancers, measures
of the physiologic function of organs involved by the
disease, such as pO
2
and pCO
2
in pulmonary disor-
ders, and measures of inflammatory or immune
response to the disease, such as specific antibody
titers in certain infectious diseases. Monitoring the
impairment of an organ's physiologic function is a
particularly good way to monitor disease activity,
when possible, because it simultaneously quantifies
the clinical impact of a change in activity.
The advantages of monitoring disease activity
using direct markers of disease activity or measures
of inflammatory or immune response to the disease
are two-fold. First, for many diseases, there is no
physiologic function marker that correlates with
disease activity. Second, changes in these markers
may precede the clinical effects of the change in
activity. Thus, these markers can in some cases
provide a lead time for a therapeutic response to the
change. The major disadvantage of the use of these
markers is that the relationship between disease
activity and the magnitude of the marker is
sometimes unpredictably nonlinear and often varies
over time. The relationship can even break down
altogether as the disease enters a different pathobio-
logic stage. For example, a tumor metastasis may
not elaborate the marker substance produced by the
primary tumor so metastatic progression of the
tumor may not be associated with any change in the
plasma concentration of the tumor marker.
Because monitoring of disease activity is a serial
process with multiple specimens obtained over time,
a monitoring rule must be employed to determine
when a significant change in marker concentration
has occurred. The rules to consider are exactly the
same as those for the monitoring of physiologic
status. The application of the 1
3s
, 2
2s
, and 4
1s
monitoring rules is discussed as an example.
The results of monthly monitoring of plasma
carcinoembryonic antigen (CEA) concentrations in a
patient who had primary surgery for breast cancer
are shown in Figure 5.6 (the data are taken from
Winkel
et al.
1982). CEA concentrations were
monitored because some of the patients who subse-
quently go on to experience recurrence of their
cancer show a significant elevation in CEA concen-
trations before disease progression is evident clini-
cally. This patient had a clinically overt recurrence
of her tumor soon after the last monitoring specimen
was obtained. As is easily appreciated from the
graphical presentation of the data, the CEA concen-
trations appear to be fairly constant over the first 13
months, so those results can be used to calculate
avg
indiv
and var
indiv
; the values are 1.3 µg/L and 0.032
µg
2
/L
2
, respectively The critical value for the 1
3s
monitoring rule is 1.8 µg/L. The CEA concentra-
tion at month 14 is 3.5 µg/L, so the concentration
increase evident in the result can be considered
significant even though the result is far below the
population-based critical value of 7.4 µg/L. The
population-based critical value was calculated using
postoperative CEA concentrations seen in patients
who, after long-term follow-up, didn't experience a
tumor recurrence. The subsequent results are all
also greater than 1.8 µg/L so there is a persistent
significant elevation in CEA concentration. The
critical value for the 2
2s
rule is 1.6 µg/L which is
exceeded by the consecutive results at months 14 and
15 as well as and all subsequent consecutive result
pairs. The 4
1s
rule's critical value is 1.5 µg/L which
is exceeded by the consecutive results at months 14
through 17. Thus, all three monitoring rules demon-
strate a significant increase in the CEA
concentration. Any of the rules could have been
used to detect the CEA elevation in this patient but
Monitoring
5-7
