Journal of Neuro-Oncology 57: 241–245, 2002.© 2002 Kluwer Academic Publishers. Printed in the Netherlands.

Clinical Study

Imaging response to chemotherapy with RMP-7 and carboplatin inmalignant glioma: size matters but speed does not

Robert Grant1, Mark Walker1, Donald Hadley2, Tina Barton3 and Chester Osborn3

1Edinburgh Centre for Neuro-Oncology, Western General Hospital, Edinburgh, Scotland, UK;2Department of Neuro-radiology, Institute of Neurological Sciences, Glasgow, Scotland, UK;3Alkermes Europe Ltd., Cambridge, England, UK

Key words: malignant glioma, clinical outcome, tumor volume, MRI scan

Summary

Introduction In recurrent malignant glioma, early imaging response to two courses of chemotherapy is generallyconsidered to be predictive of good survival. We studied the relationships between initial tumour volume and speedof imaging response to chemotherapy in malignant glioma.

Methods In 43 chemotherapy naı̈ve patients, 26 glioblastoma multiforme (GBM) + 17 anaplastic astrocytoma(AA), median age 45 years, MRI responses to intravenous cereport and carboplatin were assessed at baseline, thenat 2 monthly intervals. Patients were classified as fast responders if they had reached a partial response (PR) aftertwo courses, and slow responders if PR was achieved after three or more courses of chemotherapy.

Results PR occurred in four patients with GBM (15%) and nine patients with AA (53%). Likelihood of responsewas related to initial tumour enhancing volume in GBM but not in AA. PR occurred in four of five GBM patients(80%) with initial volume <15,000 mm3 and none of the 21 cases with an initial volume >15,000 mm3. In patientsachieving a PR, there was no association between speed or duration of response and eventual survival. Fast responderswith AA were significantly older than slow responders (p = 0.033).

Conclusions Initial enhancing volume of GBMs may be an important predictor of imaging response. This hasimplications where response rates of phase II studies are reported and in stratification for phase III trials. Furtherwork is necessary to confirm these findings with other types of chemotherapy and examine the relationship betweenproliferation markers and speed of response.

Introduction

Assessment of response to chemotherapy in phase IIstudies of malignant glioma depends on comparingclinical, imaging and steroid usage between two timepoints based on criteria suggested by Macdonald et al.[1]. The Macdonald criteria have been crucially impor-tant in standardizing what is meant by a response tochemotherapy. The criteria are based on (a) clinicalstate (better, stable, and worse), (b) imaging (completeresponse (CR), partial response (PR), stable disease(SD), and progressive disease (PD)), and (c) steroiddosage (increased, stable, reduced, and no steroids).Change in clinical state can be objectively standard-ised using the Edinburgh Functional Impairment Tests(EFIT) [2–4]. Change in imaging can be objectively

measured by change in size of the tumour. Change insteroid dose has not been standardised as there is noagreement in what represents a significant change insteroid dosage.

Despite the above criteria there is controversy overhow to measure tumour size (diameter, area, andvolume), whether the initial size is important and howsustained a change has to be before it is acceptedas significant (e.g. two ‘courses’, ‘two images’, and‘x weeks’). Methodology for measuring change andthe way the results are analysed are often not welldescribed in studies. In recurrent malignant glioma,early imaging response to two courses of chemotherapyis generally considered to be predictive of good survivalwhereas SD following two courses of chemotherapyis often considered less good. Nevertheless, we have

242

not found a difference in time to progression/survivalbetween patients who achieved a PR after two coursesof chemotherapy and those who achieved a minimalresponse (25–50% reduction in tumour size) or sta-ble disease after two courses of chemotherapy [5].We have previously hypothesised that there are slowresponders and that these patients, who may only showa PR after three or more courses of chemotherapy,have no worse prognosis than fast responders [5].

To evaluate the relationships between initialtumour volume and speed of imaging response tochemotherapy in patients with malignant glioma, weanalysed serial standardised gadolinium enhancedvolumetric MRI scans from a multi-centre phase IIstudy of RMP-7 and carboplatin used in malignantglioma at first relapse [4].

Methods

We analysed the serial imaging response from an openlabel, single arm phase II study of RMP7 and carbo-platin. Patients were recruited from centres in France,Sweden, and the United Kingdom. The inclusion andexclusion criteria for the study are shown in Table 1.

Table 1. List of inclusion and exclusion criteria

Inclusion criteria

1. Age ≥18 years.

2. Histologically proven case of AA or GBM.

3. Patient had previously undergone a surgical procedure (eitherbiopsy or resection) and a radical course of external beamradiotherapy.

4. Patient must have experienced a radiologically provenrecurrence (defined as a documented radiological increase intumour size following a progression free interval of at least 3months after radiotherapy).

5. Karnofsky performance score (KPS) ≥ 60.

6. Life expectancy ≥ 8 weeks.

7. Presence of an enhancing, measurable lesion on MRI.

8. Patient must be chemotherapeutically naive.

Exclusion criteria

1. The presence of biochemical and/or haematologicalparameters outside defined levels of clinical significance.

2. The patient required further cyto-reductive surgery.

3. The presence of other clinically significant disease.

4. Pregnancy, lactation or inability to use suitable contraceptiveprecautions.

Appropriate regulatory and ethical approval wasobtained and all patients enrolled to the study wererequired to provide written consent. All histology wassubject to central pathology review.

Treatment was provided in 4-week cycles for a max-imum of 12 cycles. Carboplatin was administered as a15-min IV infusion. The appropriate dose was calcu-lated using the Calvert formula and subsequent doseswere increased or decreased as appropriate. A 10-minIV infusion of 300 ng/kg of RMP-7 was started 10 minafter the commencement of the carboplatin infusion.

Clinical evaluation of treatment effect consisted ofpatient assessment using the EFIT, MRC NeurologyStatus Score and the KPS. Clinical evaluations wererecorded on the day of dosing of each treatment cycle.MRI assessments were performed at baseline and at theend of cycles 2, 4, 6, 9, and 12. Patients were assessedat baseline and at 2-monthly intervals thereafter. Com-puter software was used to calculate the volume ofGadolinium enhancing tumour based on aggregates oftumour slice area and thickness. Enhancing tumour vol-umes were manually traced on each digital image usinga computer program (Evergreen Technologies Inc. Cas-tine, ME), which summed the slice volumes to give atotal volume of the enhancement for the whole tumour.An independent neuro-radiologist, blinded to patientdetails, clinical data and order of assessment, mea-sured the tumour volume in mm3. Imaging responsewas assessed by MRI using Macdonald criteria.

Definitions

‘Fast responders’ were defined as patients who hadreached a ‘partial response’ after two courses.

‘Slow responders’ were defined as patients who wereclassified as having ‘stable disease’ after two coursesbut subsequently reached a ‘partial response’.

Best response was defined as the smallest enhancingtumour volume as a percentage of the baseline scan(maximum % reduction).

Best course was defined as the number of courses ofchemotherapy received before achieving best response.

Survival was calculated from the first day ofchemotherapy to death.

Results

There were 26 patients with GBM and 17 patients withAA. Ages ranged from 24 to 67 years (median age

243

45 years) (Table 2). On Central Pathology Review, a39-year-old female had the diagnosis changed fromAA to anaplastic oligodendroglioma and a 42-year-old female had the diagnosis change AA to anaplasticoligoastrocytoma (see Table 2)

Age versus initial tumour volume

There was no significant association between age andinitial tumour volume at the start of treatment forglioblastoma, (r = 0.058 (95% CI for r, −0.350 to0.467), p-value = 0.772) or AA (r = 0.159 (95% CIfor r, −0.348 to 0.594), p-value = 0.543) either singlyor combined, r = 0.120 (95% CI for r , −0.187 to0.406), p-value = 0.437).

Age versus partial response

Likelihood of achieving a partial response is stronglyrelated to age (<= 45 years) in glioblastoma (FishersExact Test p-value = 0.019), but not in AA (FishersExact Test p-value = 0.619)

Tumour grade versus partial response

Partial response occurred in 4 of 26 patients withglioblastoma (15%, median age 38 years) and 9 of17 patients with anaplatic astrocytoma (53%, medianage 35 years). The proportion of patients with AAundergoing a partial response was significantly greaterthan those with glioblastoma (chi-square test, df = 1,p-value = 0.009).

Initial tumour enhancing volume versuspartial response

(a) Glioblastoma multiforme. Four patients with GBMhad a partial response (median volume 4594 mm3)whereas 22 patients with GBM did not achieve a par-tial response (median volume = 36,776 mm3) (MannWhitney Test, p-value = 0.003) (Table 3). Fourof five patients with GBM volume <=15,000 mm3

achieved a partial response versus 0 out of 21 of thosewith volume >15,000 mm3 (Fishers Exact Test, p-value = 0.001). Initial tumour volume was, therefore,an important factor in determining whether a PR waslikely with RMP-7 and carboplatin chemotherapy.

(b) Anaplastic astrocytoma. Nine patients with AAhad a partial response (median volume = 10,014 mm3)and eight patients did not have a partial response

Table 2. Best volumetric response depending on grade, age,and volume

Age Sex Initial Best volume Survival(years) volume from baseline (days)

(mm3) (%)

Anaplastic astrocytoma26 F 8485 −95.0 61528 M 45,481 +144 10031 F 29,820 −62.9 37331 M 61,480 +38.9 39632 M 3540 −76.8 47232 M 6675 −26.4 21433 M 28,046 Died before

scanned49

34 M 1313 −64.8 32135 F 15,990 −81.8 923 Alive39 F 77,375 −41.7 24242 F 23,885 −34.5 64146 F 69,439 −74.2 28750 F 40,602 −60.0 54252 M 10,014 −97.5 17953 M 109,302 Died before

scanned28

55 F 16,292 +68.6 13759 M 2103 −100 825 Alive

Glioblastoma multiforme24 M 63,247 +135.0 28826 F 32,511 −11.3 59729 M 38,546 −6.2 44836 M 5186 −71.5 650 Alive36 F 41,499 +29.4 19536 F 59,915 −4.5 24438 F 4001 −82.0 66638 M 14,806 −79.1 33840 F 4508 −39.1 573 Alive44 F 1548 −97.4 43145 F 116,433 Died before

scanned92

46 M 27,634 +48.7 10848 M 33,709 Died before

scanned108

50 M 61,433 30.9 13452 M 44,270 +17.8 12553 M 21,153 +75.5 15953 M 35,007 −1.85 14554 F 72,733 Died before

scanned41

56 F 28,891 −46.2 41257 M 51,358 Died before

scanned52

58 M 61,704 −28.5 9558 F 83,518 +13.9 10659 M 15,481 +44.4 11959 M 24,549 +361.5 11466 M 30,982 −34.6 43567 M 27,510 +17.2 67

(−) denotes a reduction from baseline tumor volume.

244

Table 3. Partial responders divided by grade, age, and speed of response

Tumour Age Tumour Response Best Best course SurvivalType (years) Volume (mm3) After 2 courses response (%) (days)

AA 26 8485 SD −95 After 4 courses 615AA 31 29,820 SD −63 After 4 courses 373AA 32 3540 SD −77 After 4 courses 472AA 34 1313 PR −64 After 6 courses 321AA 35 15,990 SD −82 After 8 courses 923 aliveAA 46 69,438 SD −74 After 6 courses 287AA 50 40,602 PR −60 After 4 courses 542AA 52 10,014 PR −97 After 3 courses 179AA 59 2102 PR −100 After 9 courses 825 aliveGBM 36 5186 SD −71 After 4 courses 650 aliveGBM 38 4001 SD −82 After 4 courses 666GBM 38 14,806 PR −79 After 2 courses 338GBM 44 1548 PR −97 After 2 courses 435

(median volume = 36,763 mm3) (Mann Whitney Test,p-value = 0.075). Five of six patients with AAand initial tumour enhancing volume <=15,000 mm3

achieved a partial response versus four out of eleven ofthose with volume >15,000 mm3. (Fishers Exact Test,p-value = 0.086). Initial tumour volume was not a par-ticularly important factor in determining whether a PRwas likely with RMP-7 and carboplatin.

Speed of response versus best response(maximum % reduction) in partial responders

There was no clear relationship between speed ofresponse and best response amongst patients undergo-ing partial response who had a GBM or a AA (Pearsoncorrelation, r = −0.041 (95% CI for r , −0.579 to0.5218). p-value = 0.893).

Speed of response versus age inpartial responders

The median age of fast responders (n = 4) with AAwas greater than that of slow responders (n = 5) (fastresponders, median age = 51 years; slow responders,median age = 32 years, Mann Whitney test for smallsamples (one tailed), p-value = 0.033). The numbersof patients with GBM who had a response (fast or slow)was too small to calculate meaningful statistics (n = 4).

Speed of response versus prognosis

Slow responders had at least as good a prognosis asfast responders (AA and glioblastoma combined). Slowresponders (n = 7), had a median survival 615 days,

(IQR 373–666) while fast responders (n = 6) hada median survival of 338 days, (IQR 321–542 days).Log-rank test for difference in survival (chi-square =0.8841, df = 1, p-value = 0.3748) indicated no sig-nificant difference in survival between fast and slowresponders. Wilcoxon test for difference in survival(chi-square = 1.3617, df = 1, p-value = 0.2432) alsoindicated no difference in survival between fast andslow responders.

Conclusions

This study has identified two potentially importantfacts about measuring tumour volume in patients withmalignant glioma. Firstly, initial tumour enhancingvolume is important in glioblastoma and may be impor-tant in AA (but not statistically so in this study). Theimplications are; firstly, if a phase II trial accepts smallvolume tumours only, they may find a much betterresponse rate than another study using the same drugwhere there is no limitation on tumour size; secondly,should regular scanning be performed to identifytumour progression at an early stage when chemother-apy may increase the chance of a partial response?,and thirdly, it raises the question of whether surgicaldebulking should precede chemotherapy at relapse ifthe tumour is above a certain size? Our findings needto be confirmed on a different data set (e.g. nitrosoureaor temozolomide), but we suggest that future stud-ies should report initial tumour size and analyse andpresent results accordingly. The latter two questionsare important health economic and management issues(regular scanning versus scanning on clinical relapse)(debulking and chemotherapy versus chemotherapy

245

alone) and could be subjected to randomised controlledtrials, although the more important issues relate towhether either of these policies improves quality of lifeor survival, rather than their effect on imaging response.

Secondly, we have found that speed of response(in partial responders) bares no relationship to progno-sis/eventual survival. We found about 50% of patientswith glioblastoma or AA are fast responders and 50%are slow responders and that fast responders are olderthan slow responders. Fast responders do no better thanslow responders, and may be destined to do worse.These findings help explain our previous problems inidentifying a survival difference between patients whoachieved a partial response after two courses of nitro-surea based chemotherapy, and those who achievedonly a minimal response or stable disease [5]. Thisraises another methodological issue. If there are fastresponders that are also fast relapsers, the more fre-quent the imaging, the greater the chance of identifyingfast responders and the higher the likelihood of hav-ing a sustained response, i.e. the greater the numberof partial responders. Response usually has to be con-firmed on a further scan after a period of time [1] andif scanning is done frequently (e.g. every two or fourweeks [6,7], there are likely to be more responses iden-tified and more cases defined as a sustained responsethan if repeat scans are done after only ‘two courses’ ofchemotherapy (e.g. from eight to ten weeks) [8]. Themore frequent the imaging the higher the likelihood ofhaving a sustained response, i.e. the greater the numberof partial responders. As scanning has become moreavailable and frequency of scanning in recent trials hasincreased, it is not legitimate to compare the percent-age of partial responses in recent phase II trials withhistorical phase II trials.

It may be more logical to record the bestresponse (maximum % reduction) for each case givenchemotherapy and duration of treatment before relapse(radiological or clinical). Speed of response may be try-ing to tell us something about the cellular mechanismsof response (or chemo-resistance in the fast relapsers)and correlative studies with proliferation indices wouldbe interesting and informative.

Dynamics of tumour response to chemotherapyrequires further study. We found age and tumourgrade were important factors in predicting a partialresponse as has been noted many times previously [7],but in addition suggest initial tumour volume is alsoimportant. If the findings that initial volume of contrast

enhancing tumour is predictive of response and speedof response is not predictive for duration of response[5] are replicated in other data sets, changes in howtrials are designed and reported may be necessary.

Acknowledgements

We would like to thank Alkermes Europe Ltd. forallowing us access to their outcome results from theirstudy and all those investigators involved in the study.

References

1. Macdonald DR, Cascino TL, Schold SC Jr: Response crite-ria for phase II studies of supratentorial malignant glioma.J Clin Oncol 8: 1277–1280, 1990

2. Grant R, Slattery J, Gregor A, Whittle IR: Recording neu-rological impairment in clinical trials of glioma. J Neuro-Oncol 19: 37–49, 1994

3. Clyde Z, Chataway SJ, Signorini D, Gregor A, Grant R:Significant change in tests of neurological impairment inpatients with brain tumours. J Neuro-Oncol 39: 81–90, 1998

4. Gregor A, Lind M, Newman H, Grant R, Hadley DM, BartonT, Osborn C: Phase II studies of RMP-7 and carboplatin inthe treatment of recurrent high-grade glioma. J Neuro-Oncol44: 137–145, 1999

5. Grant R, Liang BC, Slattery J, Greenberg HS, Junck L:Chemotherapy response criteria in malignant glioma.Neurology 48(5): 1336–1340, 1997

6. Oldfield EH, Ram Z: Protocol guide of patient referrals:gene therapy for the treatment of brain tumours usingintra-tumoral transduction with thymidine kinase gene andintravenous ganciclovir 92-N-0246. Hum Gene Ther 4(1):60–69, 1993

7. Yung WKA, Prados MD, Yaya-Tur R, Rosenfeld SS,Brada M, Friedman HS, Albright R, Olson J, Chang SM,O’Neill AM, Friedman AH, Bruner J, Yue N, Dugan M, Zak-noen S, Levin VA: Multicentre phase II trial of temozolo-mide in patients with anaplastic astrocytoma or anaplasticoligoastrocytoma at first relapse. J Clin Oncol 17: 2762–2771, 1999

8. Gregor A, Rampling R, Aapro M, Malnstrom P, Whit-tle IR, Rye R, Stewart M, Sellar R, Demierre B, IronsideJW, Wahlby S, Smyth JF: Phase II study of Tauromus-tine in malignant glioma. Eur J Cancer 28(12): 1959–1962,1992

Address for offprints: Robert Grant, Edinburgh Centre for Neuro-Oncology, Western General Hospital, Crewe Road, EdinburghEH4 2XU, UK; Tel.: +44 131 537 3266; Fax: +44 131 537 1132;E-mail: Rgrant@skull.dcn.ed.ac.uk