Back to WCC Published Papers


A Gynecologic Oncology Group Study

Nick M. Spirtos, M.D.1, Cynthia M. Westby, M.S., M.A.2, Hervy E. Averette, M.D.3, and John T. Soper, M.D.4.

  1. Director, Women's Cancer Center of Northern California, Palo Alto, CA 94304.
  2. Biostatistician, Gynecologic Oncology Group Statistical Office,
    Roswell Park Cancer Institute, Buffalo, NY
  3. American Cancer Society Professor of Clinical Oncology, Sylvester
  4. Professor and Director, Division of Gynecologic Oncology, Jackson
  5. Memorial Medical Center, University of Miami School of Medicine,
  6. Miami, FL 33136.
  7. Professor, Division of Gynecologic Oncology, Department of Obstetrics and Gynecology,
    Duke University Medical Center, Durham, NC 27710.

Corresponding Author:
Nicola M. Spirtos, M.D.
Women's Cancer Center at
Community Hospital Los Gatos
815 Pollard Road
Los Gatos, CA 95030
PHONE: 408-866-3843
FAX: 408-866-3939

Reprint address:
GOG Administrative Office
Suite 1945
1234 Market Street
Philadelphia, PA 19107

The following are participating institutions and the National Cancer Institute grants supporting this study.

University of Alabama at Birmingham (CA 12484), Oregon Health Sciences Center*, Duke University Medical Center (CA 12534), Temple University Health Science Center Hospital (CA 27816), University of Rochester Medical Center (CA 12482), Walter Reed Army Medical Center (CA 23501), University of Minnesota Medical School (CA 23088), University of Southern California Medical Center at Los Angeles (CA 37535), University of Mississippi Medical Center (CA 13633), Colorado Foundation for Medical Care (CA 15975), University of California Medical Center at Los Angeles (CA 13630), University of Miami School of Medicine (CA 37234), University of Utah Medical Center (CA 15469), The Milton S. Hershey School of Medicine of the Pennsylvania State University (CA 16386), Georgetown University Hospital (CA 16938), University of North Carolina School of Medicine (CA 23073), University of Texas Health Science Center at Dallas (CA 28160), Indiana University Medical Center (CA 21720), Bowman Gray School of Medicine of Wake Forest University (CA 21946), The Albany Medical College of Union University (CA 27469), University of California Medical Center at Irvine (CA 23765), Tufts New England Medical Center (CA 37569), Illinois Cancer Council (CA 27806), University of Pittsburgh School of Medicine*, St. Louis University Medical Center (CA 35571), Stanford University Medical Center (CA 35640), University of Connecticut Medical Center*, State University of New York Downstate Medical Center (CA 34477), University of Michigan Medical School*, University of Puerto Rico*, Jacksonville University Hospital* and Latter Day Saints Hospital*.

* Unfunded.



Recurrence and survival in 504 patients with Stage I squamous cell carcinoma of the cervix was not affected by perioperative blood transfusion.


Objective: To determine if perioperative blood transfusion adversely affected risk of recurrence in 504 evaluable patients with Stage I squamous cell carcinoma of the cervix accessioned prospectively in a Gynecologic Oncology Group study.

Methods: Seventy-seven percent of the patients received blood products within two weeks of surgery.

Results: The number of units transfused was found to be significantly related to recurrence-free interval (RFI) and survival using univariate analysis.

Conclusion: When adjusted for clinical tumor size, capillary-lymphatic space involvement and depth of tumor invasion using multivariate analysis, the number of units transfused was no longer statistically significant with respect to either RFI or survival.


Immunosuppression secondary to allogeneic blood transfusion is well documented in the medical literature.1-5 The mechanism responsible for this effect of blood transfusion is not entirely understood.4-6 and its clinical significance in cancer patients remains unclear.7-19 Only a few retrospective studies address this issue in patients with cervical cancer and arrive at opposite conclusions.20-23 Using the data base accumulated by the Gynecologic Oncology Group (GOG) Protocol #4924, a prospective surgical pathological study of Stage I squamous cell carcinoma of the cervix, we performed an analysis to determine if perioperative blood transfusion adversely affected the risk of recurrence.


From 1981 to 1984, 1125 patients with previously untreated, histologically confirmed Stage I cervical cancer with more than 3 mm of invasion were entered into this study. Written informed consent had been obtained from all patients prior to entry on study fulfilling all institutional, state and federal regulations. One hundred eighty-five patients were excluded because they did not meet original eligibility criteria, leaving 940 evaluable patients.24 An additional 295 were eliminated from the current analysis because of cell type (208 patients) or the presence of gross disease outside of the uterus and cervix or microscopic involvement of the aortic lymph nodes (87 patients). This left 645 evaluable patients from the initial study population who had squamous carcinoma of the cervix without evidence of metastatic disease. One hundred forty-one patients did not have transfusion information available leaving 504 evaluable patients for this analysis.

The original protocol directed that a radical hysterectomy, pelvic and para-aortic lymph node dissection be undertaken within uniform guidelines provided therein.24 The pathology material was previously reviewed and analyzed by two pathologists of the GOG Pathology Committee. Patients were followed closely, with strict guidelines for documentation of disease status. Clinical data collected and analyzed in addition to that detailed by Delgado at al24 included estimated blood loss, transfusion status, rate of recurrence and survival. All transfused patients received allogeneic blood products.

Either the Pearson χ2 test25 or Fisher's exact test26 assessed the association of categorical clinical/pathological factors (e.g., performance status, histologic grade) with respect to transfusion status. The Kruskal-Wallis test27 assessed the association of continuous clinical/pathological factors (e.g., age, maximum clinical tumor diameter, depth of invasion) with respect to transfusion status. Recurrence-free interval (RFI) was defined as the time from study entry to physical or radiological evidence of disease recurrence or date last seen. Survival was defined as the time from study entry to death or date last seen. Life tables were computed using the product-limit estimate by Kaplan and Meier.28 Cox's proportional hazards model29 was used to identify and simultaneously evaluate the independent prognostic factors associated with survival and RFI.


Of the 504 evaluable patients, 391 (77.6%) were transfused with allogeneic blood products while 113 (22.4%) were not. Patients were considered transfused if they received blood within two weeks of surgery. A significant linear association exists between the estimated blood loss and the amount of blood transfused (Figure 1). The distribution of clinical/pathological factors with respect to transfusion status is displayed in Table 1. The factors that were significantly associated with transfusions were depth of invasion (p = 0.004), gross primary tumor (p = 0.03) and age (p = 0.04). Figures 2 and 3 display univariate analyses of RFI and survival time by transfusion status, respectively. Neither RFI nor survival time differences were statistically significant (p = 0.07 and p = 0.2, respectively).

The effect of the quantity of blood transfused on RFI and survival was evaluated. In the univariate case, units-of-blood-transfused was found to be significant with respect to both RFI and survival (p = 0.02 and p = 0.03, respectively). Table 2 displays the risks associated with survival by the units-of-blood-transfused (Recurrence risks were similar to that of survival).

The Cox model was used to evaluate the independent prognostic value of blood transfusion, with the risk factors used for modeling RFI and survival being identified by Delgado et al.24 Those factors included clinical tumor size, capillary-lymphatic space involvement and depth of tumor invasion. After adjusting for these factors, units-of-blood-transfused was no longer found to be a significant factor with respect to RFI and survival (p = 0.20 and p = 0.16, respectively). Table 2 also displays the risks associated with survival in this adjusted analysis.

The patient population was then separated into high and low-risk groups. High-risk was based on the presence of deep-third invasion, middle-third invasion with 10 mm or greater invasion, clinical tumor size greater than 4 cm, or positive capillary-lymphatic space involvement.24 The original analysis was repeated for the patients grouped as high-risk (N = 340) and similar results were obtained. Units-of-blood-transfused was again found to be a significant factor with respect to both RFI and survival when conducting the univariate analysis but was no longer found to be significant in the adjusted analysis.

For the patients grouped as low risk (N = 164), analysis of RFI and survival could not be done because of the lack of statistical power which is a function of the number of events (e.g., recurrence). Only eight patients recurred and seven patients died (Table 3).


The relationship between perioperative blood transfusion and disease-free survival has been examined in two series20,21 and one brief report24 of patients with early carcinoma of the cervix. Eisenkop et al concluded that transfusion adversely affected the rate of recurrence20 while Dalrymple found no such relationship.21 Both groups attempted to limit the variable that might impact the effect of blood transfusion by studying patients with early lesions (IB or IIA). Eisenkop et al went so far as to eliminate patients with lymphatic metastases, vascular space involvement, or had undergone perioperative radiation therapy.20 Dalrymple analyzed 150 similar patients with early lesions not requiring radiation therapy.21 Despite these efforts, both series could be criticized as the data were collected retrospectively, the sample size was relatively small, and the statistical analysis not multivariate. In a brief communication, Soper et al22 used multivariate analysis to assess the impact of transfusion on survival and PFI in 320 patients undergoing radical hysterectomy for early cervical cancer. Transfusions had no adverse effect on outcome in either the entire patient population or a subset of 236 patients with disease confined to the cervix.22

Using data accumulated by the Gynecologic Oncology Group, some of the limitations of these studies could be overcome. It was thought that 504 evaluable patients with early squamous cell carcinoma would provide an adequate statistical sample to determine if perioperative blood transfusion adversely affected the risk of recurrence after radical hysterectomy. Specifically, this series of patients would allow for the detection of a relative risk of two between the transfused and non-transfused groups with the probability of type I error fixed at 0.05 and a power of 0.80 in the two-tailed test.

Using the risk factors identified by Delgado et al, this population was divided into high and low-risk groups.24 In the high-risk population (N = 340), units-of-blood was again found to be significantly related to RFI and survival. However, multivariate analysis, taking into account the effect of tumor size, capillary-lymphatic space involvement and depth of tumor invasion, revealed that units-of-blood- transfused was not a significant independent factor with respect to either RFI or survival.

The 164 patients identified as low risk were analyzed separately. As only eight patients in this group recurred and seven died, meaningful statistical analysis of RFI and survival was impossible, despite the fact that all of the patients who recurred had been transfused. Based on this, it is our opinion that it is unlikely that a large enough group of low-risk patients can be prospectively accrued that would allow for a meaningful analysis of the effect of transfusion on RFI and survival.

Therefore, it is doubtful that a study will ever be completed that will fully resolve this issue.

The prognostic effects of transfusions has been extensively studied in patients with colorectal malignancies. Over 30 relatively small retrospective studies have attempted to address this issue, with conflicting results.9-11 Two recent meta-analyses of this literature found a statistically significant poor prognostic effect of allogeneic blood transfusions, but the authors cautioned that transfusions might not be the actual cause for an increased risk of recurrence.9-10 Given the limitations of the data, the need for transfusions might reflect subtle differences in prognostic characteristics of the tumor in individual patients, or differences in surgical technique.9-10

A subsequent retrospective analysis of over 1000 patients with colorectal malignancies undergoing curative resections failed to substantiate an adverse effect of allogeneic transfusions upon disease recurrence.11 Furthermore, two studies that randomized patients undergoing resections of colorectal malignancy to receiving autologous versus allogeneic blood reached divergent conclusions. Heiss et al30 found no difference in outcome when analyzing their 120 patients by intent-to-treat, but did report that the use of any allogeneic blood products had a significant independent adverse effect on disease-free survival. In contrast, among 427 patients randomized by Busch et al,31 the use of allogeneic blood products did not significantly affect disease-free survival, while patients who received any blood products had an increased risk for recurrence. They concluded that circumstances affecting the need for transfusions most likely reflect undefined tumor prognostic factors or the skill of the surgeon, rather than an immunosuppressive effect of allogeneic transfusions.32 Clearly, every effort should be made to reduce the rate of transfusion, if only to reduce the risk of transmitting infectious diseases. In our series, over 75% of the patients were transfused. Other groups have published transfusion rates between 44 and 55 percent.20,24,32 Benjamin et al32 analyzed a subgroup of patients operated on between July 7, 1991 and June 30, 1993, a time period during which their awareness of transfusion-related infection was reflected in clinical practice.

Admirably, their transfusion rate fell from 91% (1980-1983) to 44% (1991-1993). In this later group, only 19% of the patients received heterologous blood while 81% received autologous units donated prior to surgery. The estimated blood loss and mean units transfused were also less in patients operated on during the period from 1991 to 1993.32

In conclusion, RFI and survival in patients with squamous cell cancer of the cervix could not be shown to be independently related to transfusion status in our study. The raw data and the risks of transmitting infectious disease calls for caution, and every effort to decrease transfusion rates in patients undergoing radical hysterectomy should be made.


1. Opelz G, Terasaki P. Dominant effect of transfusions on kidney graft survival. Transplantation 1980;29:153-8.

2. Opelz G, Graver B, Terasaki PI. Induction of high kidney graft survival rate by multiple transfusions. Lancet 1981;1:1223-5.

3. Opelz G, Lenhard V. Immunological factors influencing renal graft survival. Ann Rev Med 1983;34:133-44.

4. George CD, Morello PJ. Immunologic effects of blood transfusions upon renal transplant, tumor operations, and bacterial infections. Am J Surg 1986;152:329-36.

5. van Twuyver E, Mooijaart RJD, ten Berge IJ, van der Horst AR, Wilmink JM, Kast WM, Melief CJ, de Waal LP. Pretransplantation blood transfusion revisited. N Engl J Med 1991;325:1210-13.

6. Fischer E, Lenhard V, Seifert P, Kluge A, Johannsen R. Blood transfusion-induced suppression of cellular immunity in man. Hum Immunol 1980;1:187-94.

7. Look KY, Reisinger M, Stehman FB, Miser M, Ehrlich CE, Sutton GP. Blood transfusion and the risk of recurrence in squamous carcinoma of the vulva. Am J Obstet Gynecol 1993;168:1718-21.

8. Blumberg N, Heal JM. Transfusion and recipient immune function. Arch Pathol Lab Med 1989;113:246-53.

9. Chung M, Steinmetz OK, Gordon PH. Perioperative blood transfusion and outcome after resections for colorectal carcinoma. Br J Surg 1993;80:427-32.

10. Vainvakas E, Moore SB. Perioperative blood transfusion on colorectal recurrence: A qualitative statistical overview and meta-analysis. Transfusion 1993;33:754-65.

11. Donohue JH, Williams S, Cha S, Windschitl HE, Witzig TE, Nelson H, Fitzgibbons RJ Jr, Wieand HS, Moertel CG. Perioperative blood transfusions do not affect disease recurrent of patients undergoing curative resection of colorectal carcinoma: A Mayo/North Central Treatment Group study. J Clin Oncol 1995;13:1671-8.

12. Foster RS Jr, Foster JC, Costanza MC. Blood transfusions and survival after surgery for breast cancer. Arch Surg 1984;119:1138-40.

13. Voogt PJ, Cornelis JH, Brand A. Perioperative blood transfusion and cancer prognosis. Cancer 1987;59:836-43.

14. Tartter P, Papastestas AG, Lesnick G, Burrows L, Aufses A Jr. Breast cancer recurrence is associated with perioperative blood transfusion. Proc Am Soc Clin Oncol 1983;2:51 (abstr).

15. Tartter PI, Burrows L, Papatestas AE, Lesnick G, Aufses AH Jr. Perioperative blood transfusion has prognostic significance for breast cancer. Surgery 1985;97:225-32.

16. Tartter PI, Burrows L, Kirschner P. Perioperative blood transfusion adversely affects prognosis after resection of stage 1 (NO) non-oat cell cancer. J Thorac Cardiovasc Surg 1984;88:659-84.

17. Nathanson SD, Fox BA, Westrick PW, Haas GP. Effects of allogenic blood transfusion in C57BL/6 mice with and without melanomas, abstracted. Proc Am Assoc Cancer Res 1984;2:270 (abstr).

18. Hyman NH, Foster RS, DeMueles JE, Costanza MC. Blood transfusion and survival after lung cancer resection. Am J Surg 1985;149:502.

19. Rosenberg SA, Seipp CA, White DE, Wesley R. Perioperative blood transfusions are associated with increased rates of recurrence and decreased survival in patients with high-grade soft-tissue sarcomas of the extremities. J Clin Oncol 1985;3:698-709.

20. Eisenkop SM, Spirtos NM, Montag TW, Moossazedeh J, Warren P, Hendrickson M. The clinical significance of blood transfusion at the time of radical hysterectomy. Obstet Gynecol 1990;76:110-13.

21. Dalymple JC, Monaghan JM. Blood transfusion and disease free survival in cervical carcinoma. J Obstet Gynecol 1988;8:356-9.

22. Soper JT, Berchuck A, Clarke-Pearson DL. The clinical significance of blood transfusion at the time of radical hysterectomy (letter) Obstet Gynecol 1991;77:165-6.

23. Morris PC, Haugen J, Tomjack J, Anderson B, Buller RE. Blood transfusion and the risk of recurrence in stage IB cervical cancer. Gynecol Oncol 1995;57:401-6.

24. Delgado G, Bundy BN, Zaino R, Sevin B, Creasman WT, Major F. A prospective surgical pathological study of disease-free interval in patients with stage IB squamous cell carcinoma of the cervix (A Gynecologic Oncology Group study) Gynecol Oncol 1990;35:352-7.

25. Snedecor GW, Cockran WG. Statistical Methods. Ames, Iowa: Iowa State University Press, 1967;211-13.

26. Agresti A. Categorical Data Analysis. New York: John Wiley & Sons, 1990.

27. Kruskal WH, Wallis WAL. Use of ranks in one-criterion variance analysis. J Am Stat Assoc 1952;47:583-621.

28. Kaplan EL, Meier P. Non-parametric estimation from incomplete observations. J Am Stat Assoc 1958;53:457-81.

29. Cox DR. Regression models and life table (with Discussion) J R Stat Soc 1972;B34:187-202.

30. Heiss M, Mempel W, Delanoff C, Jauch K, Gabka C, Mempel M, Dieterich HJ, Eissnr HJ, Schildberg FW. Blood transfusion-modulated tumor recurrence: first results of a randomized study of autologous versus allogenic blood transfusion in colorectal cancer surgery. J Clin Oncol 1994;12:1859-67.

31. Busch ORC, Hoop WCJ, van Papendrecht MA, Marquet RI, Jeekel J. Blood transfusions and prognosis of patients with colorectal cancer. N Eng J Med 1993;13:1372-6.

32. Benjamin I, Barakat RR, Curtin JP, Jones WB, Lewis JL, Hoskins WJ, Blood transfusion for radical hysterectomy before and after the discovery of transfusion-related human immunodeficiency virus infection. Obstet Gynecol 1994;84:974-8.


Distribution of Clinical/Pathological Factors by Transfusion Status
                               Transfusion Status
                                     No             Yes
Variable                        N  (%)         N  (%)      P-Value

Age                                                           0.04
    	  30                      29  (26)       56  (14)
   31 - 40                      36  (32)      140  (36)
   41 - 50                      18  (16)      100  (26)
   51 - 60                      20  (18)       60  (15)
   61+                          10  ( 9)       35  ( 9)

GOG Performance Status                                        0.1
   0                           103  (91)      334  (85)
   1 - 2                        10  ( 9)       57  (15)

Gross Primary Tumor                                           0.03
   Gross                        75  (66)      299  (76)
   Occult                       38  (34)       92  (24)

Maximum Clinical Tumor Diameter                               0.8*
   Occult                       38  (34)       92  (24)
   0.1 - 1.0 cm                  9  ( 8)       38  (10)
   1.1 - 2.0 cm                 22  (19)       82  (21)
   2.1 - 3.0 cm                 20  (18)       76  (19)
   3.1+ cm                      24  (21)      103  (26)

Gross Primary Tumor Description                               0.2*
   Occult                       38  (34)       92  (24)
   Endophytic                   27  (24)       83  (21)
   Exophytic                    31  (27)      156  (40)
   Ulcerated                    17  (15)       59  (15)

Histologic Grade                                             0.1
   1                            22  (19)       57  (15)
   2                            55  (49)      230  (59)
   3                            36  (32)      104  (27)

Depth of invasion                                             0.004
    	   5 mm                   39  (35)      101  (26)
    6 - 10 mm                   45  (40)      140  (37)
   11 - 15 mm                   17  (15)       88  (23)
   16 - 20 mm                    7  ( 6)       33  ( 9)
   21+ mm                        4  ( 4)       21  ( 5)

   Inner third                  42  (37)      114  (30)      0.3
   Middle third                 35  (31)      126  (33)
   Outer third                  36  (32)      146  (38)

Uterine extension                                            0.5
   Negative                    101  (89)      337  (87)
   Positive                     12  (11)       51  (13)

Surgical margins                                             1.0
   Negative                    109  (97)      379  (97)
   Positive                      3  ( 3)       12  ( 3)

Parametrial extension                                        0.7
   Negative                    105  (94)      362  (93)
   Positive                      7  ( 6)       29  ( 7)

Capillary/lymphatic spaces                                   0.3
   Negative                     68  (60)      213  (55)
   Positive                     45  (40)      177  (45)

Cytology                                                     0.2
   Negative                    110  (99)      383 (100)
   Positive                      1  ( 1)        0  ( 0)


Relative Risk (95% Confidence Intervals)* of
Survival by Units-of-Blood-Transfused**
Transfused         Univariate          Adjusted

    0            1.0 (0.90, 1.10)    1.0 (0.90, 1.11)
    2            1.2 (1.15, 1.34)    1.2 (1.06, 1.27)
    4            1.6 (1.46, 1.65)    1.4 (1.25, 1.46)
    6            1.9 (1.84, 2.04)    1.6 (1.47, 1.68)
    8            2.4 (2.32, 2.52)    1.8 (1.72, 1.93)
   10            3.0 (2.92, 3.11)    2.1 (2.02, 2.23)
*  Linear relationship assumed between log of relative 
   risk and units of blood transfused.
** Based on 502 cases.


Status of Low Risk Patients by Transfusion Status
                         Transfusion Status
Status                   No              Yes   
Dead                      0                7        
Alive                    38              119
Recurrence                0                8 
No Recurrence            38              118