Clinical comparison of two linear-quadratic model-based isoeffect fractionation schemes of high-dose-rate intracavitary brachytherapy for cervical cancer

C

R1(cAD

Int. J. Radiation Oncology Biol. Phys., Vol. 59, No. 1, pp. 179–189, 2004Copyright © 2004 Elsevier Inc.

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

doi:10.1016/j.ijrobp.2003.10.025

LINICAL INVESTIGATION Cervix

CLINICAL COMPARISON OF TWO LINEAR-QUADRATIC MODEL-BASEDISOEFFECT FRACTIONATION SCHEMES OF HIGH-DOSE-RATEINTRACAVITARY BRACHYTHERAPY FOR CERVICAL CANCER

CHONG-JONG WANG, M.D.,* ENG-YEN HUANG, M.D.,* LI-MIN SUN, M.D.,* HUI-CHUN CHEN, M.D.,*FU-MIN FANG, M.D.,* HSUAN-CHIH HSU, M.D.,* CHAN-CHAO CHANGCHIEN, M.D.,† AND

STEPHEN WAN LEUNG, M.S., M.D.‡

Departments of *Radiation Oncology and †Gynecology Oncology, Kaohsiung Chang Gung Memorial Hospital; ‡Department ofRadiation Oncology, Yuan’s General Hospital, Kaohsiung, Taiwan

Purpose: Two linear-quadratic model–based isoeffect fractionation schemes of high-dose-rate intracavitarybrachytherapy (HDR-IC) were used to treat cervical cancer in two consecutive periods. Patient outcomes andcomplications were analyzed and compared.Methods and Materials: Between November 1987 and December 1996, a total of 541 women diagnosed withcervical cancer were treated with curative-intent radiotherapy. Patients were categorized into two groupsaccording to the two isoeffect schemes used. Group 1 consisted of 254 patients treated with external beamradiotherapy (EBRT) plus 7.2 Gy HDR-IC to Point A for three fractions in the first period. Group 2 consistedof 284 patients treated with EBRT plus 4.8 Gy HDR-IC for five fractions in the second period. The goal of thenew scheme for the latter group was to deliver an isoeffect dose that maintained similar tumor control butreduced normal tissue complications. The calculated biologically effective dose (BED10, assuming an�/� ratio �10) of EBRT plus HDR-IC for tumor and acute responding tissue in Groups 1 and 2 was 90 Gy10 (52.8� 37.2Gy) and 88.6 Gy10 (53.1� 35.5 Gy), respectively. The corresponding BED3 for late responding tissue (assumingan �/� ratio � 3) in Groups 1 and 2 was 146.7 Gy3 (73.3� 73.4 Gy) and 134.4 Gy3 (72 � 62.4 Gy), respectively.Patients were followed for 6.1–15.2 years (median, 9.8 years).Results: Overall, 66 patients (12.2%) developed pelvic recurrence. Of these, 53 patients had central recurrence.Of the 53 patients with central recurrence, 24 (9.4%) were in Group 1 and 29 (10.1%) in Group 2 (p � 0.722).The actuarial pelvic control rate for Groups 1 and 2 was 88.2% and 86.3% at 5 years and 87.3% and 85.5% at10 years, respectively (p � 0.504). The actuarial overall survival rate for Groups 1 and 2 was 63.5% and 56.1%at 5 years and 47.8% and 49.3% at 10 years, respectively (p � 0.734). The actuarial proctitis rate for Groups 1and 2 was 49.7% and 32.7% at 5 years and 50.5% and 32.7% at 10 years, respectively (p <0.001). Most of thedecrease in the rate of proctitis was a result of a decrease in the incidence of low-grade proctitis (38% vs. 22%).The incidence of high-grade complications remained unchanged, 8% vs. 7%. The actuarial cystitis rate forGroups 1 and 2 was 14.3% vs. 11.4% at 5 years and 24.1% vs. 15% at 10 years, respectively (p � 0.134).Multivariate analysis revealed that the fractionation scheme (three fractions vs. five fractions) was a significantfactor influencing the proctitis rate (p � 0.004, hazard ratio� 0.807; 95% confidence interval, 0.697–0.934), butnot the local pelvic control rate, overall survival rate, or cystitis rate.Conclusion: The treatment results of the two groups maintained similar outcomes, while the complicationsdecreased. The linear-quadratic model correctly predicted this outcome. Biologically, the manipulation of thefraction size in our study suggested that the sensitivity of the late responding tissue to the fractional change from7.2 Gy to 4.8 Gy in HDR-IC is high and detectable clinically. The success, however, had its limitations, and theimprovement was confined to low-grade complications. © 2004 Elsevier Inc.

Cervical carcinoma, Linear-quadratic model, High-dose-rate, Fractionation scheme, Brachytherapy.

Reprint requests to: Chong-Jong Wang, M.D., Department ofadiation Oncology, Kaohsiung Chang Gung Memorial Hospital,23 Ta-Pei Rd., Niao-Sung Hsian, Kaohsiung Hsien, Taiwan. Tel:�886) 7-7317123, ext. 2600; Fax: (�886) 7-7322813; E-mail:[email protected]—We thank Dr. Yeh-Chi Lo, whose visit to ourepartment in 1993 and help in formulation substantiated the

change in the fractionation scheme used in this report. Dr. Lo wasthen a medical physicist at the Memorial Sloan-Kettering CancerCenter in New York, and is now at Mount Sinai Hospital, NewYork.

Received Jul 17, 2003, and in revised form Sep 29, 2003.Accepted for publication Oct 15, 2003.

179

WroTectbPtaGoectfsc

H(deltliatcmlhrtapt

P

ctcss(1t1yc

Pcew

E

t(cstEw(tsGzwpi1a1apcd

H

toHpfeuitnwwpRraw

RstGtwl

180 I. J. Radiation Oncology ● Biology ● Physics Volume 59, Number 1, 2004

INTRODUCTION

ith the data accumulated during the past years, high-dose-ate intracavitary brachytherapy (HDR-IC) has gained rec-gnition and popularity in the treatment of cervical cancer.he optimal fractionation scheme for this technique, how-ver, remains undetermined. A meta-analysis of the subjectonducted by Petereit and Pearcey (1) in 1999 concludedhat the optimal schedule was still unknown and could beased only on single institutions with significant experience.ublications on the clinical comparison of different frac-

ionation schemes in HDR-IC were limited. Le Pechoux etl. (2) comparing two different fractionation schemes of 5y twice weekly vs. 6 Gy once weekly, concluded thatnce-weekly HDR combined with properly adjusted pelvicxternal beam radiotherapy was safe and effective. Re-ently, Wong et al. (3) reported a comparison of two frac-ionation schemes, 7 Gy for three fractions vs. 6 Gy for fourractions, and concluded that the two schemes were not atatistically significant prognosticator in predicting diseaseontrol and complications.

The biologic basis of these fractionation schemes forDR-IC and its intercomparison with the low-dose-rate

LDR) setting was derived from a linear-quadratic model, asescribed by several authors (4–9). The model largely cat-gorized normal tissues into acute (including tumor) andate responding tissues. Characteristically, late respondingissues exhibited a survival curve of a flat shoulder at aow-dose region and a faster bend at a high-dose region,ndicating that the tissues responsible for late complicationsre more sensitive to fraction size (smaller �/� ratio). Inheory, if we split larger fraction dose into smaller ones, wean decrease complications without compromising the tu-or control rate. Whether the theory works and can trans-

ate into the expected outcome in the brachytherapy setting,owever, had not been thoroughly tested clinically. In thiseport, we analyzed 541 HDR-IC patients in two consecu-ive periods using two isoeffect schemes. Patient outcomend treatment-related complications are reported and com-ared. The validity of the model on the prediction of normalissue complications in the HDR-IC setting was examined.

METHODS AND MATERIALS

atient characteristicsThe analysis included 541 women who were treated with

urative-intent radiotherapy (RT) for cervical cancer be-ween November 1987 and December 1996. Patients wereategorized into two groups according to the fractionationcheme used in the two consecutive periods. Group 1 con-isted of 254 patients treated with external beam RTEBRT) plus three-fraction HDR-IC between November987 and April 1993. Group 2 consisted of 287 patientsreated with EBRT plus 5-fraction HDR-IC between May993 and December 1996. Patient age ranged from 31 to 86ears (median 61). The Federation Internationale de Gyne-ologie et d’Obstetrique (FIGO) staging system was used.

retreatment evaluation and staging workup, includingomplete blood count, chest radiography, intravenous py-lography, CT of the pelvis, cystoscopy, and sigmoidoscopyere done as appropriate.

xternal beam RTThe technical details of the EBRT and HDR-IC used in

his department have been previously published elsewhere10, 11). Because the departmental policies and facilitieshanged during the two treatment periods, the present de-cription focused mainly on the contrast between the tworeatment groups. All patients were treated initially withBRT. In Group 1, whole pelvic EBRT was administeredith 10-MV photons through AP parallel-opposing ports

AP–PA). The daily dose was 2 Gy. A four-field “box”echnique was not routinely used in these patients. Thetandardized dose to the central cervix ranged from 40 to 44y in 20–22 fractions. An additional boost of 0–14 Gy in

ero to seven fractions was given to the bilateral pelvic wall,ith a 4-cm midline block to shield the central structures inatients with parametrial or suspected pelvic lymph nodenvolvement, so that total dose to the pelvic wall in Groupwas maintained at 40–58 Gy. With the implementation ofmultileaf collimator-equipped linear accelerator in early

995, the daily dose was reduced from 2 Gy to 1.8 Gy, andfour-field box technique was routinely used in all the

atients treated thereafter. Thus, in Group 2, the dose to theentral cervix was 39.6–45 Gy in 22–25 fractions, and theose to the side wall was maintained at 39.6–59.6 Gy.

igh-dose-rate ICThe HDR-IC techniques were similar across the two

reatment periods. HDR-IC was given after the completionf EBRT in all patients. An interdigitated arrangement ofDR-IC during the EBRT course was not applied in ouratients. On a routine basis, the first application was per-ormed under general anesthesia. Tumor regression wasvaluated, and the endocervical canal was dilated. Theterine sounding was measured, and the applicator wasnserted. The applicator used in all patients was Henschke’sype, consisting of one curved central tandem and twoonshielded ovoids. At the end of the procedure, the vaginaas packed tautly with contrasted gauze. A urinary catheterith a urograffin-filled balloon was installed to outline theosterior bladder wall. The International Committee onadiation Units and Measurements recommended points for

ectal/bladder dose evaluation in Report No. 38 were notvailable until 1991. Before this, orthogonal maximal pointsere used for rectal/bladder dose evaluation.The HDR-IC facility used in Group 1 was mainly the

alstron system (Toshiba, Japan) using 60Co sources. Theystem was replaced by a microSelectron system (Nucle-ron) using a 192Ir source in early 1993. The patients inroup 2 were all treated with this new system. During the

ransition of the two systems, brachytherapy in 40 patientsere performed at a neighboring institution using the Se-

ectron system. These patients were not included in the

ao

ttmgiwcf�rtGopp

baedttptTdcidcut0adrTcicvw

C

socp

F

rw

OfrPtttwr

ecbirtimcpwtrTdwrUobtpcwcPs

G

2vm2ufwecedrstT

181Two isoeffect schemes of HDR-IC ● C.-J. WANG et al.

nalysis because the method of the source dwell and planptimization were not consistent with ours.The standard fractionation scheme in Group 1 was 7.2 Gy

o Point A for three fractions with each fraction adminis-ered 2 weeks apart. On the basis of the linear-quadraticodel, the scheme was changed to 4.8 Gy for five fractions

iven twice weekly in May 1993. The goal was to deliver ansoeffect dose that maintained a similar tumor control rate,hile hopefully reducing normal tissue complications. The

alculated biologically equivalent dose (BED10) of HDR-ICor tumor and acute responding tissue (assuming �/� ratio

10) in Groups 1 and 2 was 37.2 Gy10 and 35.5 Gy10,espectively. The corresponding BED3 for late respondingissue (assuming �/� ratio � 3) for the 2 groups was 73.4y3 and 62.4 Gy3. For medical considerations, a reductionf 10–20% of the total HDR-IC dose was allowed in agedatients with a favorable tumor response after EBRT or inatients with severe chronic disease in both groups.The rules of source dwell and plan optimization have

een previously published (10). These rules were consistentcross the two treatment periods. The method was partlyxtended from the Manchester technique. The concept ofifferential loading in tandem was generally applied to ourreatment. This was achieved by modulating the weight ofhe dwell time in the tandem. The rule was that the highestoint of the tandem near the fundus was weighted 1.5 andhe lowermost point near the cervical os was weighted 0.8.he rest of tandem points were equally weighted at 1.0. Theose contribution of the central tandem vs. both ovoids wasonsistently 3:1. For the Ralstron unit, the stepwise distances adjustable in 0.1-cm intervals; usually 0.9–1.3 cm (me-ian, 1.1 cm) was used. The stepwise distance of the mi-roSelectron unit is formatted, and 0.5 cm was commonlysed. The dwell positions in tandem were adjusted so thathe lowermost point (with 0.8 weighting) was generally.6–0.8 cm above the cervical os. This constraint was set tovoid an excessive dose to the rectovaginal septum. Theose to the point of cervix os was routinely calculated andequired to be maintained about twofold of that of Point A.he dose to Point B and the contralateral Point A werealculated in Group 1, but not in Group 2. A newer point ofnterest, defined as 2 cm directly caudal to the cervix os, wasalculated later. The constraint was set to ensure adequateaginal coverage. The dose required for this vaginal pointas about 100% of Point A.

hemotherapyA total of 32 patients received chemotherapy in this

tudy. Its administration was individualized and was largelyn a neoadjuvant basis. The selection criteria included lo-ally advanced stage, bulky tumor, positive nodes, andatient preference.

ollow-up and statistical analysisPatient, tumor, and treatment characteristics are summa-

ized in Tables 1 and 2. After treatment completion, patientsere regularly followed at the Department of Radiation

ncology and Gynecology. Living patients were followedor 6.1–15.2 years (median, 9.8 years). Patients withoutegular visits were interviewed by telephone and/or letter.atients lost to follow-up were censored and the survival

ime was counted from the day of treatment completion tohe day of the last contact/visit. For a meaningful analysis,he small number of patients with Stage IIA, IIIA, and IVAere included with those with Stage IB, IIB, and IIIB,

espectively.Pelvic recurrence was categorized as central or periph-

ral. Central recurrence was defined as recurrence in theervix, parametrium, or upper vagina that was encompassedy brachytherapy. Proof of recurrence by pathologic exam-nation in these patients was usually possible. Peripheralecurrence, on the other hand, was defined as recurrence inhe side wall, obturator nodes, iliac nodes, lower vagina, orntroitus. Any extrapelvic disease was considered distantetastasis. The duration of pelvic control and survival was

ounted from the day of RT completion. The actuarialelvic control rate and overall survival rate were estimatedith the Kaplan-Meier product-limit method, and their sta-

istical significance tested by log–rank test. The severity ofectal complications was graded according to the Radiationherapy Oncology Group criteria (12). For analysis andiscussion, Grade 1-2 vs. Grade 3-4 rectal complicationsere considered low-grade vs. high-grade complications,

espectively. The International Committee on Radiationnits and Measurements recommended point for evaluationf the rectal/bladder dose was reported, but not analyzed,ecause this information was not available in patients fromhe earlier period. Confounding factors that may influenceelvic control, overall survival, and rectal/bladder compli-ations were tested with Cox’s regression model in a step-ise forward manner. Data processing and statistics were

arried out on a personal computer using the Statisticalackage for Social Sciences (SPSS) Software, version 11,oftware (SPSS, Chicago, IL) for Windows.

RESULTS

eneral dataAs of March 2003, 263 patients (49%) were alive, and

78 patients (51%) had died; 66 patients (12.2%) had de-eloped tumor recurrence in the pelvis and 110 (20%) hadetastasis outside the pelvis. Rectal bleeding developed in

01 patients (37%) to a varying degree. Colostomy wasndertaken in 16 patients (3%) because of stenosis or fistulaormation in the rectosigmoid area. Hemorrhagic cystitisas observed in 71 patients (13%). Nineteen patients (4%)

xperienced chronic abdominal cramp pain, possibly be-ause of partial intestinal obstruction. Mild to moderate legdema occurred in 41 patients (8%). Thirteen patients (2%)eveloped ureteral obstruction and chronic hydronephrosiselated to the treatment. Nineteen patients (4%) developed aecond malignancy. The distribution of patient outcome andreatment-related complications by group is summarized inable 3.

P

ro(diprei

aw581t0

G�arG

06

O

o5o24

a63o6fI

R

v(sm

182 I. J. Radiation Oncology ● Biology ● Physics Volume 59, Number 1, 2004

elvic controlOf the 66 patients (12.2%) who developed pelvic recur-

ence, 30 patients were in Group 1 and 36 in Group 2. Thenset of the pelvic recurrence ranged from 3 to 149 monthsmedian, 8 months) for Group 1 and 3–87.7 months (me-ian, 8.7 months) for Group 2. Of the 30 pelvic recurrencesn Group 1, 24 were categorized as central recurrence, 5 aseripheral recurrence, and 1 as both. Of the 36 pelvicecurrences in Group 2, 29 were central and 7 were periph-ral. The distribution by group was not statistically signif-cant (p � 0.722; Table 4).

Regardless of treatment group or disease stage, the actu-rial pelvic control rate for the entire cohort of 541 patientsas 87.2% at 5 years and 86.8% at 10 years. By group, the-year actuarial pelvic control rate for Groups 1 and 2 was8.2% and 86.3%, respectively. The corresponding rate at0 years for Groups 1 and 2 was 87.3% and 85.5%, respec-ively. The difference was not statistically significant (p �.504, Fig. 1).

By stage, the 5-year actuarial pelvic control rate forroups 1 and 2 was 96.9% and 88.9% for Stage IB-IIA (p0.037); 91.6% and 88.6% for Stage IIB-IIIA (p � 0.526);

nd 68.9% and 69.9% for Stage IIIB-IVA (p � 0.784),espectively. The 5-year actuarial rate at 10 years forroups 1 and 2 was 94.4% and 87.2% for IB-IIA (p �

Table 1. Patients a

Characteristics

Treatment periodAge (range, 31–86 y; median, 61)

�60�60

BMI (range, 14.8–44.1 kg/m2; median, 23.6)�24�24Unknown

Hemoglobin (range, 4.8–16 g/dL; median, 12)�12�12Unknown

Pathologic typeSCCNon-SCC

StageIB � IIA (135 � 47)IIB � IIIA (248 � 12)IIIB � IVA (96 � 3)

Diabetes mellitusYesNo

HypertensionYesNo

Follow-upRangemedian

Abbreviation: BMI � body mass index (weigData in parentheses are percentages, unless ot

.037); 91.6% and 88.6% for IIB-IIIA (p � 0.526); and2.1% and 69.9% for IIIB-IVA (p � 0.784).

verall survival rateRegardless of treatment group or stage, the actuarial

verall survival rate for the entire cohort of 541 patients was9.8% at 5 years and 47.9% at 10 years. The 5-year actuarialverall survival rate was 63.5% and 56.1% for Groups 1 and, respectively. At 10 years, for Groups 1 and 2, it was7.8% and 49.3%, respectively (p � 0.734, Fig. 2).

The 5-year actuarial overall survival rate for Groups 1nd 2 was 81.4% and 67.5% for Stage IB-IIA (p � 0.352);5.8% and 54.8% for Stage IIB-IIIA (p � 0.514); and8.3% and 30.8% for Stage IIIB-IVA (p � 0.301). Theverall survival rate at 10 years for Groups 1 and 2 was3.8% and 58.3% for IB-IIA (p � 0.352); 49.4% and 51.3%or IIB-IIIA (p � 0.514); and 25.7% and 27.7% for IIIB-VA (p � 0.301).

ectal complicationsOverall, 201 patients developed rectal complication of

arying degree. Of these, 117 (46%) were in Group 1 and 8429%) in Group 2. The median onset for Groups 1 and 2 wasimilar, 11 months (range, 3.1–147 months) and 11.3onths (range, 3.1–40 months), respectively.

or characteristics

Overall(n � 541)

Group 1(n � 254)

Group 2(n � 287)

987–1996 1987–1993 1993–1996

247 (46) 120 (47) 127 (44)294 (54) 134 (53) 160 (56)

285 (53) 136 (54) 149 (52)252 (46) 117 (45) 135 (47)

4 (1) 1 (1) 3 (1)

235 (43) 111 (44) 124 (43)292 (54) 138 (54) 154 (54)

14 (3) 5 (2) 9 (3)

517 (96) 244 (96) 273 (95)24 (4) 10 (4) 14 (5)

182 (34) 70 (27) 112 (39)260 (48) 124 (49) 136 (47)

99 (18) 60 (24) 39 (14)

74 (14) 38 (15) 36 (13)467 (86) 216 (85) 251 (87)

104 (19) 41 (16) 63 (22)437 (81) 213 (84) 224 (78)

6.1–15.2 9.7–15.2 6.1–9.69.8 12.4 7.6

ht in square meters).e noted.

nd tum

1

ht/heigherwis

4G�

carlc

�7ctm

B

G

183Two isoeffect schemes of HDR-IC ● C.-J. WANG et al.

The 5-year actuarial proctitis rate for Groups 1 and 2 was9.7% and 32.7%, respectively. At 10 years, the rate forroups 1 and 2 was 50.5% and 32.7%, respectively (p0.001, Fig. 3).The number of patients with Grade 1, 2, 3 and 4 compli-

ation for Groups 1 and 2 was 56 (22%), 40 (16%), 14 (5%),nd 7 (3%) and 39 (14%), 24 (8%), 12 (4%), and 9 (3%),espectively. When the complications was categorized asow (Grade 1-2) and high (Grade 3-4) grade, the low-gradeomplication rate for Groups 1 and 2 was 38% and 22% (p

0.001), and the high-grade complication rate was 8% and%, respectively (Table 5). The reduction in rectal compli-ations in Group 2 mainly occurred in low-grade complica-ions, with the incidence of high-grade complications re-aining unchanged.

ladder complicationsOverall, 71 patients developed radiation cystitis, 45 in

roup 1 and 26 in Group 2. The median onset for Groups 1

Table 2. Treatment characteristics

CharacteristicOverall

(n � 541)Group 1

(n � 254)Group 2

(n � 287)

EBRTFour-field box technique

Yes 195 (36) 20 (8) 175 (61)No 346 (64) 234 (92) 112 (39)

Daily dose (Gy)1.8 223 (41) 0 (0) 223 (78)2 318 (59) 254 (100) 64 (22)

Dose to cervix (Gy) (median, 45)39.6–43.9 38 (7) 9 (4) 29 (10)44–45 430 (80) 207 (81) 223 (78)45.1–59.4 73 (13) 38 (15) 35 (12)

Dose to the parametrium (Gy) (median, 54)39.6–43.9 143 (26) 60 (24) 83 (29)44–54 152 (28) 34 (13) 118 (41)54.1–59.6 246 (46) 160 (63) 86 (30)

BED10 (range, 47–70, median, 53)47–52 32 (6) 8 (3) 24 (8)53 443 (82) 210 (83) 233 (81)54–70 66 (12) 36 (14) 30 (11)

BED3 (range, 63–107, median, 73)63–71 32 (6) 8 (3) 24 (8)72–73 443 (82) 210 (83) 233 (81)74–107 66 (12) 36 (14) 30 (11)

HDR-ICActive length (range, 1–9 cm; median, 4.5)

�4.5 251 (46) 137 (54) 114 (40)�4.5 290 (54) 117 (46) 173 (60)

Fractionation schemeStandardized � minor modification (�20%) 468 (87) 201 (79) 267 (93)�20% modification 73 (13) 53 (21) 20 (7)

BED10 (range, 12.4–52.2; median, 35.5)�28 56 (10) 49 (19) 7 (3)28–37.2 470 (87) 202 (80) 268 (93)�37.3 15 (3) 3 (1) 12 (4)

EBRT � HDR-ICTotal treatment days (range, 45–228; median, 71)

�70 242 (45) 34 (13) 208 (72)�70 299 (55) 220 (87) 79 (28)

Total BED10 (range, 70.8–106.8; median, 88.6)70.8–80.9 36 (17) 31 (12) 5 (2)81–90 453 (84) 207 (82) 246 (86)90.1–106.8 52 (9) 16 (6) 36 (12)

ChemotherapyYes 32 (6) 17 (7) 15 (5)No 509 (94) 237 (93) 272 (95)

Abbreviations: ERBT � external beam radiotherapy; BED � biologically effective dose; HDR-IC �high-dose-rate intracavitary brachytherapy.

Data in parentheses are percentages, unless otherwise noted.

a3

11

r

U

pwvm�Ivcpbtla

9i

0C�f

�h(w

(h10s

(w

riiO(s�tTspcf(Sr(emi

wep(midluat5tn

APDPCCCLUD

w

NP

184 I. J. Radiation Oncology ● Biology ● Physics Volume 59, Number 1, 2004

nd 2 was similar, 53 months (range, 10.3–167 months) and8.9 months (range, 10.9–85 months), respectively.

The 5-year actuarial cystitis rate for Groups 1 and 2 was4.3% and 11.4%, respectively. At 10 years, it was 24.1% and5% for Groups 1 and 2, respectively (p � 0.134, Fig. 4).

The actuarial rates of pelvic control, overall survival, andectal and bladder complications are summarized in Table 6.

nivariate and multivariate analysesTo examine the influence of confounding factors, 16

atient-, tumor-, and treatment-related variables were testedith univariate and multivariate analysis (Table 7). Theseariables included pathologic type (squamous vs. nonsqua-ous), age (�60 vs. �60 years), body mass index (�24 vs.24), hemoglobin (�12 vs. �12), stage (FIGO Stage IB-

IA vs. IIB-IIIA vs. IIIB-IVA), daily fraction size (1.8 Gys. 2 Gy), box technique (yes vs. no), EBRT dose to theervix (�44 vs. 44–45 vs. �45.1 Gy), EBRT dose to thearametrium (39.6–45 vs. 45.1–54 vs. 54.1–59.6 Gy), dia-etes (yes vs. no), hypertension (yes vs. no), HDR-IC frac-ionation scheme (three fractions vs. five fractions), activeength (�4.5 vs. �4.5 cm), total BED10 (� 88.6 vs. �88.6),nd total treatment days (�70 vs. �70).

For pelvic control, age (p �0.001, hazard ratio � 0.367,5% confidence interval [CI], 0.214–0.629), body massndex (p � 0.020; hazard ratio � 0.541; 95% CI, 0.322–

Table 3. Distribution of patient outcome and treatment-relatedcomplications

Outcome/complicationsGroup 1

(n � 254)Group 2

(n � 287) p*

live 108 (43) 155 (54) 0.008elvic recurrence 30 (12.2) 36 (13) 0.795istant metastasis 54 (21) 56 (20) 0.614roctitis 117 (46) 84 (29) �0.001olostomy 7 (3) 9 (3) 0.982ystitis 45 (18) 26 (9) 0.003hronic abdominal cramp 12 (5) 7 (2) 0.150eg edema 33 (13) 8 (3) �0.001reter obstruction 7 (3) 6 (2) 0.614ouble cancer 9 (4) 10 (4) 0.970

Data in parentheses are percentages; percentages determinedithin group.* Pearson chi-square test.

Table 4. Distribution of pelvic recurrence by group

RecurrenceGroup 1

(n � 254)Group 2

(n � 287)

one 224 (88.2) 251 (87.5)elvic recurrence 30 (11.8) 36 (12.5)Central 24 (9.4) 29 (10.1)Peripheral 5 (2) 7 (2.4)Both 1 (0.4) 0 (0)

Data in parentheses are percentages.p � 0.722.

.909), FIGO stage (p � 0.001; hazard ratio � 1.851; 95%I, 1.306–2.625), and total BED10 (p � 0.003; hazard ratio2.234; 95% CI, 1.311–3.808) were statistically significant

actors.For overall survival, hemoglobin (p �0.001; hazard ratio0.575; 95% CI, 0.450–0.734), FIGO stage (p �0.001;

azard ratio � 1.649; 95% CI, 1.388–1.959), and diabetesp � 0.025; hazard ratio � 1.459; 95% CI, 1.049–2.029)ere statistically significant factors.For radiation proctitis, the EBRT dose to the parametrium

p � 0.007; hazard ratio � 1.291; 95% CI, 1.074–1.533),ypertension (p � 0.020; hazard ratio � 1.502; 95% CI,.065–2.119), and HDR-IC fractionation scheme (p �.004; hazard ratio � 0.807; 95% CI, 0.697–0.934) weretatistically significant factors.

For radiation cystitis, the EBRT dose to the parametriump � 0.045; hazard ratio � 1.350; 95% CI, 1.006–1.811)as the only statistically significant factor.

DISCUSSION

Although the pelvic control rate has been commonlyeported and used to compare different treatment modalitiesn cervical cancer, central recurrence is a more criticalndicator to evaluate the tumoricidal effect of HDR-IC.verall, the central recurrence rate in this series was 10.4%

56 of 541), and its distribution between the two groups wasimilar, 9.4% vs. 10.1% for Groups 1 and 2, respectively (p

0.722). This result showed that the tumoricidal effect ofhe two schemes was comparable in terms of central control.he patterns of recurrence in the pelvis in the HDR-ICetting have not been well analyzed. Wong et al. (3) re-orted 34 (16%) local failures in 205 patients achieving aomplete remission. Of the 34 cases, 7 (21%) were centralailure only, 15 (44%) were pelvic failure only, and 1235%) were mixed failure. Takeshi et al. (13), treating 265tage III patients with HDR-IC, reported an overall pelvicecurrence rate of 24% (63 of 265). Of the 63 cases, 5283%) were central recurrence and 11 (17%) were periph-ral recurrence. Similar to the findings of Takeshi et al. (13),ost pelvic recurrence in our patients was central, account-

ng for 80% (53 of 66) of all recurrence in the pelvis.The tumoricidal effect of HDR-IC for advanced cases

as a concern. Petereit et al. (14), comparing the Wisconsinxperience of HDR vs. LDR, reported significantly poorerelvic control for Stage IIIB cases in HDR-treated patients44% vs. 75% for HDR vs. LDR, respectively). In anothereta-analysis, Petereit and Pearcey (1) observed several

nstitutions reporting favorable pelvic control in Stage IIIBisease with a tumoricidal dose less than the LDR equiva-ent of 75 Gy. Barillot et al. (15) also raised the concern ofnderdosage and/or accurate staging for Stage IIIB casesfter analyzing and comparing 642 LDR-treated patients inhree periods, 1970–1978, 1979–1984, and 1985–1994. The-year local control rate in Stage III patients fell from 75%o 55% when the complication rate decreased simulta-eously with time. In our study, the 5-year actuarial pelvic

cG

a

185Two isoeffect schemes of HDR-IC ● C.-J. WANG et al.

ontrol rate for Stage IIIB-IVA was 68.9% vs. 69.9% for

ppeared slightly more favorable than the findings of Pete-

Fig. 1. Actuarial pelvic control rate by group (p � 0.504). 3-Fr � Group 1; 5-Fr � Group 2.

eit et al. (14) and Barillot et al. (15).

Fig. 2. Actuarial overall survival rate by group (p � 0.734). 3-Fr � Group 1; 5-Fr � Group 2.

acAiaTacTq

iop

fiatwaomHtsoptegsba

wmatoSG0a0

p �0.

NL

H

186 I. J. Radiation Oncology ● Biology ● Physics Volume 59, Number 1, 2004

The parametrial dose was a statistically significant factorssociated with an increased risk of both rectal and bladderomplications in this and one of our previous reports (16).

high parametrial dose has also been reported as a factor ofncreased small bowel complications by Ferrigno et al. (17),lthough it was not statistically significant (p � 0.260).hey suggested the boost field should be lower than S2-S3,nd the dose �54 Gy. As we pointed out earlier, mostervical cancer recurrences in the pelvis remained central.he value of a high dose to the side wall with EBRT is thusuestionable and needs further study.The overall treatment time was not a statistically signif-

cant predictor of local control or survival in our study. Theverall treatment time has been reported as an independentrognostic factor on pelvic control in cervical carcinoma in

Fig. 3. Actuarial proctitis rate by group (

Table 5. Distribution of severity of rectal complication by group

Group 1(n � 254)

Group 2(n � 287)

one 137 (54) 203 (71)ow grade 96 (38) 63 (22)Grade 1 56 (22) 39 (14)Grade 2 40 (16) 24 (8)

igh grade 21 (8) 21 (7)Grade 3 14 (5) 12 (4)Grade 4 7 (3) 9 (3)

Data in parentheses are percentages.p � 0.001.

ve large series between 1992 and 1995 (18–22). Theseuthors reported a loss of pelvic control ranging from 0.5%o 1.2% daily if treatment was prolonged. These resultsere all LDR based. Together with our findings, we noted

t least four HDR series that reported the negative influencef treatment prolongation on pelvic control/survival usingultivariate analysis (3, 13). The latest detailed analysis ofDR series reported by Chen et al.(23) concluded that

reatment prolongation negatively influences cause-specificurvival and pelvic control. To the best of our knowledge,nly one HDR study by Ferrigno et al. (17) has reported aositive influence from treatment prolongation. We believehe issue remains debatable. Additionally, in practice, short-ning the overall treatment time excessively may not beood for advanced cases, because substantial tumor regres-ion may not have been achieved yet for optimal coveragey brachytherapy. This was stressed by Petereit et al. (14)nd the report by the American Brachytherapy Society (24).

The 5-year actuarial overall survival rate for both groupsas similar (p � 0.734). This result was comparable withost of the recent HDR reports. Ferrigno et al. (17) reported5-year overall survival rate of 53.7% in 138 patients

reated with HDR-IC. Takeshi et al. (13) reported a 5-yearverall survival rate of 50.7% in 265 Stage III patients.tage for stage, the 5-year actuarial overall survival rate forroups 1 and 2 was 81.4% vs. 67.5% for Stage IB-IIA (p �.352); 65.8% vs. 54.8% for Stage IIB-IIIA (p � 0.514);nd 38.3% vs. 30.8% for Stage IIIB-IVA, respectively (p �.301). This result is also comparable with existing LDR

001). 3-Fr � Group 1; 5-Fr � Group 2.

rrFS7oaat

iipfmitttWdcd(Pledcl50

sft

P

O

P

C

187Two isoeffect schemes of HDR-IC ● C.-J. WANG et al.

eports. Coia et al. (25) reported a 5-year overall survivalate for Stage I-III in the Pattern of Care Study of 33–74%.yles et al. (26) reported a 5-year overall survival rate fortage I-III in the Princess Margaret Hospital study of 41–8%. Ferrigno et al. (27) reported on the Brazil experiencef cobalt teletherapy plus LDR treatment. The 5-year over-ll survival rate for Stage I-III was 46–83% (27). Barillot etl. (28) reported a result of 45–84% in a French Coopera-ive study.

The proctitis rate in our study was significantly decreased

Fig. 4. Actuarial cystitis rate by group (p

Table 6. Actuarial rate of local control, overall survival, rectaland bladder complications

End pointOverall

(%)Group 1

(%)Group 2

(%) p

elvic control (y)5 87.2 88.2 86.3 0.50410 86.8 87.3 85.5

verall survival (y)5 59.8 63.5 56.1 0.73410 47.9 47.8 49.3

roctitis (y)5 40.8 49.7 32.7 �0.00110 41.7 50.5 32.7

ystitis (y)5 12.8 14.3 11.4 0.13410 21.5 24.1 15.0

n Group 2. We were very pleased to see this result, becauset was the main goal of our initial design. This was the firstrospectively designed protocol in HDR-IC that success-ully demonstrated that a linear-quadratic model–basedathematical calculation can convert into an expected clin-

cal outcome. The success through biologic manipulation ofhe fraction size in our study suggests that the sensitivity ofhe late responding tissue to the fraction change from 7.2 Gyo 4.8 Gy in HDR-IC is high and detectable clinically.

ong et al. (3) and Le Pechoux et al. (2) also compared twoifferent schemes in HDR-IC but failed to demonstrate anyomplication benefit. A possible explanation is that theifference in the fraction size in their studies was too narrow7 Gy vs. 6 Gy in Wong et al. and 6 Gy vs. 5 Gy in Leechoux et al.). The fraction sensitivity may have been too

ow to be detected. Furthermore, the small fraction differ-nce may need a larger sample size or longer follow-up toetect any differences. Wong et al. (3) reported an overallomplication rate for the four-fraction regimen that wasower than that of the three-fraction regimen (44.2% vs.5.8%), with a p value of marginal significance (p �.0672).Our finding has important implications. The fractionation

cheme in HDR-IC has varied widely from 2 to 9 Gy/raction in the past. With rare exception (29), we observedhat the fractionation schemes used in recent studies have

34). 3-Fr � Group 1; 5-Fr � Group 2.

mtScrrtso

tod(h(to((idpadcadvetrnb

aiA7rwsaspttubwtGto

wcrqTtdtTsc

PABHSDDDBDHAHTT

r

188 I. J. Radiation Oncology ● Biology ● Physics Volume 59, Number 1, 2004

ostly converged between 4 and 7 Gy, which agrees withhe recent recommendation of the American Brachytherapyociety (24) and the Wisconsin experience (1, 14). Clini-ally, it is not appropriate to interpolate or extrapolate theesults of this study. The fraction sensitivity in the doseange of brachytherapy interest may not be totally linear, sohat the clinical results of a dose between 4.8 and 7.2 Gyhould not be extrapolated to, for instance, 4.8 Gy and 3 Gy,r 7.2 Gy and 10 Gy.Although the overall proctitis rate decreased, it appeared

hat the manipulation had its limitations. When the severityf proctitis was compared grade by grade, we found that theecrease was confined to only the low-grade complications38% vs. 22% for Groups 1 and 2, respectively). Theigh-grade complication rate remained virtually unchanged8% vs. 7% for Groups 1 and 2, respectively). Hospitaliza-ion or blood transfusion because of severe rectal bleedingr anemia (Grade 3) was needed in 5% (14 cases) and 4%12 cases) of Groups 1 and 2, respectively. ColostomyGrade 4) was required in 3% of patients, distributed equallyn both groups. Barillot et al. (15) reported the effect ofifferent techniques of radiotherapy in the treatment of 642atients with cervical cancer in three periods between 1970nd 1994. The rate of Grade 3 rectal/bladder complicationsecreased over time from 16% to 6%, and no Grade 4omplications was encountered in the last period. Theyttributed these complications to three factors: a high EBRTose, a high dose rate at the reference points, and wholeagina brachytherapy with cylinder. The analysis of Barillott al. (15) highlighted that severe complications in radio-herapy for cervical cancer are more technically or dosimet-ically related. It is thus understandable that biologic ma-ipulation of the fraction size in our study may not haveeen able to change the rate of severe complications. Radi-

Table 7. Univariate

Variable

Pelvic control

UVA MVA

athologic finding 0.042 0.064ge �0.001 �0.001*MI 0.014 0.020*emoglobin 0.003 0.204 �tage �0.001 0.001* �aily dose (EBRT) 0.296 0.057ose to cervix (EBRT) 0.003 0.248ose to parametrium (EBRT) 0.002 0.806 �ox field 0.823 0.390iabetes 0.132 0.303ypertension 0.227 0.699ctive length 0.453 0.630DR-IC scheme 0.504 0.172otal BED10 0.040 0.003*otal treatment days 0.432 0.357

Abbreviations: UVA � univariate analysis; MVA � multivadiotherapy; BED10 � biologically equivalent dose, assuming �/

* Statistically significant in MVA.

tion-induced rectal bleeding is a complicated issue, involv-ng damage of both acute and late responding tissue (30).lthough the rate of Grade 3-4 severe rectal complications,–8%, in this study was acceptable, we believe there is stilloom to improve, particularly in Stage I-II cases. We agreeith Petereit and Pearcey’s conclusion in their meta-analy-

is that the optimal fractionation schedule is still unknownnd presently can be based only on single institutions withignificant experience (1). The treatment policy in our de-artment will continue to evolve, including decreasing theotal EBRT dose to the whole pelvis, restricting the parame-rial boost dose/field in Stage I-II cases, and increasing these of concurrent chemotherapy in selective cases on theasis of recent encouraging results (31–33). Additionally,e noted that the fractionation schemes of HDR-IC have

ended to converge further between the dose of 4.5 and 6y, the fraction sensitivity of this range deserves clinical

esting. A grant-supported randomized trial is underway inur department.

CONCLUSION

On the basis of the experiences of treating 541 patientsith two isoeffect schemes, we found that the central re-

urrence, actuarial pelvic control rate, and overall survivalate were similar and the proctitis rate decreased. The linear-uadratic model correctly predicted this clinical outcome.his study demonstrated that fraction sensitivity between

he dose of 4.8 Gy and 7.2 Gy in HDR-IC is high andetectable clinically. This finding has important implica-ions for the prescription of the fraction dose in HDR-IC.he success through biologic manipulation of the fractionize in this range has limitations, and the improvement wasonfined to a reduction in low-grade complications.

ultivariate analysis

rall survival Proctitis rate Cystitis rate

MVA UVA MVA UVA MVA

0.066 0.949 0.982 0.283 0.3560.218 0.639 0.460 0.242 0.2340.110 0.174 0.153 0.466 0.489

�0.001* 0.202 0.326 0.566 0.446�0.001* 0.001 0.881 0.048 0.457

0.767 0.008 0.318 0.399 0.7350.807 0.026 0.194 0.060 0.3250.862 0.003 0.007* 0.027 0.045*0.907 0.040 0.804 0.508 0.4750.025* 0.951 0.598 0.806 0.6970.317 0.247 0.020 0.303 0.6090.906 0.327 0.325 0.378 0.6030.304 0.001 0.004* 0.134 0.2970.217 0.463 0.256 0.043 0.2260.747 0.011 0.864 0.190 0.891

analysis; BMI � body mass index; EBRT � external beam0; HDR-IC � high-dose-rate intracavitary brachytherapy.

and m

Ove

UVA

0.0290.8290.0110.0010.0010.2870.0180.0010.3130.1430.6230.9240.7430.1450.077

ariate� � 1

1

1

1

1

1

1

1

1

189Two isoeffect schemes of HDR-IC ● C.-J. WANG et al.

REFERENCES

1

1

2

2

2

2

2

2

2

2

2

2

3

3

3

3

1. Petereit DG, Pearcey R. Literature analysis of high dose ratebrachytherapy fractionation schedules in the treatment of cer-vical cancer: Is there an optimal fractionation schedule? Int JRadiat Oncol Biol Phys 1999;43:359–366.

2. Le Pechoux C, Akine Y, Sumi M, et al. High dose ratebrachytherapy for carcinoma of the uterine cervix: Compari-son of two different fractionation regimens. Int J Radiat OncolBiol Phys 1995;31:735–741.

3. Wong FCS, Tung SY, Leung T-W, et al. Treatment results ofhigh-dose-rate remote afterloading brachytherapy for cervicalcancer and retrospective comparison of two regimens. Int JRadiat Oncol Biol Phys 2003;55:1254–1264.

4. Ling CC, Sahoo N, Leibel S, et al. High dose rate gynecolog-ical applications: Radiobiological considerations based on the�/� model. Radiother Oncol 1992;25:103–110.

5. Brenner DJ, Hall EJ. Fractionated high dose rate vs. low doserate regimens for intracavitary brachytherapy of the cervix.Br J Radiol 1991;64:133–141.

6. Fowler JF. The linear-quadratic formula and progress in frac-tionated radiotherapy. Br J Radiol 1989;62:679–694.

7. Dale RG. The use of small fraction numbers in high dose rategynecological afterloading: Some radiobiological consider-ations. Br J Radiol 1990;63:290–294.

8. Sood B, Garg M, Avadhani J, et al. Predictive value oflinear-quadratic model in the treatment of cervical cancerusing high dose rate brachytherapy. Int J Radiat Oncol BiolPhys 2002;54:1377–1387.

9. Orton CG. High dose rate vs. low dose rate brachytherapy forgynecological cancer. Semin Radiat Oncol 1993;3:232–239.

0. Wang C-J, Leung SW, Chen H-C, et al. High-dose-rate int-racavitary brachytherapy (HDR-IC) in treatment of cervicalcarcinoma: 5-year results and implication of increased low-grade rectal complication on initiation an HDR-IC fraction-ation scheme. Int J Radiat Oncol Biol Phys 1997;38:391–398.

1. Chen M-S, Lin F-J, Hong C-H, et al. High-dose-rate after-loading technique in the radiation treatment of uterine cervicalcancer: 399 cases and 9 years experience in Taiwan. Int JRadiat Oncol Biol Phys 1991;20:915–919.

2. Cox JD, Stetz J, Pajak TF. Toxicity criteria of the RadiationTherapy Oncology Group (RTOG) and the European Organi-zation for Research and Treatment of Cancer (EORTC). Int JRadiat Oncol Biol Phys 1995;31:1341–1346.

3. Takeshi K, Katsuyuki K, Yoshiaki T, et al. Definitive radio-therapy combined with high dose rate brachytherapy for StageIII carcinoma of the uterine cervix: Retrospective analysis ofprognostic factors concerning patients characteristics andtreatment parameters. Int J Radiat Oncol Biol Phys 1998;41:319–327.

4. Petereit DG, Sarkaria JN, Potter DM, et al. High-dose-rate vs.low-dose-rate brachytherapy in the treatment of cervical can-cer: Analysis of tumor recurrence—The University of Wis-consin experience. Int J Radiat Oncol Biol Phys 1999;45:1267–1274.

5. Barillot I, Horiot JC, Maingo P, et al. Impact on treatmentoutcome and late effects of customized treatment planning incervix carcinoma: Baseline results to compare new strategies.Int J Radiat Oncol Biol Phys 2000;48:189–200.

6. Huang EY, Lin H, Hsu HC, et al. High external parametrialdose can increase the probability of radiation proctitis inpatients with uterine cervix cancer. Gynecol Oncol 2000;79:406–410.

7. Ferrigno R, Novaes PERDS, Pellizzon ACA, et al. High-dose-

rate brachytherapy in the treatment of uterine cervix cancer:Analysis of dose effectiveness and late complication. Int JRadiat Oncol Biol Phys 2001;50:1123–1135.

8. Fyles A, Keane TJ, Barton M, et al. The effect of treatmentduration in the local control of cervix cancer. Radiother Oncol1992;25:273–279.

9. Petereit DG, Sakaria JN, Chappell R, et al. The adverse effectof treatment prolongation in cervical carcinoma. Int J RadiatOncol Biol Phys 1995;32:1301–1307.

0. Lanciano RM, Pajak TF, Martz K, et al. The influence oftreatment time on outcome for squamous cell cancer of theuterine cervix. Cancer 1992;69:2124–2130.

1. Girinsky T, Rey A, Roche B, et al. Overall treatment time inadvanced cervical carcinoma: A critical parameter in treat-ment outcome. Int J Radiat Oncol Biol Phys 1993;27:1051–1056.

2. Perez CA, Grigsby PW, Castro-Vita H, et al. Carcinoma of theuterine cervix: Impact of prolongation of treatment time andtiming of brachytherapy on outcome of radiation therapy. IntJ Radiat Oncol Biol Phys 1995;32:1275–1288.

3. Chen SW, Liang JA, Yang SN, et al. The adverse effect oftreatment prolongation in cervical cancer by high-dose-rateintracavitary brachytherapy. Radiother Oncol 2003;67:69–76.

4. Nag S, Erickson B, Thomadsen B, et al. The AmericanBrachytherapy Society recommendations for high dose ratebrachytherapy for carcinoma of the cervix. Int J Radiat OncolBiol Phys 2000;48:201–211.

5. Coia L, Won M, Lanciano R, et al. The patterns of careoutcome study for cancer of the uterine cervix: Results of thesecond National Practice Survey. Cancer 1990;9:431–436.

6. Fyles AW, Pintilie M, Kirkbridge P, et al. Prognostic factorsin patients with cervix cancer treated by radiation therapy:Results of a multiple regression analysis. Radiother Oncol1995;35:107–117.

7. Ferrigno R, Oliveira Faria SLC, Weltman E, et al. Radiother-apy alone in the treatment of uterine cervix cancer withtelecobalt and low dose rate brachytherapy: Retrospectiveanalysis of results and variables. Int J Radiat Oncol Biol Phys2003;55:695–706.

8. Barillot I, Horiot JC, Pigneaux J, et al. Carcinoma of the intactuterine cervix treated with radiotherapy alone: A French co-operative study—Update and multivariate analysis of prog-nostic factors. Int J Radiat Oncol Biol Phys 1997;38:969–978.

9. Sood BM, Gorla G, Gupta S, et al. Two fractions of high-dose-rate brachytherapy in the management of cervix cancer:Clinical experience with and without chemotherapy. Int JRadiat Oncol Biol Phys 2002;53:702–706.

0. Dubray BM, Thames HD. Chronic radiation damage in therate rectum: An analysis of the influences of fractionation,time, and volume. Radiother Oncol 1994;33:41–47.

1. Morris M, Eifel P, Lu J, et al. Pelvic radiation with concurrentchemotherapy compared with pelvic and para-aortic radiationfor high-risk cervical cancer. N Engl J Med 1999;340:1137–1143.

2. Rose PG, Bundy N, Watkins EB, et al. Concurrent cisplatin-based radiotherapy and chemotherapy for locally advancedcervical cancer. N Engl J Med 1999;340:1144–1153.

3. Green JA, Kirwan JM, Tierney JF, et al. Survival and recur-rence after concomitant chemotherapy and radiotherapy forcancer of the uterine cervix: A systematic review and meta-analysis. Lancet 2001;358:781–786.