INTRODUCTION
It has been apparent for quite a while that cancer
disrupts the lives of patients and their families (Weisman & Worden, 1976). There are many side
effects, both psychological and physical, that accompany the disease
and its therapy that are worrisome. Therefore, an effective and
efficient screening program is invaluable for identifying at-risk
patients in a triage-like fashion so that early intervention strategies
can be employed before the onset of a crisis event (Zabora, 1998). Large batteries of tests and
assessment instruments are simply not practical for many oncology
settings, due to cost and time pressures. With a wide variety of issues
to assess and attend to, the use of brief self-report screens are the
only way such a large number of people and potential problem areas can
be identified.
One of the more dangerous physical side effects faced by cancer
patients is the increased potential for nutritional risk and cachexia,
which is a complex syndrome typically associated with a combination of
factors including decreased food intake and a variety of metabolic
dysfunctions (Fearon & Moses, 2002).
Indeed, this factor is one of the foremost causes of early death seen
in cancer patients (Inui, 1999). Although
typically associated with the advanced cancer patient, cachexia may
arise in the early stages of tumor growth before there are any other
signs of malignancy (Kern & Norton,
1988).
With the need to identify multiple patient difficulties with the
least amount of resource depletion, there is a need for using screening
instruments for multiple purposes. The idea of utilizing portions of a
depression scale such as the Zung Self-Rating Depression Scale (ZSDS)
for multiple purposes has been explored by our group in previous
studies. For instance, we examined the criterion validity and
sensitivity and specificity of a single item to rapidly screen patients
in ambulatory oncology clinics for cancer-related fatigue (Kirsh et al., 2001). In an effort to expand the
utility of the scale as a screen for other symptoms while
simultaneously screening for depression, we were interested in
examining the utility of single items, such as the fatigue item, on the
scale as a separate screen. The fatigue item from the ZSDS, item 10,
reads: “I get tired for no reason” and was found to be
highly correlated with much longer, existing measures of fatigue.
Similarly, we tried to address the need for screening to identify the
nature of what oncology patients often complain about, saying that
their “mind does not seem to be clear.” This subjective
perception, sometimes referred to as “chemo brain,” may be
due to situational stressors, psychological disorders, organic factors,
or effects of neurotoxic medications. Cognitive decline not only
diminishes quality of life, but can also interfere with a
patient's ability to make decisions regarding complex treatment
issues. This study (Kibiger et al., 2003)
investigated the utility of using item 11 (“My mind is as clear
as it used to be”) of the ZSDS as a cognitive screen. Although
the results were mixed, we concluded that although the perception of
cognitive impairment is common in cancer patients, there may be
problems in interpreting the nature of these complaints, particularly
in separating them from their depressive preoccupation.
Finally, we attempted to examine issues of criterion validity and
detection of insomnia utilizing a single item from the ZSDS as a means
to rapidly screen cancer patients in ambulatory oncology clinics (Passik et al., 2003). The sleep item reads “I
have trouble sleeping through the night” and is rated on a
4-point Likert scale ranging from none or little of the time
to most or all of the time. A sample of oncology patients was
administered the ZSDS and further evaluated with the Pittsburgh Sleep
Quality Index (PSQI). The sensitivity and specificity of the ZSDS sleep
item was investigated utilizing various cutoffs as they predicted
results of the PSQI, which were used as a criterion. Results revealed
poor sensitivity and specificity of the ZSDS single-item screen for
detecting insomnia in cancer patients, and that the relationship
between insomnia and depression is more complicated than anticipated.
We concluded that the use of this single item, or perhaps any single
item, as a means of screening for sleep disturbances in cancer patients
may be problematic and a better understanding of insomnia and its
measurement are worthwhile areas of study.
The present study was an attempt to gather initial data to determine
if two items from the Zung Self-Rating Depression Scale could be used
as short screening items for detecting cachexia in cancer patients.
Specifically, ZSDS item 5, “I eat as much as I used to,”
and item 7, “I notice I am losing weight,” were chosen for
the study. A random chart review was conducted of patients who had
completed a variety of screening measures in our cancer network.
METHODS
Participants and Procedure
An a priori simplified power analysis was conducted to
determine the sample size requirement for the correlational analyses of
this study. Assuming an alpha of 0.05, a medium effect size (ES =
r = 0.50), and a power of 0.80, 50 total subjects would be
required (Cohen, 1992).
A random sample of 50 charts was drawn for review from the population
of patients enrolled in 1 of 31 Community Cancer Care Inc. clinics in
urban and rural areas throughout Indiana. Any patients undergoing
treatment for malignancy were eligible to participate if they were able
to read and write English, were over 18 years of age, and did not give
evidence of cognitive impairment severe enough to preclude giving
informed consent.
Instruments
Demographics
Basic demographic information was gathered on the sample including
gender, height, weight, appetite levels, and nausea as well as albumin
and lymphocyte levels. In addition, the BMI was calculated for each
patient to assess ratios of overall body size as a marker of health
risk factors (Bray & Gray, 1988).
Scored Patient-Generated Subjective Global Assessment
The PG-SGA is a combination self-report and clinician-derived scale
to assess nutritional risk in patients (Ottery,
1994). It yields a total score that can be used to determine
what interventions, if any, a patient may need. Scores on the lower end
of the spectrum indicate better functioning and scores above 9 indicate
severe nutritional risk.
Zung Self-Rating Depression Scale
The ZSDS (Zung, 1967a,
1967b) is a 20-item self-report measure
of the symptoms of depression. Subjects rate each item regarding how they
felt during the preceding week using a 4-point Likert scale, with 4
representing the most unfavorable response. After correcting for items that
are reverse-scored, the 20 items are summed to create a total score. Scores
are not meant to offer strict diagnostic guidelines but rather denote
levels of depressive symptomatology that may be of clinical
significance. The ZSDS has been shown to be both valid and reliable,
with high internal consistency of 0.84 and test–retest
reliability of 0.86 (Tate et al., 1993; Dugan et al., 1998). Gabrys and Peters (1985) reported that the ZSDS had an interrater
reliability of 0.89, internal consistency reliability of 0.88
(Cronbach's alpha), mean item-total correlations of 0.85,
split-half reliability of 0.94, and showed preliminary evidence of
discriminant validity, significantly differentiating between
nondepressed and depressed clients (t = 30.85, p <
0.0001). For purposes of this study, special attention was given to
items 5 (“I eat as much as I used to”) and 7 (“I
notice I am losing weight”) as predictors of cachexia.
Statistical Analyses
A series of descriptive statistics, chi squares, one-way ANOVAs with
Bonferroni post hoc analyses, and sensitivity and specificity
statistics were calculated. Specificity refers to a statistic designed
to elucidate the number of correctly classified subjects without
cachexia, via subthreshold scores on the ZSDS items 5 or 7, divided by
the total number of subjects without fatigue (as measured by the
PG-SGA) multiplied by 100. Thus, the statistic offers the percentage of
cases wherein the ZSDS items correctly identified people who are not at
nutritional risk (i.e., its ability to identify true negatives).
Sensitivity refers to a statistic calculated by the number of correctly
classified subjects with cachexia, via high scores on the ZSDS items,
divided by the total number of true subjects with nutritional risk or
cachexia (as measured by the PG-SGA) multiplied by 100. Thus, the
statistic yields the percentage of cases wherein the ZSDS items
correctly identified people who are cachcectic (i.e., its ability to
identify true positives).
RESULTS
To protect the privacy of patients chosen for the chart review study,
and due to the preliminary nature of the work, only limited demographic
data were gathered by a research assistant. However, based on our prior
work with this population (Dugan et al.,
1998; Kirsh et al., 2001; Kibiger et al., 2003; Passik et
al., 2003), we know that our patients typically fall into an
average range of 59–63 years old, are married (67–70%), are
largely Caucasian (∼90%), and usually have breast (19–36%),
lung (10–23%), or colon (10–13%) cancer. The present sample
was comprised of 33 women (66%) and 17 men (34%). Average weight for
the group was 165.53 lbs. (SD = 38.63) and average height was
65.38 in. (SD = 3.60). These measurements yielded an average
BMI score of 27.04 (SD = 5.38, range = 14–48). According
to the accepted BMI guidelines (Bray & Gray,
1988), this means that 4% (n = 2) of the sample was
underweight, 36% (n = 18) were normal, 40% (n = 20)
were overweight, 16% (n = 8) were obese, and 4% (n =
2) fell into the extremely obese category. Patients' charts were
also examined 6 months later and weights were found to average 164.73
lbs. (SD = 38.75) where data was available (n = 32).
Albumin and lymphocyte levels were also recorded. Patients had an
average albumin score of 3.62 (SD = 0.45) and lymphocyte
levels of 1.39 (SD = 0.60).
Patients were also routinely asked a number of questions pertaining
to their health and nutrition and these responses were noted. When
asked about their current eating habits, 12% (n = 6) reported
eating very little solid food and 24% (n = 12) reported eating
solid foods but in smaller quantities, whereas the remainder (70%,
n = 35) reported no changes in food consumption. Twenty
percent (n = 10) reported having no appetite and the remainder
(80%, n = 40) reported normal appetite. The patients were also
asked about a number of symptoms that might affect their eating
behaviors. Patients reported a number of symptoms including feelings of
quickly becoming satiated (8%, n = 4), changes in perception
of taste (6%, n = 3), dry mouth (6%, n = 3), nausea
(4%, n = 2), diarrhea (4%, n = 2), vomiting (2%,
n = 1), constipation (2%, n = 1), difficulty
swallowing (2%, n = 1), pain (2%, n = 1), and mouth
sores (2%, n = 1).
Although not the main focus of this study, the total score on the
ZSDS was examined. Patients had a mean score of 32.5 (SD =
7.24) with a range from 22 to 52. Only two of the patients (4%) scored
above the threshold for depression on the ZSDS. When examining the
relationship of the ZSDS to the PG-SGA, however, there was a
significant relationship (r = 0.63, p < 0.01),
indicating that a greater presence of depressive symptomatology was
associated with greater nutritional risk. Also, as expected, item 5
(r = 0.62, p < 0.01) and item 7 (r =
0.58, p < 0.01) were significantly related to the ZSDS as a
whole.
Item 5: “I Eat as Much as I Used to.”
A series of analyses were conducted to explore the utility of item 5
to identify patients with problems of cachexia. Chi square analyses
were not significant between item 5 and gender or activity level, and
one-way ANOVAs were not significant between the item and either albumin
or lymphocyte levels. The relationship was significant, however,
between item 5 and initial weight (F3,45 = 6.06,
p < 0.001), with post hoc analyses indicating a
significant difference between those answering “none of the
time” (mean weight = 140.07 lbs.) and those answering “some
of the time” (mean weight = 195.83 lbs.; mean difference = 55.77
lbs., p < 0.05). Similarly, item 5 was significantly
related to weight at 6-month follow-up (F3,27 =
4.16, p < 0.05), with post hoc analyses showing
differences remaining between the “none of the time”
response (mean weight = 126.52 lbs) and those answering “some of
the time” (mean weight = 186.46 lbs.; mean difference = 59.94,
p < 0.05). Item 5 was also predictive of BMI score
(F3,46 = 2.89, p < 0.05) and
nutritional risk on the PG-SGA (F3,45 = 5.80,
p < 0.01). Post hoc analyses indicated differences
in the PG-SGA between the “none of the time” response (mean
PG-SGA = 4.2) and the “most of the time” response (mean
PF-SGA = 0.74; mean difference = 3.46, p < 0.05). Using
item 5 as a screen yields a sensitivity of 33% and specificity of
97%.
Item 7: “I Notice I Am Losing Weight.”
The same group of analyses was repeated to examine the utility of
item 7 for identifying patients with cachexia. Chi square analyses were
not significant between item 7 and gender or overall activity level.
Likewise, a series of one-way ANOVAs were not significant between item
7 and initial weight, weight at 6 months follow-up, BMI, albumin
levels, or lymphocyte levels. However, item 7 was a significant
predictor of nutritional risk as measured by the PG-SGA
(F3,46 = 6.01, p < 0.01). Subsequent
post hoc analyses identified the significant difference
occurring between those reporting “a good part of the time”
(mean PG-SGA = 5.75) and those reporting “none of the time”
(mean PG-SGA = 1.36) to item 7 (mean difference = 4.39, p <
0.05). Using item 7 as a screen yields a sensitivity of 40% and
specificity of 81%.
Combination of Items
It was also of interest to determine if the two items could be
combined to create a short 2-item screen. For purposes of exploration,
scores of 2–5 were chosen as indicating no problems whereas
scores of 6–8 (indicating a “good part of the time”
or more on both items) were deemed to indicate cachexia or nutritional
risk. According to this scale, we would predict eight of the patients
(16%) to have cachexia and the remainder to be normal (84%, n
= 42). Using this scale, chi square analyses found no relationships to
gender, and ANOVAs were not related to weight, albumin, or lymphocytes.
The two-item scale, however, maintained the individual items'
significant relationship to the PG-SGA (F1,48 =
13.99, p < 0.001). Using this as the criterion for
identifying nutritionally at-risk patients, the two-item screen yields
a sensitivity of 50% and specificity of 88%.
DISCUSSION
Developing quick and useful screens for identifying patients at
nutritional risk for developing cachexia is an important and worthwhile
endeavor. We conducted a promising, albeit preliminary, chart review
study that offers some evidence that developing just such a quick
screening tool is feasible. This offers the possibility of identifying
a larger number of patients at risk, and at the very least will help to
create more attention and awareness around this compelling aspect of
symptom management. It is our hope that further studies will help to
solidify the use of single- or even two-item screens that can be easily
implemented into even the busiest of clinics.
Our results indicated that item 5 of the ZSDS (“I eat as much
as I used to”) was significantly related to a large number of
variables associated with nutritional risk and cachexia, including
scores on the PG-SGA, BMI, and weight at both time periods measured.
Item 7 (“I notice I am losing weight”) was less robust
overall but added to the overall value of the predictive power of
examining the nutritional risk factors assessed by the PG-SGA when
combined with item 5. Therefore, it may offer some unique information
and can perhaps help to identify a larger cohort of patients at risk.
However, it must also be noted that although the combined items'
specificity was rather impressive, overall sensitivity was somewhat
poor.
Certainly, the present study is not without its limitations. Chart
reviews must be carefully examined and conclusions must always be of a
tentative nature due to the possibility of errors or omissions in
charting. In addition, the sample is drawn from a Midwestern sample
that may not be indicative of cachexia and weight issues in cancer
patients across the nation at large. It must be noted that the BMI
scores placed 60% of the sample in the overweight, obese, or extremely
obese categories. Finally, although adequate for this purpose, studies
with larger sample sizes and of a prospective nature are needed to
either replicate or refute these findings.
