Clinical Cancer Research Vol. 6, 4713-4718, December 2000
© 2000 American Association for Cancer Research
White Blood Cell Count: A Prognostic Factor and Possible Subset Indicator of Optimal Treatment with Low-Dose Adjuvant Interferon in Primary Melanoma
Pauline de La Salmonière,
Jean-Jacques Grob,
Brigitte Dreno,
Michèle Delaunay and
Claude Chastang
Département de Biostatistique et Informatique Médicale, Hôpital Saint-Louis, Paris [P. d. L. S., Cl. C.]; Service de Dermatologie, Hôpital Sainte Marguerite, Marseille [J-J. G.]; Service de Dermatologie, Hôpital Hôtel-Dieu, Nantes [B. D.]; and Service de Dermatologie, Hôpital Pellegrin, Bordeaux [M. D.], France.
 |
ABSTRACT
|
|---|
IFNhas recently been recognized as an adjuvant therapy to surgery in
melanoma patients. A major issue is to select patients who will benefit
from this therapy and to avoid toxicity in those who will not respond.
The aim of this exploratory analysis was to identify the predictive
factors of response to IFN.
The French cooperative group has recently shown that adjuvant therapy
of melanoma patients with low-dose IFN provides a benefit on
disease-free interval (DFI). Using this database, predictors of DFI
were investigated using Cox models and treatment-covariate interactions
were sought.
Gender, age, Breslow thickness, and baseline WBC count, given an
IFN-WBC interaction, were independent predictors of DFI. Baseline
WBC count was the only variable for which there was an interaction with
IFN, whatever the Breslow: patients with low WBC count (<6.8 x 109/liter = median) did not benefit from IFN
(HR=1.27(95%CI: 0.84–1.91); P = 0.26) whereas the
DFI of patients with high WBC was prolonged (P =
0.0001) with a hazard ratio of 0.50 (95% confidence interval,
0.35–0.71). The estimated values of WBC count for which IFN was
significantly superior to no-treatment were those 
7.2 x
109/liter. The baseline WBC count was correlated to
baseline neutrophils but not to Breslow thickness or to time since last
melanoma surgery.
IFN prolonged DFI in patients with a high WBC count but not in those
with a low WBC count. The results of this exploratory analysis, if
confirmed by other studies, may help to identify patients who are most
likely to benefit from IFN.
|
INTRODUCTION
|
|---|
It is only recently that IFN has been recognized as an
adjuvant therapy to surgery in patients with high-risk cutaneous
melanoma. A 1-year high-dose IFN -2b regimen has been shown to
improve the disease-free and overall survival of
AJCC1
stage IIB and III melanoma patients (1)
. Furthermore, the
beneficial effect on DFI of adjuvant low doses of IFN -2a in
patients with cutaneous melanomas thicker than 1.5 mm without
clinically detectable node metastases has recently been shown by our
group (2)
and confirmed by an ongoing study
(3)
. However, the exact mechanisms by which IFN exerts
antitumor effects are not clearly known (4)
, nor are the
characteristics of patients who will respond well defined. Thus, as
IFN is emerging as an adjuvant therapy in high-risk cutaneous
melanoma patients, it is becoming critical to identify clinical,
immunological, or molecular features that will enable the selection of
patients who are likely to benefit from IFN therapy (5)
,
thereby limiting toxicity and impairment of quality of life in patients
who will not benefit from IFN.
The aim of this exploratory study is to investigate, in the high-risk
primary melanoma patients of the French database (2)
, the
baseline clinical, histological, and biological characteristics that
are predictive of relapse and the characteristics that are associated
with a beneficial response to IFN.
|
PATIENTS AND METHODS
|
|---|
Patients.
The 489 eligible patients enrolled in a randomized clinical trial,
which was recently published by our group, were studied
(2)
. Patients had a high-risk primary cutaneous melanoma
(Breslow thickness, 1.5 mm) without clinically detectable node
metastases; after tumor resection, they were randomized to receive
either 3 x 106 IU of IFN three times weekly
for 18 months or no treatment. Using a sequential procedure, IFN
demonstrated a significant benefit for DFI, which was the primary end
point (P = 0.038). After a median follow-up of 5.0
years, 487 patients were evaluable for relapse-free interval. There
were 100 relapses among the 244 IFN-treated patients and 119 among the
243 evaluable control patients; in IFN-treated patients, there was a
significant prolongation of relapse-free interval (P =
0.035, log-rank test) with an estimated HR of 0.75 (95% CI,
0.57–0.98) and a clear trend toward an increase in overall survival
(P = 0.059, log-rank test) compared with control
patients.
Statistical Methods.
The variables that were studied were first those usually recognized as
having a prognostic significance in melanoma patients (6)
,
namely gender, age, Breslow thickness, site (back, chest, face, arm,
leg, hand-foot, and other), histology (superficial spreading, nodular,
acral, and other), and Clark level (four categories because there were
no in situ melanomas). Ulceration was not systematically
recorded and was, therefore, not taken into account in the analysis.
Secondly, the prognostic value of the following baseline biological
variables was evaluated: WBC count, hemoglobin, platelets, bilirubin,
alkaline phosphatase, liver enzymes aspartate aminotransferase and
alanine aminotransferase, 
glutamyl transferase, creatinine, and
urea. Gender was coded 0 for female and 1 for male; variables with more
than two categories were coded with dummy variables (site, histology,
and Clark level). The functional form of each continuous variable was
investigated (7)
; age and Breslow were coded as continuous
variables over two separate intervals with cutoffs at 50 years and 4
mm, respectively, and the other continuous variables and, in
particular, WBC count, were kept as such. For illustrative purposes
(Figs. 1
2
3)
, Breslow and WBC count were dichotomized with cutoffs at 4 mm and
6.80 x 109/liter (median value),
respectively.
View larger version (19K):
[in this window]
[in a new window]
|
Fig. 1. DFI according to baseline WBC counts
(cutoff = median value = 6.8 x 109/liter)
and Breslow thickness in the whole study sample.
|
|
View larger version (18K):
[in this window]
[in a new window]
|
Fig. 2. A, DFI of patients according to
their treatment and their baseline WBC count. B, DFI of
patients with baseline WBC counts 6.80 x 109/L
(high WBC). C, DFI of patients with baseline WBC counts
below 6.80 x 109/L (low WBC).
|
|
View larger version (27K):
[in this window]
[in a new window]
|
Fig. 3. A, DFI according to baseline WBC
counts (cutoff = median value = 6.8 x
109/liter) and Breslow in the control group.
B, DFI according to baseline WBC counts (cutoff =
median value = 6.8 x 109/liter) and Breslow in
the IFN group.
|
|
DFI was the end point. Treatment-covariate interactions were sought. A
quantitative interaction arises when there is a variation in the
magnitude, but not in the direction, of treatment effects among
subsets. Quantitative interactions are said to occur when one treatment
is superior for some subsets of patients and the alternative treatment
is superior for other subsets (8)
. The search for
treatment-covariate interactions was carried out using the method
proposed by Byar and Green (9)
. The variables that had a
prognostic significance in the control group, the IFN group, or both
combined were determined (9)
: DFI was estimated by the
Kaplan Meier method (10)
, and comparisons between groups
were made by the log-rank test; the variables that had a prognostic
significance were introduced in a Cox model (11)
, and
those that were selected after a step-down procedure with a
significance level of 0.05 were considered for the final model. First,
the Cox model which predicted relapse in the whole study population was
selected by introducing in a Cox model the prognostic variables found
to be significant in at least one of the three groups, the treatment
term (1 for IFN and 0 for no treatment) and the corresponding
treatment-variable interaction terms; second, interactions between
baseline characteristics were sought. Quantitative interactions were
formally tested with a test for heterogeneity (8)
; an
overall test for qualitative interaction was performed for significant
interactions (8)
. The values of WBC count for which IFN
treatment led to a significantly prolonged DFI compared with no
treatment were estimated (12)
. For the purpose of
comparing DFI by change in the WBC count at 1 month, a landmark
analysis (13)
was conducted.
The Spearman nonparametric test was used to determine whether there was
a correlation between continuous variables, and the Wilcoxon rank test
was used to compare distributions of continuous variables. All
Ps corresponded to two-sided statistical tests, and a
significance level of 0.05 was used to indicate statistical
significance. The SAS (SAS Institute, Inc., Cary, NC) and Splus
(StatSci, Seattle, WA) softwares were used.
|
RESULTS
|
|---|
The clinical and biological variables found to have a prognostic
significance for the control group, the IFN group, or both combined are
shown in Table 1
. In the multivariable analysis, Breslow thickness was a predictive
factor of relapse in all three groups. Among the baseline biological
variables, only WBC count was predictive of relapse and that only in
the control group, an increment of 1 x
109/liter WBCs being associated with a HR of
relapse of 1.23 (i.e., with an increased risk of relapse of
23%).
View this table:
[in this window]
[in a new window]
|
Table 1 Prognostic factors of the disease-free interval
for the control group, the IFN group, and both combined in univariable
and multivariable analysis with estimated HR ratio of
relapse
|
|
The Cox model that predicted DFI when considering interactions of
prognostic factors with treatment (Table 2)
fits the main effects of gender, age, Breslow, WBC count, and
treatment as well as the WBC treatment interaction term. The prognostic
value of Breslow thickness and WBC count, using 4 mm, which is the
cutoff value between AJCC stage IIA and IIB for Breslow thickness and
the median value of 6.80 x 109/liter as
cutoff value for WBCs, is presented for the whole study sample (Fig. 1)
. The time to 25% relapse was, among patients with Breslow <4 mm,
2.1 years (95% CI, 1.5–3.6) in low WBC count patients and 1.5 years
(95% CI, 1.2–2.5) in high WBC count patients, and among patients with
Breslow 4 mm, was 1.0 year (95% CI, 0.5–2.1) in low WBC count
patients and 0.5 year (95% CI, 0.4–1.2) in high WBC count patients.
The quantitative WBC treatment interaction was highly significant
(P < 0.005; test for heterogeneity) (8)
;
however, the Gail and Simon overall test for a qualitative interaction
between treatment and WBC count was not significant. The interaction
between baseline WBC count and treatment is shown in Fig. 2
. There was a significant difference in the DFI of the four groups
determined by treatment and baseline WBC count (P =
0.0001, log-rank test; Fig. 2
A). Among patients with
baseline WBC count 6.80 x 109/liter (high
WBC), the DFI was highly significantly prolonged [HR, 0.50 (95% CI,
0.35–0.71); P = 0.0001, log-rank test] in the
IFN-treated group compared with the control group (Fig. 2
B),
whereas there was no treatment effect among patients with baseline WBC
count below 6.80 x 109/liter (low WBC)
[HR, 1.27 (95% CI, 0.84–1.91), P = 0.26; Fig. 2
C]. Because the WBC treatment interaction was found with
WBC count coded as a continuous variable, we attempted to estimate the
WBC count cutoff value (i.e., the "low" values for which
there was no treatment effect, and the "high" values for which
there was a beneficial effect of IFN). Using the method proposed by
Shuster and van Eys (12)
, the estimated values of WBC
count for which IFN was significantly superior to no treatment were
those 7.22 x 109/liter.
A different way of illustrating the WBC-IFN interaction, whatever the
Breslow thickness, is shown in Fig. 3
. In control patients, WBC count had a prognostic value (Fig. 3
A and Table 1
), whereas WBC count had no prognostic value
in the IFN group (Fig. 3
B and Table 1
), the DFI curves of
high WBC IFN-treated patients being similar to those of low WBC count
control patients. Among patients with Breslow thickness <4 mm, IFN
improved the DFI of patients with a high baseline WBC count [HR, 0.52
(95% CI, 0.34–0.79); P = 0.002, log-rank test] but
not that of patients with a low WBC count [HR, 1.32 (95% CI,
0.82–2.13); P = 0.25]. Among patients with Breslow
4 mm, there was a trend toward an increased DFI in high baseline WBC
IFN-treated patients compared with control patients [HR, 0.56 (95%
CI, 0.29–1.08); P = 0.08], whereas in low baseline
WBC counts, IFN did not improve DFI [HR, 0.96 (95% CI, 0.43–2.13);
P = 0.91]. When the baseline NC was considered instead
of WBC count, similar results were obtained, with the NC being a
predictor of DFI (P = 0.0001) and there being a
significant interaction between NC and IFN (P =
0.0001). However, when baseline lymphocyte count or eosinophil count
were considered, no predictive value of these counts and no significant
interaction between them and IFN were found (data not shown).
Baseline WBC count, mean (±SD) 7.2 (±1.96) x
109/liter, was positively correlated only with
baseline neutrophils (Spearman correlation coefficient,
r = +0.91; P = 0.0001), lymphocytes
(r = +0.43; P = 0.0001), and platelets
(r = +0.26; P = 0.0001) and negatively
correlated only with baseline age (r = -0.14;
P = 0.001) and bilirubin (r = -0.12;
P = 0.01). The baseline WBC count was not correlated
with Breslow thickness [median, 2.51 mm (range, 1.5–20.0);
r = +0.04; P = 0.32] nor with
eosinophils (r = +0.02, P = 0.72).
There was no significant difference in WBC count between site, Clark
level, or gender subgroups. The baseline WBC count, which was measured
at a median of 19 (range, 0; 52) days after last melanoma surgery, was
not correlated to time since last melanoma surgery (r =
-0.06; P = 0.21).
The mean (±SD) change in WBC count at 1 month was, among patients with
WBC counts <6.8 x 109/liter, +0.24
(±1.22) x 109/liter in control patients
and -1.22 (±0.95) x 109/liter in
IFN-treated patients, and, among patients with WBC 6.8 x
109/liter, -0.78 (±2.05) x
109/liter in control patients and -2.96
(±2.32) x 109/liter in IFN-treated
patients. Among the 234 evaluable IFN-treated patients [one relapse
the first month and nine (4%) missing data], the relative decrease in
WBC count (median, -30%; range, -81, +46) was nearly significantly
predictive of relapse [HR, 1.011 (95% CI, 0.999–1.022),
P = 0.06], patients having greater decreases of WBCs
being at lower risk of relapse.
In the final model to predict DFI, significant interactions between WBC
count and Breslow thickness and between gender and age were found
(Table 3)
, although the test for heterogeneity (8)
of these two
interactions was not significant (0.05 < P <
0.10 and 0.10 < P < 0.25 respectively).
Regarding the age-gender interaction, among patients younger than 50
years, female patients had a prolonged DFI compared with males
(P = 0.003); however, for patients above 50 years,
gender had no prognostic value (P = 0.29).
View this table:
[in this window]
[in a new window]
|
Table 3 Multivariable Cox model with significant
interactions to predict disease-free interval in the whole study
population
|
|
|
DISCUSSION
|
|---|
In this study, gender, age, Breslow thickness, and WBC count,
given that there was an interaction between WBC count and IFN, were
predictive of relapse. The baseline WBC count was the only clinical,
histological, or biological characteristic for which there was an
interaction with treatment and which predicted a beneficial response to
adjuvant low-dose IFN; patients with low baseline WBC counts
experienced no benefit from IFN, whereas the DFI of patients with high
baseline WBC counts was highly significantly prolonged in IFN-treated
patients compared with control patients, and this, whatever the Breslow
thickness. Similar results were found when the baseline NC was
considered instead of WBC count.
The prognostic value of Breslow thickness, age, and gender are well
known (6)
. To our knowledge, the prognostic value of
clinical and histological factors in primary melanoma has been
extensively studied, whereas the prognostic significance of baseline
biological variables has received little attention (6
, 14)
. In our study, the high prognostic value of baseline WBC
counts in the no-treatment group, high values being associated with a
poor prognosis, was unexpected.
In this study, the baseline WBC count was analyzed as a continuous
variable because this coding best modeled the influence of WBC count on
DFI. Although the median value of WBC count (6.8 x
109/liter) was used as a cutoff value in the
figures, the values of baseline WBC counts for which IFN was estimated
to be significantly superior to no treatment were those 7.2 x
109/liter. These values were determined on a
population of resected AJCC stage IIA–IIB melanoma patients whose
inclusion criteria were, among others, WBC count 4.0 x
109/liter. However, only very few patients were
not included in the study because of WBC counts below 4.0 x
109/liter; the WBC count of our study population
is, therefore, presumably representative of the WBC counts of patients
with primary melanoma thicker than 1.5 mm without clinically detectable
node metastases.
To our knowledge, the WBC-IFN interaction has not been reported
previously. We believe that this finding merits attention. Indeed, the
difference in DFI between the IFN and control group among patients with
high baseline WBC counts was highly significant. Furthermore, the
baseline WBC count represents a basic approach to the inflammatory and
immunological status that may have some relevance to the mechanisms
through which low doses of IFN -2a exert their effects. The precise
mechanisms by which IFN compromises tumor cell survival are not
clear, but seem to involve nonnecrotic pathways such as effects on
tumor cell differentiation, induction of apoptosis, or possibly
senescence; immunological mechanisms may be important, particularly
with small tumor volume (4)
. The baseline WBC count may be
explained by the inflammatory reaction due to the surgical removal of
the primary tumor, as is seen by the slight decrease at 1 month of the
WBC counts of control patients (-0.78 x
109/liter), but the baseline WBC count was not
correlated with the time since the last melanoma surgery. To account
for the relation between WBC count and prognosis, one could have
speculated the baseline WBC count to be related to regional lymph node
involvement; indeed, as the primary tumor thickness increases, so does
the likelihood of involvement of sentinel lymph nodes
(15)
. However, the baseline WBC count was not
significantly associated with Breslow thickness or Clark level in this
study, thus rendering this hypothesis unlikely. Two other hypotheses
could account for the predictive role of the WBC count. First, the WBC
count may be a surrogate measure for the stimulation of the
inflammatory pathways and the immune system by the melanoma. Second,
the WBC count may reflect yet another aspect of the immune status that
could be involved in the mechanisms of response to IFN. Indeed, WBC
counts have been reported to be a marker of response to cytokines such
as interleukin 2; it has been shown that response to interleukin 2 in
adoptive immunotherapy of advanced cancer was associated with a
baseline lymphocyte count above 1.4 x
109/liter (16)
.
The search for a treatment by variable interaction (i.e.,
for a different response to treatment according to different levels of
a variable), which was carried out in this study, was not initially
planned and, therefore, belongs to exploratory data analysis
(17)
. The method used is one that allows an understanding
of the prognostic role of the clinical, histological, and biological
variables in the three groups considered (i.e., the IFN
group, the control group, and both combined). Exploratory analyses,
unlike hypotheses testing, are concerned with the manipulation and
summarization of data to make it more comprehensible to the human mind
and with discovering patterns that may be concealed in complex data
sets (18)
. Patterns should be sought and examined to see
whether they lead to interesting questions; one must bear in mind,
however, that the surest way to assess the truth of these patterns is
by finding them in other data sets (9)
. Our results should
be validated in similar groups of patients with low-dose IFN regimens,
ideally in a prospective setting. Moreover, they should be sought in
different melanoma populations and with different IFN regimens; indeed,
because the WBC-IFN interaction is presumably linked to the mechanism
of action of IFN, it might not be found in high-dose regimens where
different mechanisms of IFN are at work (i.e.,
immunomodulatory versus cytotoxic effects).
In this exploratory study, the only initial characteristic that
predicted a beneficial response to low-dose adjuvant IFN was a high
baseline WBC count. The fact that such a simple and rough marker could
be related to prognosis and response to IFN may surprise, especially
when subtle mechanisms are at work. However, one should remember that
simple markers such as sedimentation rate are very useful markers in
the management of many inflammatory diseases. If our results are
confirmed by other studies, WBC counts may help to identify patients
who would benefit from IFN and avoid unnecessary toxicity and quality
of life impairment in those who would not respond to IFN.
|
FOOTNOTES
|
|---|
The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.
1 The abbreviations used are: AJCC, American Joint
Committee on Cancer; HR, hazard ratio; NC, neutrophil count; DFI,
disease-free interval; CI, confidence interval. 
Received 7/12/00;
accepted 10/ 3/00.
|
REFERENCES
|
|---|
-
Kirkwood J., Strawderman M., Ernstoff M., Smith T., Borden E., Blum R. Interferon -2b adjuvant therapy of high-risk resected cutaneous melanoma: the Eastern Cooperative Oncology Group Trial EST 1684. J. Clin. Oncol., 14: 7-17, 1996.[Abstract]
-
Grob J., Dreno B., de La Salmonière P., Delaunay M., Cupissol D., Guillot B., Souteyrand P., Sassolas B., Cesarini J., Lionnet S., Lok C., Chastang C., Bonerandi J. Randomised trial of interferon IFN-2a as adjuvant therapy in resected primary melanoma thicker than 1. 5 mm without clinically detectable node metastases. Lancet, 351: 1905-1910, 1998.[CrossRef][Medline]
-
Pehamberger H., Soyer H., Steiner A., Kofler R., Binder M., Mischer P., Pachinger W., Auböck J., Fritsch P., Kerl H., Wolff K. Adjuvant interferon -2a treatment in resected primary stage II cutaneous melanoma. J. Clin. Oncol., 16: 1425-1429, 1998.[Abstract/Free Full Text]
-
Gutterman J. Cytokine therapeutics: lessons from interferon . Proc. Natl. Acad. Sci. USA, 91: 1198-1205, 1994.[Abstract/Free Full Text]
-
Kirkwood J. Adjuvant IFN2 therapy of melanoma (Editorial). Lancet, 351: 1901-1903, 1998.[CrossRef][Medline]
-
Koh H. Cutaneous melanoma. N. Engl. J. Med., 325: 171-182, 1991.[Medline]
-
Therneau, T. Extending the Cox Model. Rochester, MN: Mayo Clinic, 1996.
-
Gail M., Simon R. Testing for qualitative interactions between treatment effects and patient subsets. Biometrics, 41: 361-372, 1985.[CrossRef][Medline]
-
Byar D., Green S. The choice of treatment for cancer patients based on covariate information. Bull. Cancer (Paris), 67: 477-490, 1980.
-
Kaplan E., Meier P. Non parametric estimation from incomplete observations. J. Am. Stat. Assoc., 53: 457-481, 1958.[CrossRef]
-
Cox D. Regression models and life tables (with discussion). J. R. Stat. Soc. B, 34: 187-220, 1972.
-
Shuster J., van Eys J. Interaction between prognostic factors and treatment. Control. Clin. Trials, 4: 209-214, 1983.[Medline]
-
Anderson J., Cain K., Gelber R. Analysis of survival by tumor response. J. Clin. Oncol., 1: 710-719, 1983.[Abstract]
-
Garrison M., Nathanson L. Prognosis and staging in melanoma. Semin. Oncol., 23: 725-733, 1996.[Medline]
-
Haddad F., Stall A., Messina J., Brobeil A., Ramnath E., Glass L., Cruse C., Berman C., Reintgen D. The progression of melanoma nodal metastasis is dependent on tumor thickness of the primary lesion. Ann. Surg. Oncol., 6: 144-149, 1999.[Abstract]
-
West W., Tauer K., Yannelli J., Marshall G., Orr D., Thurman G., Oldham R. Constant-infusion recombinant interleukin-2 in adoptive immunotherapy of advanced cancer. N. Engl. J. Med., 316: 898-905, 1987.[Abstract]
-
Simon R., Altman D. Statistical aspects of prognostic factor studies in oncology. Br. J. Cancer, 69: 979-985, 1994.[Medline]
-
Andrews D. Exploratory data analysis Kruskal W. Tanur J. eds. . International Encyclopedia of Statistics, : 97-107, Free Press New York 1978.
Top
ABSTRACT
INTRODUCTION
