10970142, 1983, 1, Downloaded from https://acsjournals.onlinelibrary.wiley.com/doi/10.1002/1097-0142(19830701)52:1<126::AID-CNCR2820520123>3.0.CO;2-Y by Cochrane Mexico, Wiley Online Library on [24/10/2022].

See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License
Peripheral Lymphocyte and Eosinophil Counts as
Indicators of Prognosis in Primary Breast Cancer

HELEN E. OWNBY, PHD, LARRY D. ROI, PHD, RUTH R. ISENBERG, BA, MICHAEL J. BRENNAN, MD, AND
THE BREAST CANCER PROGNOSTIC STUDY ASSOCIATES

As a part of a major study on the pathophysiologic indices for recurrence of human breast cancer,
preoperative eosinophil and lymphocyte counts were determined on 419 and 581 primary breast cancer
patients, respectively. Patients with lymphocyte counts less than or equal to 1500/mm3 and/or eosin-
ophil counts of less than 55/mm3 had significantly higher risk of recurrent disease than those patients
who had normal or high levels of eosinophils and/or lymphocytes. These findings may indicate that
the immunologic activities of eosinophils and lymphocytes enhance the patients’ ability to respond
against disease.
Cancer 52:126-130, 1983.

A N IN-DEPTH ANALYSIS O f factors associated with

early recurrence of metastatic disease was under-
taken in 1975 by the Michigan Cancer Foundation’s
with exception of metastasis to the lymph nodes; (3) the
patient had no previous history of breast cancer within
five years of this surgery; and (4) lymph nodes had been
Breast Cancer Prognostic Study.’ One important aspect removed and examined for the presence of metastatic.
of this study is the examination of the prognostic po- foci. The 592 patients studied with at least one follow-
tential of the clinical data routinely collected by physi- up and with either a presurgical eosinophil or lympho-
cians in their care of the breast cancer patient. cyte counts or both form the subset of study patients for
The role of the immune system in the prognosis of whom results, are reported here. These patients were
breast cancer has been examined at great length.2 Part followed from two to 75 months prior to recurrent dis-
of the current understanding of the host-tumor rela- ease onset, with an average time in the study of 48
tionship is that the ultimate fate of the patient is deter- months. The patients were contacted by nurse-coordi-
mined by her immunological ~ o m p e t e n c e . ~
An, ~ ideal nators at three-month intervals as standard protocol for
measurement of this would predict survival. Unfortu- the Study. The 28 patients with a previous cancer more
nately, to date, no series of practical immunologic tests than five years earlier (including 15 with contralateral
have been identified which clearly permit such predic- breast cancer) who are included in this study, did not
tion. differ in distribution of lymphocyte or eosinophil values
This report examines the correlation between pre- from other patients.
surgical lymphocyte and eosinophil counts and early Presurgical leukocyte and differential counts were
recurrent disease. performed by the clinical hematology laboratories of the
various participating hospitals on samples obtained at
Materials and Methods the time of hospital entry. Either manual or mechanical
methods were used. Protocol in all hospitals required a
The criteria for entry into the Breast Cancer Prog-
minimum of 100 leukocytes classified in each case, ex-
nostic Study included the following: (1) the patient must
cept in cases of an abnormal count where 200 were clas-
be female; (2) the diagnosis was primary breast cancer
sified. The number of cells per cubic millimeters was
with no known metastasis present at the time of surgery
calculated from the total leukocyte count and the per-
cent cell type.
From the Michigan Cancer Foundation, Detroit, Michigan 4820 I .
Supported by the National Institute of Health Research Grant CA
Life table analysis’ was used to determine whether
16175. individual categorical factors were significantly related
Address for reprints: Helen E. Ownby, PhD, Michigan Cancer Foun- to time to recurrence. Breslow’s modification of Gehan’s
dation, I10 East Warren Avenue, Detroit, MI 48201.
The authors thank R. McCune, RN. for patient follow-up and L.
generalized Wilcoxin test was used to compute the sig-
Howard, MPH, and J. Chu, MS, for their statistical asssistance. nificance level for the life table analysis6 The traditional
Accepted for publication April 9, 1982. chi-square test was employed for testing the association

O008-543X/83/0701/0126$ I .05 0 American Cancer Society

126
 10970142, 1983, 1, Downloaded from https://acsjournals.onlinelibrary.wiley.com/doi/10.1002/1097-0142(19830701)52:1<126::AID-CNCR2820520123>3.0.CO;2-Y by Cochrane Mexico, Wiley Online Library on [24/10/2022]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License
No. I EOSINOPHIL
A N D LYMPHOCYTE
LEVELSIN BREASTCA Ownby et al. 127

of variables in contingency tables. Cox's proportional

hazards regression method' was used for all multivari-
able analyses relating a set of factors to time to recur-
rence. Significance levels for the addition of a factor were
computed using the maximum likelihood ratio chi-
square. Other statistical methods were utilized as needed
from the Michigan Interactive Data Analysis System
(MIDAS).' An age-adjustment of lymphocyte counts
utilizing the normal values of MacKinney' was em- (n = 383)
c55 ce11s/rnm3
ployed to determine if the patients' age influenced the (n = 36
cell counts.

Results
'"1
0
p = 0178

14 28 42 56 70
2

A comparison of patients groups with counts read Time (Months)

mechanically (38%)versus those read by manual (62%)
methods showed no significant difference in the capacity FIG. I . Time to recurrence for patients with peripheral blood eo-
sinophil counts less than 55 cells/mm3 (- - -) and equal to or greater
of the counts to determine prognosis (P> 0.5) for either than 55 cells/mm3 (-).
eosinophils or lymphocytes.
Low lymphocyte and eosinophil counts are highly
year estimated recurrence rate of 30 f 5%. Those pa-
associated measurements. As seen in Table 1, patients
tients with greater than 1500 cells/mm' had an esti-
with low eosinophil counts tend to have low lymphocyte
mated recurrence rate at two years of 19.6 f 3%. The
counts. The Spearman rank order correlation between
all lymphocyte and eosinophil counts is .20, which is estimated percent of disease-free patients as a function
of time is plotted in Figure 2. The time to recurrence
significantly different from 0 (P< 0.01).
Nine percent of the patients seen at least once at fol- is significantly different between the two groups (P
low-up for whom eosinophil counts were available (n -= 0.02). Further examination of the lymphocyte counts
= 419) had eosinophil counts less than 55 cells/mm3
showed that dividing the unadjusted patient counts into
prior to their surgery. The estimated 24-month recur- high and low groups at any point between 1300 and
rence rate for this group is 34 f 8%, whereas patients 1900 cells/mm3 also showed a significant difference be-
with 55 or more eosinophils/mm3have an estimated 24- tween groups with respect to time to recurrence. The
month recurrence rate of 21 f 2% (life table analysis). level of 1500 cells/mm3, two standard deviations below
The life table estimate for the percent of patients disease- the mean" was chosen as the cut-off point. After age
free as a function of time is plotted for both groups in adjustment of the lymphocyte counts using the normal
Figure 1. As standard convention, groups which de- values of MacKinney,' the optimal cut-off point is 2000
creased to less than ten patients were not plotted on the
graphs. The time to recurrence for the high esinophil
group was significantly longer than that for the low eo-
sinophil group (P < 0.02). The cut-off point of 55 cells/
mm3 was chosen because it gave the greatest significant
statistical difference between the high and low groups.
Five hundred and eighty-one patients from whom
lymphocyte counts are available have been seen at least
once at follow-up. Twenty-two percent had lymphocyte
counts less than or equal to 1500 cells/mm3 and a two-
-
n
m
C
L--,
L----
L 1500 cells/rnrn3
(n = 452)

$ 60- c 1500 ce~mrn3

a (n = 129)

TABLE1 . Relationship Between Eosinophil and Lymphocyte 5 0 1 p = 0200

Counts in Breast Cancer Patients
_________
-f
~

Eosinophil No. 96 Lymphocytes

0 14 28 42 56 70
counts patients 5 1500 c e ~ ~ s / m m 3
Time (Months)
46
19 FIG.2. Time to recurrence for patients with unadjusted peripheral
blood lymphocyte counts less than 1500 cells/mm' (- - -) and equal
408 (Y*) P < 0.0003 to or greater than I500 cells/mm3 (-).
 10970142, 1983, 1, Downloaded from https://acsjournals.onlinelibrary.wiley.com/doi/10.1002/1097-0142(19830701)52:1<126::AID-CNCR2820520123>3.0.CO;2-Y by Cochrane Mexico, Wiley Online Library on [24/10/2022]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License
128 CANCERJuly 1 1983 Vol. 5 2

three, and four years is 3 1.2 k 5.1, 3 1.2 f 5.1, and 35.3
k 6.3, respectively. In contrast, 302 patients having lym-
phocyte counts greater than 1500 cells/mm3 and eosin-
ophil counts greater than 54 cells/mm3 had estimated
respective two, three, and four-year recurrence rates of
19.1 k 2.6, 25.8 f 3.1, and 29.3 f 3.6% for the three
time periods. As shown in Figure 4,the curves for the
two groups are significantly different (P< .01).After age
adjustment of the lymphocyte counts, 143 patients hav-
ing either low lymphocyte (<2000 cells/mm3) and low
eosinophil ( 4 5 cells/mm3) counts or a combination of
low eosinophil, high lymphocyte or high eosinophil, low

01
4 lymphocyte counts, the estimated recurrence rates for
this group at two, three, and four years are 29.6 f 4.3,
0 14 28 42 56 70
32.3 -t 4.5, and 37.1 -t 5.3, respectively. In contrast, 265
Time (Months)
patients with high lymphocyte and eosinophil counts
FIG.3. Time to recurrence for patients with age-adjusted peripheral had estimated respective two, three, and four-year re-
lymphocyte counts less than 2 0 0 0 cells/mm' (- - -) and equal to or currence rates of 18.1 f 2.8, 24.5 f 3.3, and 26.9
greater than 2000 cells/mm' (-).
f 3.6% for the three time periods. As shown in Figure
5 , the curves for the two groups are significantly different
cells/mm3. The two-year estimated recurrence rate is (P< 0.002).
28.5 f 3.9%, for patients with less than 2000 cells/mm3 The above analysis indicates that these associations
and 18.9 t 2.3% for those patients with greater than or between cell counts are highly significant predictors of
equal to 2000 cells/mm3. In Figure 3 the percent of recurrence. Further analysis of our data was camed out
patients disease-free as a function of time is plotted for to see whether lymphocyte and eosinophil counts are
both groups. The time to recurrence is significantly dif- independent predictors of recurrence or, alternatively,
ferent between the two groups (P< 0.005). whether their significance may be explained by associ-
One hundred and six of the patients examined above ation with other predictors. The percentage of counts
have either low eosinophil and low lymphocyte counts, in the low group for both eosinophils and lymphocytes
or a combination of low eosinophil, high lymphocyte is associated with the clinical stage (TNM)" of the dis-
or high eosinophil, low lymphocyte counts. The esti- ease (Tables 2 and 3), in that a higher percentage of
mated percent recurrence rates for this group at two, patients in Stage 4 have low eosinophil counts and/or

100 100

90 90

(u a,
2
Lf 60
:
Lt
80
e,
m v,
m
0
c High ;
-
n
e,
C
70
L - --- -- -- -
(n = 302) -
m
C
70

0 Low
2 60 (n = 106) $
a
6C

5c "= 002
50 p = 0252

0 0
0 14 28 42 56 70 0 14 28 42 56 70

Time (Months) Time (Months)

FIG.4. Time to recurrence for patients with unadjusted lymphocyte FIG.5. Time to recurrence for patients with age-adjusted lympho-
and eosinophil counts, low for both or low for either lymphocyte or cyte and eosinophil counts, low for both or low for either lymphocyte
eosinophil counts (- - -)and high for both lymphocyte and eosinophil or eosinophil counts (- - -) and high for both lymphocyte and eosin-
counts (-). ophil counts (-).
 10970142, 1983, 1, Downloaded from https://acsjournals.onlinelibrary.wiley.com/doi/10.1002/1097-0142(19830701)52:1<126::AID-CNCR2820520123>3.0.CO;2-Y by Cochrane Mexico, Wiley Online Library on [24/10/2022]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License
No. 1 EOSINOPHIL
AND LYMPHOCYTE
LEVELSIN BREASTCA * Ownby ef a/. 129

low lymphocyte counts. Examination of the number of TABLE3. Relationship of Lymphocyte Count
lymphocytes and eosinophils however by stage did not to Clinical Stage (TNM)
show any groups with significantly different counts (< 1650/mm’)
(Kruskal-Wallis, Student’s t, or median test) for all pa- Stage No. patients Ilow lymph
tients or for those in age groups 20-39, 40-59, 60-79, 1 213 23.9
or older than 80 years (Table 4 and 5). Neither lym- 2 252 18.3
phocyte nor eosinophil counts were related to the age 3 39 20.5
of the patient by Spearman’s rho rank correlation coef- 4 -71 33.8
ficient (P> 0.05). Also, the cell counts are not associated 575*
individually by other prognostic indicators such as nodal ( x z ) P < 0.05 for differences among all four stages
status (P< 0.2), maximum tumor diameter (P< O.l), P < 0.02 for stages 1-111 versus IV
Bloom and Richardson tumor grade (P < 0.3), or es-
trogen receptor content of the tumor (P< 0.3). * Six patients with lymphocyte counts were not staged.
After taking into account nodal status, maximum tu-
mor diameter, tumor grade and estrogen receptor, nei- TABLE4. Pretreatment Peripheral Lymphocyte
Counts by Stage and Age
ther the eosinophil count (P< 0.2) nor the lymphocyte
count (P> 0.3) were significantly associated with time Age (YB)
to recurrence by likelihood ratio chi-square in the Cox
Stage 20-39 40-59 60-79 80+ All ages
regression analysis.
I 27/2214* 10412282 9212 176 I 112085
Discussion (1041) (966) (772) (873)
I1 1612231 14512197 1 1312I80 1312009
To our knowledge, this is the first time that a low (740) (847) (791) (870)
Ill 0 312166 251201 5 1312207 312654
eosinophil count has been implicated as an indicator of ( 1404) (896) (684) (969)
poor prognosis in breast cancer. The role of the eosin- IV 312217 3012095 4411859 213045
ophil has remained one of mystery since its discovery (778) (939) (697) (822)
by Ehrlich over a century ago.” The current hypothesis * No. patientstmean count (SD).
for the role of these cells include two very diverse modes
of operation; one which considers the role of eosinophils parallel the magnitude of eosinophilic response in tissues
as protective against parasitic and antimicrobial infec- to a variety of material^.'^"^ Therefore, it is impossible
tions and another which considers the eosinophil as a to predict the patient’s eosinophilic response to her tu-
modulator of anaphylaxis. ‘**I3 mor by her peripheral blood cell counts. Over 50 cases
Jong and KlebanoP4 have shown that intact stimu- of hypereosinophilia among patients with various can-
lated guinea pig eosinophils are cytotoxic to mouse as- cers including two with breast cancer have been re-
cites lymphoma cells (LSTRA) by the mechanism of the In most cases, metastases were already pres-
eosinophil peroxidase-peroxide-halide system. The as- ent, and the high eosinophil counts were considered to
sociation of low peripheral eosinophil count with early be indicators of poor prognosis.
development and metastatic breast cancers may have The lymphocyte count has been implicated as an in-
such a basis. However, other studies have established dicator of prognosis for breast cancer patients by nu-
that the circulating eosinophil level does not necessarily merous investigators. Riesco” found a positive corre-

TABLE2. Relationship of Eosinophil Count TABLE5. Pretreatment Peripheral Eosinophil

to Clinical State (TNM) Counts by Stage and Age

(<55/mm3) Age
Stage No. patients Ilow EOS
Stage 20-39 40-59 60-79 80+ All ages
1 152 7.2
2 183 7.7 I 21/155* 791160 581166 1641163
3 30 3.3 ( I 40) (134) (118) (128)
II 81129 1041231 881162 2 101196
4 -50 20
( 1 16) (69 1) ( 1 33) (495)
415. III 41143 211126 51157 331138
( x 2 ) P < 0.02 for differences among all four stages (120) (70) (73) (75)
IV 312 17 2011 10 311164 561149
( x 2 ) P < 0.003 for Stages 1-111 versus 1V ( 1 19) (90) (1 70) (144)
* Four patients with eosinophil counts were not staged. * No. patientstmean count (SD).
 10970142, 1983, 1, Downloaded from https://acsjournals.onlinelibrary.wiley.com/doi/10.1002/1097-0142(19830701)52:1<126::AID-CNCR2820520123>3.0.CO;2-Y by Cochrane Mexico, Wiley Online Library on [24/10/2022]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License
130 CANCERJuly 1 1983 Vol. 52

lation between breast cancer “curability” and the total of eosinophils and lymphocytes in breast cancer patients
number of peripheral lymphocytes. In a study of 305 who subsequently develop metastatic disease.
breast cancer patients Papatestas and Kark4 showed that
patients with a five year disease-free period postmastec- REFERENCES
tomy had significantly higher pretreatment counts than 1. Rich MA, Brennan MJ, the Scientific, Pathology and Surgical
those who developed second primaries and/or metas- Associates of the MCF Breast Cancer Prognostic Study. The Breast
tases. They also showed that differences in lymphocyte Cancer Prognostic Study Program: A study of the metastatic process.
counts between the means of the four stages were sig- In: Brennan MJ, McGrath CC, Rich MA, eds. New concepts in Etiol-
ogy and Control. New York Academic Press, 1980; 29-51.
nificant among patients aged 20-39 years as well as the 2. Heppner GH. Immunology: Breast cancer. Recent Resulfs Can-
paired means of similar age groups in Stages I and 11. cer Res. 1976; 57:95.
In a more recent study, Papatestas and associates2’ 3. Meyer KK. Survival after mastectomy for breast cancer: Predic-
tive value of lymphocyte count and its change after mastectomy in
showed that patients with positive axillary nodes and/ patients with metastases to one to three axillary nodes. World J Surg
or locally invasive tumor involving skin or pectoral fas- 1978; 2:331-339.
4. Papatestas AE, Kark AE. Peripheral lymphocyte counts in breast
cia and pretreatment lymphocyte counts >2000/mm3 carcinoma: An index of immune competence. Cancer 1974; 34:2014-
had a significantly longer disease-free survival than those 2017.
patients with <2000/mm3. A relationship of lymphocyte 5 . Gross AJ, Clark VA. Survival Distributions: Reliability appli-
counts to clinical stage was also observed by Bainbridge cations in the biomedical sciences. New York: John Wiley and Sons,
1975; 31.
ef ~ 1 . More
~ ’ recently Fodo32examined presurgical lym- 6. Breslow N. A generalized Kruskal-Wallis test for comparing K
phocyte counts in 66 long-term survivors (<20 years) samples subject to unequal patterns of censorship. Biometrika 1970;
related to 78 who died within five years of initial diag- 57379-594.,
7. Cox DR. Regression models and life tables. J R Stat Soc (Serjes
nosis and mastectomy. Absolute lymphocyte counts for B) 1972; 34: 187-220.
short survivors were significantly lower at diagnosis than 8. Fox DJ, Guire KE. Documentation for MIDAS ed. 3. The Uni-
the corresponding counts for the long term survivors. versity of Michigan, Statistical Research Laboratory, 1976.
9. MacKinney AA. Effect of ageing on the peripheral blood lym-
Our results support their findings with respect to prog- phocyte count. J Gerontol 1978; 33:2 13-2 16.
nosis, but are contrary to Papatestas and Kark’s findings 10. Beeson PB, McDermott W, Wyngaarden JB, eds. Cecil Text-
with respect to age and stage. We have noted that un- book of Medicine. Philadelphia: W. B. Sanders, 1979.
11. American J. Committee for Cancer Staging and End Results
adjusted low levels of lymphocytes show a significant Reporting. Manual for Staging Cancer, 1978.
relationship to Stage IV when Stages 1-111 are combined 12. Samter M. Eosinophils: Nominated but not elected. N Engl J
(P < 0.02). This finding also holds for low eosinophil Med 1980; 303:1175-1176.
13. Mahmound AAF, Austen KR, eds. The Eosinophil in Health
counts where Stage IV has a significantly higher pro- and Disease. New York: Grune and Stratton, 1980.
portion of patients in this group (P< 0.003). 14. Jong EC, Klebanoff SJ. Eosinophil-mediated mammalian tu-
The changes in lymphocyte levels in breast cancer mor cell cytotoxicity: Role of the perioxidase system. J Immunol1980
124: 1949-1953.
have frequently been attributed to the method of treat- 15. Riddle JM, Barnhart MI. The eosinophil as a source of profi-
ment in radiotherapy or ~hemotherapy.~ Our results in- brinolysin in acute inflammation. Blood 1965; 25:775-794.
volve patients prior to primary therapy and thus do not 16. Rebuck JW, Kelly AP, Sweet LC. Monitoring leukoytic Retic-
uloendothelial system functions in man. In: Rebuck JW, Berard CW,
contribute pro or con to this idea. Abell MR, eds. The Reticuloendothelial System. Baltimore: Williams
Our overall experience agrees with Wanebo et aLz3 & Wilkins Co., 1975; 81-133.
in that the mean lymphocyte counts for all patients in 17. Fendler JP, Tutin M, Pasquier P, Roman M, Bour H.Les hy-
pereosinophiles au cows des cancers. La Presse Medicale 1965;
our study are within the normal range, 1 500-3000/mm3, 73:2239-2242.
for adult females.22The utilization of age adjustment 18. Bostrom SG, Hart WR. Carcinomas of the cervix with intense
somewhat increases the significance of the number of stromal eosinophilia. Cancer I98 1; 47:2887-2893.
19. Riesco A. Five year cancer cure: Relation to total amount of
lymphocytes as an indicator of recurrent disease. Nu- peripheral lymphocytes and neutrophils. Cancer 1970; 25: 135-140.
merous studies of normal lymphocyte levels all indicate 20. Papatestas AE, Lesnick GJ, Genkins G, Aufses AH Jr. The
a significant decrease in total lymphocyte count after prognostic significance of peripheral lymphocyte counts in patients
with breast carcinoma. Cancer 1976; 37:164-168.
age 70 year^.^,^^." Therefore, the comparison of unad- 21. Bainbridge ET, Ford CHJ, Newman CE. Total lymphocyte
justed and adjusted lymphocyte counts is useful. counts in breast cancer. Lancet: 1978; 1:1203-1204.
Zacharski and Linman26 have suggested that lym- 22. Fodor J. Peripheral blood lymphocyte counts and survival in
breast cancer. Neoplasma 1976; 23:3 1 1-3 13.
phocytopenia is frequently seen in patients with various 23. Wanebo HJ, Thaler HT, Hansen JA, ef al. Immunologic reac-
malignancies and may be a result of either a lympho- tivity in patients with primary operable breast cancer. Cancer 1978;
cytopenia-inducing effect exerted by highly malignant 4 194-94.
24. Rosenthal M, Steinmann A. Lebensalter und Inmunitat. Dtsch
tumors or that cellular immune mechanisms mediated Med Wochenschr 1978; 103:409-412.
by lymphocytes are important for cancer homeostastis 25. Jamil NAK, Millard RE. Studies of T, B, and ‘null’ blood lym-
phocytes in normal persons of different age groups. Gerontology 198 I :
and a failure of this mechanism is reflected in the lym- 27:79-84.
phocytopenia.26Further study will be necessary to elicit 26. Zacharski LR, Linman JW. Lymphocytopenia: Its causes and
the biological mechanisms which cause the low levels significance. Mayo Clin Proc I97 1; 46: 168- 173.