1. Evaluation of Patients with Advanced Cancer Using the
Karnofsky Performance Status
JEROME W. YATES, MD, BRUCE CHALMER, MS, F. PATRICK McKEGNEY, MD
The Karnofsky Performance Status Scale (KPS) was designed to measure the level of patient activity and
medical care requirements. It is a general measure of patient independence and has been widely used
as a general assessment of patients with cancer. Although there is a long history of use of the KPS for
judging cancer patients, its reliability and validity have been assumed without formal investigation.
The interrater reliability of the KPS was investigated in two ways, both of which gave evidence of moder-
ately high reliability. The patients evaluated in their home were usually assigned a lower KPS score com-
pared with a similar evaluation at the same time done in the outpatient clinic. Construct validity of the
KPS was demonstrated by strong correlation with several variables relating to physical function. On-
study KPS scores accurately predicted early death, but high initial KPS scores did not necessarily predict
long survival. Patient deterioration with subsequent death within a few months could be predicted to a
limited extent by a rapidly dropping KPS. These results suggest that the KPS has considerable validity
as a global indicator of the functional status of patients with cancer and might be helpful for following
other patients with chronic disease.
Cancer 45:2220-2224, 1980.
W I T H T H E INCREASE in clinical trials designed to
evaluate chemotherapeutic agents in the treat-
ment of cancer, it became evident that some method
Methods
A group of patients with advanced cancer, all of
whom were being followed as part of the Cancer
of quantifying patients’ status relative to degree of in- Rehabilitation Project at the Vermont Regional Cancer
dependence in carrying out normal activities and self- Center, were assigned KPS scores at the time of their
care was needed. In 1948, Karnofsky and Burchena15 admission to the project.6 One criterion for admission
described a numerical scale for this purpose. This has to the study was a probable survival of three months
subsequently become known as the Karnofsky Per- to one year. The KPS scale as used is shown in Table 1 .
formance Status Scale (KPS). Although the KPS is The reliability of the KPS was evaluated two ways.
widely used for assessment of patients with cancer, First, scores were assigned to each patient by a nurse
its reliability and validity have generally been assumed and a social worker independent of each other. During
without formal investigation. In this paper we evaluate the course of the nurses’ training, simultaneous, in-
the reliability and validity of the KPS and its usefulness dependent ratings of the same patients by nurses and
as a clinical tool. physicians proved to be virtually identical; we won-
An earlier study from this institution of patients on dered if this finding would hold for patients evaluated
chronic hemodialysis demonstrated lower performance at nearly, but not exactly, the same time, by other
scores for patients when seen in the home as com- project professionals most directly concerned with
pared with the clinic.3 The conclusion was that a more evaluating the patients’ status. Both the nurse’s rating
realistic appraisal of activity was possible in the home. and the social worker’s rating were based on contact
Concurrent home and clinic evaluations were planned with the patient either in the clinic or hospital. Second,
as a part of this study. because part of the project involved periodic home
visits by the social workers to collect research data, it
From the Vermont Regional Cancer Center and the Departments was possible to obtain KPS scores based on contact
of Medicine, Epidemiology and Psychiatry, College of Medicine,
University of Vermont. with the patients in their home environment. The home
This work was supported by NCI grant R18 17868. scores could then be compared with the clinic scores
Address for reprints: Jerome Yates, MD, Vermont Regional as a further check on reliability, provided the two
Cancer Center, The University of Vermont College of Medicine,
Burlington, VT 05401. scores were sufficiently close in time. We considered
Accepted for publication May 2, 1979. a difference of one week or less to be sufficiently
0008-543X/80/0415/2220 $0.75 0 American Cancer Society
2220
2. No. 8 USING
EVALUATION PATIENTS
OF THE KPS . Yates et al. 222 1
close for comparison. There were 52 patients for whom TABLR Karnofsky Performance Status Scale
1.
two clinic (or hospital) ratings within one week of each 100 Normal, no complaints, no evidence of disease
other were available, and 50 for whom one clinic rating 90 Able to carry on normal activity, minor signs or symptoms
and one home rating within one week of each other of disease
80 Normal activity with effort, some signs or symptoms of disease
were available. 70 Cares for self. Unable to carry on normal activity or to do
In addition t o the KPS, data on a number of other active work
variables relating to both physical and psychological 60 Requires occasional assistance, but is able to care for most
of his needs
status were collected for each patient by the social 50 Requires considerable assistance and frequent medical care
worker through a structured interview. Questions con- 40 Disabled, requires special care and assistance
cerning satisfaction, happiness, and affect were used in 30 Severely disabled, hospitalization is indicated although death
not imminent
an attempt to elicit the patient’s own views of various 20 Hospitalization necessary, very sick, active supportive
aspects of their “quality of life.”1-2~4 interview
The treatment necessary
format was evolved locally from results of several 10 Moribund, fatal processes progressing rapidly
0 Dead
earlier versions. The two affect variables (positive
and negative) were slightly modified versions of scales
developed by Bradburn based on his model of the
structure of psychological well-being.2 The Pearson TABLE . Variables used in Validity Analysis
2
correlations between the KPS and these additional -~
variables were used to examine the construct validity Name Description
of the KPS. For the purpose of this analysis, data 1. Desire for food “How has your desire for food been in the
taken from all patients on the project in a single month past few days?”
(April 1978) were used (a total of 52 patients). The ad- 2. Sleep “How well have you been sleeping in the past
few nights?” (3-point scale: well, so-so,
ditional variables were described in Table 2.
or poorly)
Finally for those patients who died during the course 3. Difficulty with “Have you been having any difficulty
of the project (N = 104), it was possible t o evaluate the balance keeping your balance in the past few
degree to which the KPS was correlated with duration days?” (YesiNo)
4. Difficulty on “Have you been having any difficulty going
of survival. This evaluation was made in terms of both stairs up and down stairs in the past few days?”
the degree to which initial KPS scores were predictive (YesiNo)
of duration of survival (for all 104 patients) and the 5. Pain level “If zero is no pain at all, and 100 is more
pain than you could stand, what is your
degree to which successive KPS scores over time present level of pain?”
reflect the course of patients’ diseases. 6 . Happiness “Taken all together, how would you say
Characteristics of the patient samples used in this things are these days-would you say that
you are very happy, pretty happy, or
paper are given in Table 3. Although there was con- not too happy?” (2)
siderable overlap among the various samples, no two of 7 . Positive affect Number of “Yes” responses to five questions
them were identical; each sample consisted of those about positive feelings experienced in
the past few weeks (2)
patients for whom the necessary data were available. 8. Negative affect Number oT”No“ responses to five questions
about negative feelings experienced in
R e Lia bility the past few weeks (2)
9. Satisfaction “Which face (of seven, labeled from
with life delighted to terrible) comes closest to
Figure 1 shows a scatterplot of the nurse KPS rating expressing how you feel about your life
vs. the social worker ratings, with both ratings taken as a whole?” (1)
in the clinic or hospital within one week of each other. 10. Overall “What is your estimate of your overall
The Pearson correlation between the two sets of ratings condition condition right now, on a scale of zero
to loo?”
was .69 ( P< .001), indicating a moderate degree of
3.
TABLE Patient Characteristics
Sex Age Primary cancer (%)
Total
Reliability sample no. Female Male Mean Range Lung Breast Other
Nurse v . social worker 52 23 29 55 (22-82) 52 15 33
Clinic vs. home 50 25 25 59 (28-85) 36 20 44
Val idit y 52 30 22 59 (37-85) 31 27 42
Deceased patients 104 42 62 57 (22-8 1) 48 6 46
3. 2222 15 1980
CANCERA~~ Vol. 45
being taken within seven days of each other, is shown
in Figure 2. Although the Pearson correlation coef-
‘80
ficient between the home and clinic scores (.66, P
< .001) demonstrates a similar degree of interrater re-
liability for the two clinic scores, there was a significant
4.. tendency for the clinic scores to be greater than the
$
a L.
home scores; the average clinic score was over five
points higher than the average home score (81 .O vs.
n n e
75.3, P < .003 by a paired t-test).
Validity
Table 4 shows correlations (Pearson) between the
KPS and other variables for which data were gathered
at the same time. The KPS was strongly correlated
with the variables most closely related to physical
functioning (especially difficulty with balance and
difficulty on stairs), and less strongly (but still sig-
OV lb 2b A 40 &I $0 7b Bb 40 A nificantly) correlated with most of the variables related
SOCIALWORKER KPS RATING to psychological status. The degree to which the KPS
FIG. 1. Scatterplot of nurse vs. social worker KPS ratings (N = 52). may be useful as a predictor of survival can be seen in
Figure 3, which shows a plot of on-study KPS scores
vs. duration of survival for the 104 patients who have
interrater reliability. There was no tendency for either
died. A graph of the mean KPS scores for deceased
nurses or social workers to rate patients higher; average
patients taken at each of seven time points in the final
ratings for the two sets were within a half point of
phase of their disease is shown in Figure 4. Because
each other (70.2 for the nurses, 69.7 for the social
it was felt that the graph in Figure 4 might appear
workers, with no significant difference as evaluated by
different for patients with lung cancer as compared to
a paired t-test). We were unable to discern any tendency
other types of cancer, the mean KPS scores for patients
for certain types of patients to show greater disparities
with lung cancer over time were compared with those
between the nurse rating and the social worker rating.
for patients with other cancers. At each of the seven
A scatterplot of the clinic KPS ratings vs. the home
time points there was no significant difference (by t-test).
KPS ratings, again with the two ratings for each patient
Discussion
Many different measures are currently in use in
clinical trials, including indices of disease status,
checklists of symptoms, and signs and survival data.
Because one of the major concerns of patients with
advanced cancer is maintaining independence through
self-care, it is desirable to be able to assess the patient’s
degree of independent function. Earlier experience led
to the KPS assessments done in the home and clinic at
about the same time for comparison. Figure 2 indicates
0 40-
the KPS at home were often lower than those deter-
z mined in the clinic. This probably reflects a tendency for
d 30-
patients’ problems to seem less severe outside their
home and possibly a tendency for patients to put on a
20- “show” of well-being for clinic staff.“
- Correlations between the KPS and other variables
10
are shown in Table 4. The KPS showed a strong cor-
relation with positive affect, but not negative, which
suggests an explanation for the pattern of the cor-
HOME KPS RATING relations. In a general sample of adults, Bradburn
FIG.2 . Scatterplot of clinic vs. home KPS ratings (N = 50). found that negative affect was strongly associated with
4. No. 8 EVALUATION PATIENTS
OF USINGT H E KPS . Yatrs rt af. 2223
the level of physical symptoms.2 However, when only TAB1.F 4. Correlations (Pearson) between KPS and
Other Variables (45 5 N 5 49)
those individuals suffering from a physical illness were
considered, the level of symptoms was not correlated Correlation
with negative affect. Similarly, having advanced cancer Variable with KPS*
was a source of negative feelings for all of the pa- I . Desire for food .40 ( P < .002)
tients in our sample, and it appears that variation in the 2. Sleep .24 ( P < ,050)
degree to which their performance status was com- 3. Difficulty with balance .61 ( P < .001)
4. Difficulty with stairs .63 ( P < ,001)
promised did not make much difference in how much 5 . Pain level -.37 ( P < ,006)
negative feeling the patients reported. On the other 6. Happiness . I 2 (not significant)
hand, Bradburn found that positive affect was related 7. Positive affect .54 ( P < ,001)
8. Negative affect -.09 (not significant)
to participation in novel activities and becoming in- 9. Satisfaction with life .36 ( P < ,007)
volved with one’s environment. For our sample, the 10. Overall condition .39 ( P < .004)
degree to which patients were able to participate in
All variables except pain level and negative affect are coded
such activities was largely dependent on their physical
+
such that a higher score represents a higher level of functioning.
status; hence the correlation between positive affect
and the KPS. Thus it is possible that although the
KPS is a useful overall indicator of physical status in score was not predictive of long-term survival because
a number of aspects, it may not reflect variation in many of the patients with high initial scores died
psychological status beyond that associated with quickly. The KPS on Figure 4 reflects the progressive
physical dysfunction. The non-significant correlation deterioration of patients’ physical condition, with a
of the KPS with happiness and the relatively weak considerable drop in the last two months of life.
correlations of the KPS with overall condition and pain The KPS is designed to assess independent function
level (which certainly have both physical and psycho- and appears to have substantial validity as an indicator
logical bases), support this view. of overall physical status. In particular, the association
It is clear from Figure 3 that for this group of patients between low KPS scores and shortened life expectancy
a low KPS score was strongly associated with death suggests that the KPS may be valuable as a stratifica-
within a relatively short time. Only one of the patients tion variable in randomizing patients for clinical trials.
with an on-study KPS score of less than 50 survived In addition, improvement in KPS associated with re-
longer than six months. On the other hand, a high KPS sponse t o treatment can be used as one objective
100 .. .. . “ . . .
d . ..
.. . . . .
..
.....
.....
. . . . .. ..... +
.a
.
a
.. . -a
FIG.3. Scatterplot of on-study
KPS scores vs. days before death
a
* 5 0 . .. .a ..ow
( N = 104).
...
...
lot
d 1 1 I
6I 6 0I ~ !I 5 4 0 ~ 4 2 0 ~ 3 0 0 2 4 0 l m 1 m06 0
I 1 1 1 1 I
DAYSBEFORE DEATH
5. C A N C E R A ~ 1980
15 ~ Vol. 45
loo - measure of that response. It does not seem to reflect
variations in psychological well-being measures, other
90- than those associated with physical disability. How-
ever, its evident validity, reliability, and simplicity
make it quite helpful as a criterion in clinical trials
80-
for patients with cancer, and potentially for patients
-
Lj
ui
+I
70-
P I
I with other chronic diseases with a fatal outcome.
The evidence assembled in this study indicates that
the performance status, long assumed to be a useful
assessment of function, is in fact a fairly quantitative
60- measure with consistency among observers and a close
8 50- correlation with deterioration in function as measured
by other quantifiable parameters.
ff
x 40- REFERENCES
1. Andrews, F. M., and Withey, S. B.: Social Indicators of
3
z
30-
Well-Being. New York, Plenum Press, 1976.
2 . Bradburn, N . M.: The Structure of Psychological Well-
Being. Chicago, Aldine Publishing Company, 1969.
3 . Brown, T. M., Feins, A., Parke, R. C . , and Paulus, D. A,:
20-
Living with long-term home dialysis. A n n . Int. Med. 81:165-
170, 1974.
1
0 - 4. Chalmer, B. J : Measuring “Quality of Life“ in Patients with
Advanced Cancer. Master’s thesis, U. of Vermont, 1978.
5. Karnofskv, D. A., and Burchenal, J. If.: The clinical evaluation
I I I I I 1
I I I of chemotherapeutic agents in cancer. In Evaluation of chemo-