doi:10.1016/S0360-3016(03)00821-6

CLINICAL INVESTIGATION Prostate

THE ROLE OF HEMOGLOBIN CONCENTRATION IN CLINICALLYLOCALIZED PROSTATE CANCER TREATED WITH RADICAL

RADIOTHERAPY � NEOADJUVANT ANDROGEN DEPRIVATION

CHRIS PARKER, F.R.C.R.,* PADRAIG WARDE, M.D.,† ANDY NORMAN, PH.D.,‡

TONY PANZARELLA , PH.D.,† CHARLES CATTON, M.D.,† ALAN HORWICH, F.R.C.R.,*ROBERT HUDDART, F.R.C.R.,*AND DAVID DEARNALEY, F.R.C.R.*

*Academic Unit of Radiotherapy and Oncology and‡Department of Computing and Information, Institute of Cancer Research andThe Royal Marsden NHS Trust, Sutton, Surrey, United Kingdom;†Department of Radiation Oncology, Princess Margaret Hospital,

Toronto, Ontario, Canada

Purpose: Serum hemoglobin level (Hb) is a significant determinant of treatment outcome after radical radio-therapy (RT) for several cancer types, but its importance in prostate cancer is not well established.Methods and Materials: Two treatment-specific cohorts of men with localized prostate cancer (T1–4, Nx/N0, M0)were analyzed. Seven hundred six men who received radical RT at Princess Margaret Hospital between 1987 and2000 comprise the RT-alone cohort, of whom 536 had a pre-RT Hb. Six hundred fifty-eight men received 3–6months’ neoadjuvant androgen deprivation (NAD) and radical RT at Royal Marsden Hospital between 1989 and2000 and comprise the NAD � RT cohort, of whom 475 had a pre-NAD Hb and 513 a pre-RT Hb. Time tobiochemical failure (TTBF) was the primary end point. Univariate and multivariate analyses using the Coxproportional hazards regression model were used for each data set independently to study the prognostic role ofpre-RT Hb, pre-NAD Hb, nadir Hb (lowest Hb during RT), Hb decrement (pre-NAD Hb � pre-RT Hb), Gleasonscore, presenting PSA, and T stage.Results: On univariate analysis, no significant association was seen between TTBF and any of the Hb variablesfor either data set. On multivariate analysis, TTBF was associated with presenting PSA (p < 0.001), Gleasonscore (p < 0.01), and (for the NAD � RT data set) T stage (p < 0.001), but not pre-NAD Hb (p � 0.24) or pre-RTHb (p > 0.3).Conclusion: Hemoglobin level is not an important determinant of RT outcome in localized prostate cancer.© 2004 Elsevier Inc.

Prostate cancer, Radiotherapy, Hemoglobin.

INTRODUCTION

Serum hemoglobin (Hb) level is a significant determinant oftreatment outcome, independent of other clinical prognosticfactors, for patients receiving radical radiotherapy (RT) forcervical carcinoma (1), head-and-neck cancer (2), transi-tional cell carcinoma of the bladder (3), and non–small-celllung cancer (4). In an overview of 51 studies addressing theprognostic role of initial Hb level, involving over 17,000patients receiving radical RT for a variety of malignancies,39 of the studies demonstrated a statistically significantimpact of Hb level on treatment outcome (5).

Only two previous groups have addressed the prognostic

significance of Hb level in clinically localized prostatecancer (6, 7) and have yielded contrasting results. Dunphyet al. studied 914 patients treated with radical RT alonebetween 1956 and 1985, of whom 656 had an initial Hbvalue recorded (6). Using a cutpoint of 13.5 g/dL, noassociation was seen between pretreatment Hb and eitherdisease-specific survival or local control. In a recent studyfrom Boston, 133 men received 6 months’ androgen depri-vation, with radical RT starting after 2 months’ hormonetreatment; an Hb level�13 g/dL at the start of RT wasassociated with an increased risk of PSA failure (p � 0.04)(7). One possible explanation for these discrepant results is

Reprint requests to: Dr. Christopher C. Parker, B.A., M.R.C.P.,M.D., F.R.C.R., The Academic Department of Radiotherapy andOncology, The Royal Marsden NHS Trust and The Institute ofCancer Research, Downs Road, Sutton, Surrey SM2 5PT, UnitedKingdom. Tel: �44 (208) 661 642 6011, ext. 3425; Fax:�44(208) 643 8809.

This work was undertaken in The Royal Marsden NHS Trust,who received a proportion of its funding from the NHS Executive;

the views expressed in this publication are those of the authors andnot necessarily those of the NHS Executive. This work was sup-ported by the Institute of Cancer Research, the Bob ChampionCancer Trust, and Cancer Research UK Section of RadiotherapyNCRI South of England Prostate Cancer Collaborative.

Received Mar 24, 2003, and in revised form Jun 18, 2003.Accepted for publication Jun 20, 2003.

Int. J. Radiation Oncology Biol. Phys., Vol. 58, No. 1, pp. 53–58, 2004Copyright © 2004 Elsevier Inc.

Printed in the USA. All rights reserved0360-3016/04/$–see front matter

53

the use, in the Boston series, of neoadjuvant androgendeprivation (NAD), which lowers Hb levels. The magnitudeof the effect of androgen deprivation on Hb level dependson the duration of treatment (8, 9), so the Boston studycould have important implications for the timing of neoad-juvant androgen deprivation before RT and for the possiblerole of strategies to correct anemia in these patients.

The primary aim of the current study was to test thehypothesis that Hb level at the start of RT is a significantdeterminant of outcome in localized prostate cancer. Asecondary aim was to explore the possibility that the prog-nostic effect of pre-RT Hb level is specific to men receivingprior androgen deprivation. We have therefore analyzedbiochemical outcome in relation to pre-RT Hb levels in twocohorts of men with clinically localized disease, one cohorttreated with RT alone, and the other with NAD � RT.

METHODS AND MATERIALS

Radiotherapy-alone cohortBetween 1987 and 1994, 793 men were treated at Prin-

cess Margaret Hospital with external beam RT for histolog-ically proven, clinically localized prostate cancer (T1/2,NX/N0, M0). Patients who received pre-RT hormone ther-apy (85) or palliative local RT (2) were excluded, leaving706 for analysis. Staging investigations included a bonescan, CT scan of the pelvis, bipedal lymphogram in 210(30%), and lymph node biopsy or sampling in 126 (18%).The tumor stage was recorded prospectively according tothe 1987 UICC classification (10) in use at the time. Aninitial serum hemoglobin level before the start of radiationtreatment was available in 536 cases. These 536 cases donot differ significantly from the entire cohort with respect tostage, Gleason score, or initial PSA. Between 1987 and1991, it was routine practice to obtain a serum hemoglobinlevel weekly during radiation. At least one hemoglobinlevel during radiation treatment was available in 119 cases.The radiotherapy technique has been described elsewhere(11). Radiation doses were prescribed to the isocenter, andthe two most common dose/fractionation schemes usedwere 65 Gy in 35 fractions (473 cases), or 65 Gy in 30fractions (205 cases). Patient characteristics are shown inTable 1.

Neoadjuvant androgen deprivation and radical RT cohortSix hundred fifty-eight men with histologically proven,

clinically localized (T1–4, N0/NX, M0) prostate cancer,who were treated at the Royal Marsden Hospital between1989 and 2000, received NAD and radical RT. Pretreatmentinvestigations included clinical examination, serum PSAmeasurement, CT or MRI scan of pelvis, bone scan, P/Achest radiograph, and prostatic sextant biopsies or transure-thral prostate resection. Pelvic lymph node sampling wasnot performed. An initial serum hemoglobin level, beforethe start of androgen deprivation, was available in 475cases. Androgen deprivation was achieved by an initial3-week course of cyproterone acetate, 100 mg tds orally,

together with monthly s.c. leuprorelin 3.75 mg or goserelin3.6 mg, starting 1 week after cyproterone and continuinguntil the completion of RT. Radical RT was intended tocommence after 3 months of androgen deprivation, butlonger courses of initial hormone treatment were allowedfor men with bulky local disease or obstructive urinarysymptoms. A pre-RT hemoglobin level, performed within 1month before RT, was available in 513 cases. These 513cases do not differ significantly from the entire cohort withrespect to stage, Gleason score, or initial PSA. The RTtechnique has been described elsewhere (12). Adjuvant hor-mone therapy after radiotherapy was not used.

Follow-upEach cohort was followed using digital rectal examina-

tion and serum PSA 2 times a year for the first 5 years afterRT and once a year thereafter. Bone scans and other imag-ing modalities were obtained only when dictated by clinicalsuspicion of relapse.

Statistical methodsThe primary end point was the time from the start of RT

to biochemical failure (TTBF), which, for the RT data set,was defined as 3 consecutive rising PSA values, accordingto the American Society for Therapeutic Radiology andOncology consensus statement (13). For the NAD � RTdata set, given the spurious PSA rise that is often seenduring testosterone recovery (14), biochemical failure wasdefined as 2 consecutive, rising PSA values and PSA �1.5ng/mL, according to the Fox Chase Cancer Center definition(14). Candidate predictive factors for time to biochemicalfailure after radiotherapy included initial PSA, biopsy Glea-son score, tumor stage, pre-NAD Hb level (for the NAD �RT data set), pre-RT Hb level, nadir Hb level (lowest Hbvalue during RT), and Hb decrement (difference betweenpre-NAD Hb and pre-RT Hb). Continuous variables weredichotomized about the median value. In addition, pre-RTHb was also analyzed, both as a continuous variable andusing a cutpoint of 13 g/dL, specified a priori on the basisof the previous report (7). The association between each ofthe candidate predictive factors and outcome was testedwith the log–rank test or, if there was a natural ordering ofthe categories (e.g. Gleason score), the log–rank test fortrend. Multivariate analysis was performed using the Coxproportional hazards model, and the results were expressedby hazard ratios and their 95% confidence intervals. Theassumptions involved with the Cox model (linearity, pro-portional hazards) were checked and were found not to be inviolation. The two cohorts differ from one another in termsof selection factors, radiation technique, and definition ofbiochemical failure, so all analyses were performed sepa-rately for each of the two data sets.

RESULTS

For the RT-alone data set, median follow-up was 4.3years, and 5-year freedom from biochemical failure, 44%

54 I. J. Radiation Oncology ● Biology ● Physics Volume 58, Number 1, 2004

(95% confidence interval: 40%–48%). On univariate anal-ysis (See Table 2), no association was seen between TTBFand pre-RT Hb (p � 0.45) or nadir Hb (p � 0.75). For theNAD � RT data set, median follow-up was 3.9 years, and5-year freedom from biochemical failure, 48% (95% con-fidence interval: 44%–53%). On univariate analysis (SeeTable 2), no association was seen between TTBF andpre-RT Hb (p � 0.87), pre-NAD Hb (p � 0.39), nadir Hb(p � 0.36), or Hb decrement (p � 0.45). Figure 1 illustratesthe outcome in terms of time to biochemical failure bypre-RT Hb using a cutpoint at the median Hb value.

Multivariate analysis of factors associated with TTBF forthe RT-alone data set (Table 3) was based on a total of 432cases and 224 events. Initial PSA (p � 0.001) and Gleasonscore (p � 0.001), but not pre-RT Hb (p � 0.30), weresignificant predictors of outcome. Multivariate analysis offactors associated with TTBF for the NAD � RT data set(Table 3) was based on a total of 441 cases and 215 events.

Initial PSA (p � 0.001), Gleason score (p � 0.003), and Tstage (p � 0.001), but not pre-RT Hb (p � 0.98) orpre-NAD RT (p � 0.24), were significant predictors ofoutcome.

DISCUSSION

In men with clinically localized prostate cancer, no sig-nificant association was observed between hemoglobin lev-els and biochemical outcome after radical radiotherapy.This lack of association was seen both for men treated withRT alone and for those receiving NAD and RT. This con-trasts with findings in other tumor types and suggests thatthe prognostic role of pre-RT Hb is tumor site specific andthat strategies to improve radiotherapy outcomes by correct-ing anemia are not warranted in prostate cancer.

Similarly, Dunphy et al. found no association betweenpre-RT Hb and treatment outcome, but because this study

Table 1. Patient characteristics

NAD � RT cohort n (%)RT cohort n

(%)

Cases 658 (100) 706 (100)Age

Median 68 69Range 49–83 48–86

T category1 65 (10) 162 (23)2a – 122 (17)2b – 200 (28)2c 243 (37) 217 (31)2x – 5 (1)3 303 (46)4 28 (4)missing 19 (3)

Gleason score2–4 88 (13) 2–6 377 (53)5–7 446 (68) 7 217 (31)�7 112 (17) �7 81 (11)

Missing 12 (2) Missing 31 (4)Initial PSA

0–4.0 33 (5) 80 (11)4.1–10.0 145 (22) 189 (27)10.1–20.0 176 (27) 187 (26)�20 252 (38) 150 (21)Missing 51 (8) 100 (14)

Pre-NAD hemoglobinn 475Median 14.4 g/dLRange 11.1–18.3 g/dL

Pre-RT Hemoglobinn 513 536Median 13.5 g/dL 14.8 g/dLRange 10.2–16.5 g/dL 10.5–19.2 g/dL

Nadir Hemoglobinn 429 119Median 13.1 g/dL 14.1 g/dLRange 9.3–15.8 g/dL 10.5–16.2 g/dL

Hemoglobin decrementn 330Median 0.9 g/dLRange �3.7 to �3.4 g/dL

55Hb and prostate cancer radiotherapy ● C. PARKER et al.

consists of patients treated between 17 and 46 years ago, itis of uncertain relevance to contemporary clinical practice(6). D’Amico et al. studied 133 men with localized prostatecancer who were managed using radical RT (70 Gy) and 6months of androgen deprivation (2 months’ neoadjuvant,concurrent, and adjuvant). PSA failure was defined as 2consecutive rising PSA values �1.0 ng/mL after accountingfor PSA rebound due to testosterone recovery. The initial

results were presented to the American Society for Thera-peutic Radiology and Oncology in 2001, and the 3-yearestimate of PSA control was 95% vs. 63% (p � 0.03) forpatients with a pre-RT Hb �13 g/dL (n � 62) vs. �13 g/dL(n � 71), respectively. The relative risk of PSA failure forpatients with a pre-RT Hb level �13 g/dL was 6 (95%confidence interval: 1.5–9). An updated analysis of theseresults, based on a total of 110 patients, included an addi-tional variable, namely the decline in Hb during the firstmonth of androgen deprivation (15). On multivariate anal-ysis, this decline in Hb was the only significant predictor of

Table 2. Univariate analysis of time to biochemical failure withrespect to Hb variables

NAD � RT data set RT data set

n% 2-yrbNED p value n

% 2-yrbNED p value

Pre-RT Hb�median 251 69.9 p � 0.90 266 68 p � 0.45�median 262 66.0 247 69

Pre-RT Hb�13.0 g/dL 144 68.1 p � 0.87 43 69 p � 0.52�13.0 g/dL 369 67.9 470 69

Pre-NAD Hb�median 220 75.7 p � 0.39�median 255 77.9

Nadir Hb�median 212 74.3 p � 0.36 64 70 p � 0.75�median 217 70.9 50 66

Hb decrement�median 170 72.2 p � 0.45 59 65 p � 0.39�median 160 72.3 55 71

Abbreviations: Hb � hemoglobin; NAD � neoadjuvant andro-gen deprivation; RT � radiotherapy; bNED � biochemical noevidence of disease.

Fig. 1. Freedom from biochemical failure by pre-RT Hb for NAD � RT data set. Hb � hemoglobin; NAD �neoadjuvant androgen deprivation; RT � radiotherapy.

Table 3. Multivariate analysis of factors associated with time tobiochemical failure

VariableHazard ratio

(HR) 95% CI p value

NAD � RT data setInitial PSA 2.40 1.77–3.25 �0.001Gleason score* 1.71 1.24–2.35 0.001T stage (T1/2 vs.

T3/4)1.73 1.29–2.32 �0.001

Pre-RT Hb 1.00 0.77–1.32 0.98Pre-NAD Hb 1.19 0.89–1.60 0.24

RT-alone data setPre-RT PSA 1.46 1.26 to 1.69 �0.001Gleason score† 1.44 1.19 to 1.76 �0.001T stage (T1 vs T2) 1.08 0.96 to 1.22 0.21Hemoglobin 1.15 0.96 to 1.38 0.30

* GS 5–7 vs GS 8–10.† GS 2–6 vs 6 vs � 7.Abbreviations: CI � confidence interval; NAD � neoadjuvant

androen deprivation; RT � radiotherapy.

56 I. J. Radiation Oncology ● Biology ● Physics Volume 58, Number 1, 2004

time to PSA failure (p � 0.02), and pre-RT Hb was nolonger statistically significant (p � 0.11). However, theresults of this study must be interpreted with considerablecaution, because they are based on a mere 10 events. Thiscontrasts markedly with the 247 events in the current NAD� RT analysis.

Might we have found a significant association betweenpre-RT Hb and outcome had we chosen a different Hbcutpoint? The rationale for choosing to analyze Hb using acutpoint of 13 g/dL was based on the findings of the Bostonstudy (7). Studies in other tumor types that have demon-strated the prognostic significance of pre-RT Hb level haveemployed a range of other Hb cutpoints (5). There is atheoretical basis for a nonlinear relationship betweenpre-RT Hb and outcome, with an adverse effect of bothabnormally low and abnormally high Hb levels (16). As adata-driven, hypothesis-generating exercise, we consideredthe optimum method for modeling pre-RT Hb. Analysis ofboth data sets (data not shown) indicated both that there wasno Hb cutpoint for which a significant association withoutcome would be found and that there was no suggestionof an adverse effect on outcome at both high and lowextremes of Hb level.

Although racial data were not available, both cohortsanalyzed originate from predominantly Caucasian popula-tions. The African-American population has a lower distri-bution of hemoglobin levels, and so our findings may not begeneralizable to all ethnic groups.

We considered whether variation in the duration of neo-adjuvant androgen deprivation, within the permitted rangeof 3 to 6 months, might have influenced our findings.However, duration of androgen deprivation showed no sig-nificant correlation with Hb decrement (p � 0.24, Pearsoncoefficient). Furthermore, the lack of prognostic signifi-cance of pre-RT Hb persisted (p � 0.87) when duration ofandrogen deprivation was included in the multivariatemodel of outcome.

The demonstration, in other tumor types, of a significantrelationship between pre-RT Hb and treatment outcome hasled to considerable interest in the potential benefit of cor-recting low Hb levels before and during RT. Ongoing ran-domized trials are studying the effect of recombinant humanerythropoietin (rHuEPO) on the efficacy of radical radio-therapy (17). Our results suggest that such trials are lessappropriate in prostate cancer than in other tumor types.However, we have confirmed the finding that use of NAD isassociated with a decline in Hb levels, with the median Hb

level changing from 14.4 g/dL pre-NAD to 13.5 g/dLpre-RT (p � 0.0001). In 133 men receiving 3 months’ totalandrogen blockade, Strum et al. reported a mean decline inserum hemoglobin level of 1.8 g/dL (8). Asbell et al.observed a mean fall in hemoglobin of 3.1 g/dL in 141 menreceiving 4 months of goserelin and flutamide and radicalradiotherapy to the pelvis and prostate (9). Although ourresults suggest that this fall in Hb level does not impact onthe efficacy of radical radiotherapy in prostate cancer, none-theless it may account for some of the detriment in health-related quality of life seen with androgen deprivation (18).Thus, trials of rHuEPO may be warranted in men receivingandrogen deprivation for prostate cancer, with health-re-lated quality of life as the main end point.

If pre-RT Hb is a determinant of outcome for some butnot all cancer types, what could account for this tumor sitespecificity? The biologic basis for the association betweenHb levels and RT outcome in other tumor types is not wellunderstood. There are animal data to suggest that low Hblevels are associated with increased tumor hypoxia (19), andhypoxia is known to increase radioresistance (20). In humancancers, an association between Hb levels and tumor hyp-oxia has only been observed at Hb levels �11 g/dL (21),levels that are common in, say, cervix cancer but not inlocalized prostate cancer. Thus, pre-RT Hb may not impacton outcome of prostate radiotherapy, because the distribu-tion of Hb values is too high to affect tumor oxygenation.Hb level may be associated also with treatment outcome byserving as a surrogate marker of tumor extent (See Ref. 16for a review). The strength of the association between serumHb level and tumor extent is likely to vary from one cancertype to another, not least because some types of cancer (butnot prostate cancer) commonly cause blood loss. Thus, analternative hypothesis is that pre-RT Hb may predict RToutcome for cervix cancer but not prostate cancer, becausethe Hb level is a better marker of disease extent in theformer.

In summary, in a large study of men with localizedprostate cancer, treated with radical radiotherapy eitheralone or combined with neoadjuvant androgen deprivation,we have observed no statistically significant associationbetween hemoglobin levels and treatment outcome. Thiscontrasts with findings in other tumor types. The prognosticrole of hemoglobin level may be tumor site specific. Trialsto determine the effect of rHuEPO on the efficacy of radicalradiotherapy are not warranted in prostate cancer.

REFERENCES

1. Grogan M, Thomas GM, Melamed I, et al. The importance ofhemoglobin levels during radiotherapy for carcinoma of thecervix. Cancer 1999;86(8):1528–1536.

2. Warde P, O’Sullivan B, Bristow RG, et al. T1/T2 glotticcancer managed by external beam radiotherapy: The influenceof pretreatment hemoglobin on local control. Int J RadiatOncol Biol Phys 1998;41(2):347–353.

3. Cole CJ, Pollack A, Zagars GK, Dinney CP, Swanson DA,

von Eschenbach AC. Local control of muscle-invasive bladdercancer: Preoperative radiotherapy and cystectomy versus cystec-tomy alone. Int J Radiat Oncol Biol Phys 1995;32(2):331–340.

4. Dische S, Warburton MF, Saunders MI. Radiation myelitisand survival in the radiotherapy of lung cancer. Int J RadiatOncol Biol Phys 1988;15(1):75–81.

5. Grau C, Overgaard J, editors. Significance of hemoglobinconcentration for treatment outcome. Berlin: Springer; 1998.

57Hb and prostate cancer radiotherapy ● C. PARKER et al.

6. Dunphy EP, Petersen IA, Cox RS, Bagshaw MA. The influ-ence of initial hemoglobin and blood pressure levels on resultsof radiation therapy for carcinoma of the prostate. Int J RadiatOncol Biol Phys 1989;16(5):1173–1178.

7. D’Amico AV, Saegart T, Chen M, et al. The impact of anemiaon PSA outcome following radiation and androgen suppres-sion therapy in patients with high-risk prostate cancer. AS-TRO San Francisco 2001.

8. Strum SB, McDermed JE, Scholz MC, Johnson H, Tisman G.Anaemia associated with androgen deprivation in patientswith prostate cancer receiving combined hormone blockade.Br J Urol 1997;79(6):933–941.

9. Asbell SO, Leon SA, Tester WJ, Brereton HD, Ago CT,Rotman M. Development of anemia and recovery in prostatecancer patients treated with combined androgen blockade andradiotherapy. Prostate 1996;29(4):243–248.

10. Sobin LH, Hermanek P, Hutter RV. TNM classification ofmalignant tumors. A comparison between the new (1987) andthe old editions. Cancer 1988;61(11):2310–2314.

11. Catton C, Gospodarowicz M, Mui J, et al. Clinical and bio-chemical outcome of conventional dose radiotherapy for lo-calized prostate cancer. Can J Urol 2002;9(1):1444–1452(discussion 1453).

12. Parker CC, Norman AR, Huddart RA, Horwich A, DearnaleyDP. Pre-treatment nomogram for biochemical control afterneoadjuvant androgen deprivation and radical radiotherapy forclinically localised prostate cancer. Br J Cancer 2002;86(5):686–691.

13. Consensus statement: Guidelines for PSA following radiationtherapy. American Society for Therapeutic Radiology andOncology Consensus Panel. Int J Radiat Oncol Biol Phys1997;37(5):1035–1041.

14. Hanlon AL, Hanks GE. Scrutiny of the ASTRO consensusdefinition of biochemical failure in irradiated prostate cancerpatients demonstrates its usefulness and robustness. AmericanSociety for Therapeutic Radiology and Oncology. Int J RadiatOncol Biol Phys 2000;46(3):559–566.

15. D’Amico AV, Saegaert T, Chen MH, et al. Initial decline inhemoglobin during neoadjuvant hormonal therapy predicts forearly prostate specific antigen failure following radiation andhormonal therapy for patients with intermediate and high-riskprostate cancer. Cancer 2002;95(2):275–280.

16. Fyles AW, Milosevic M, Pintilie M, Syed A, Hill RP. Anemia,hypoxia and transfusion in patients with cervix cancer: Areview. Radiother Oncol 2000;57(1):13–19.

17. Antonadou D, Cardamakis E, Sarris G, et al. Effect of theadministration of recombinant human erythropoietin in pa-tients with pelvic malignancies during radiotherapy. RadiotherOncol 1998;S48:483.

18. Potosky AL, Knopf K, Clegg LX, et al. Quality-of-life out-comes after primary androgen deprivation therapy: Resultsfrom the Prostate Cancer Outcomes Study. J Clin Oncol2001;19(17):3750–3757.

19. Kelleher DK, Mattheinsen U, Thews O, Vaupel P. Blood flow,oxygenation, and bioenergetic status of tumors after erythro-poietin treatment in normal and anemic rats. Cancer Res1996;56(20):4728–4734.

20. Gray LH, Conger AD, Ebert M, et al. The concentration ofoxygen dissolved in tissues at the time of radiation as a factorin radiotherapy. Br J Radiol 1953;26:638–648.

21. Becker A, Stadler P, Lavey RS, et al. Severe anemia isassociated with poor tumor oxygenation in head and necksquamous cell carcinomas. Int J Radiat Oncol Biol Phys2000;46(2):459–466.

58 I. J. Radiation Oncology ● Biology ● Physics Volume 58, Number 1, 2004