Short Delay in Initiation of Radiotherapy May Not Affect Outcome of Patients With Glioblastoma: A Secondary Analysis From the Radiation Therapy Oncology Group Database

Abstract

Purpose

To analyze the Radiation Therapy Oncology Group (RTOG) database of patients with glioblastoma and appraise whether outcome was influenced by time to initiation of radiation therapy (RT).

Patients and Methods

From 1974 through 2003, adult patients with histologically confirmed supratentorial glioblastoma were enrolled onto 16 RTOG studies. Of 3,052 enrolled patients, 197 patients (6%) were either initially rendered ineligible or had insufficient chronologic data, leaving a cohort of 2,855 patients for the present analysis. We selected four patient groups based on the interval from surgery to the start of RT: ≤ 2 weeks, 2 to 3 weeks, 3 to 4 weeks, more than 4 weeks to the protocol eligibility limit of 6 weeks. Survival times were estimated by the Kaplan-Meier method. Multivariate analysis incorporated variables of time interval, recursive partitioning analysis (RPA) class, and treatment regimen.

Results

No decrement in survival could be identified with increasing time to initiation of RT. Among our four temporal groupings, median survival time was unexpectedly and significantly greater in the group with the longest interval (> 4 weeks) than in those with the shortest delay (≤ 2 weeks): respectively, 12.5 months versus 9.2 months (P < .0001). On multivariate analysis, with overall survival as the end point, time interval more than 4 weeks and lower RPA class were both significant predictors of improved outcome. Treatment regimen was not a significant factor.

Conclusion

There is no evident reduction in survival by delaying initiation of RT within the relatively narrow constraint of 6 weeks. An unanticipated yet significantly superior outcome was identified for patients for whom RT was delayed beyond 4 weeks from surgery.

INTRODUCTION

Randomized trials have consistently shown the value of radiotherapy as part of the optimal management of glioblastoma multiforme (GBM).¹ Even modern series continue to show a statistically significant advantage in overall survival when radiotherapy alone is compared with best supportive care in various populations.² With its emergence as a standard of care for GBM, more attention has been devoted toward optimizing the delivery of radiotherapy.^3–5

A generally recognized, straightforward means of maximizing the efficacy of cancer treatments is the prompt initiation of such therapies. Indeed, among the first principles of oncology is the expeditious inauguration of cytotoxic therapy.⁶ Because delay would be expected to have the most detrimental effect on the control of neoplasms with short doubling times,⁷ patients with rapidly growing tumors such as GBM are theoretically the most vulnerable to negative consequences from delayed initiation of radiotherapy.

The relationship between the delay in radiotherapy and the outcome of radiotherapy has been explored^8–13 in several tumor types (eg, breast, head and neck cancer) and less extensively in others (eg, lung, cervix cancer).^14,15 To date, only one single-institutional experience has specifically addressed the delayed initiation of irradiation for GBM.¹⁶ The current study was undertaken to explore this relationship by analyzing the database of the Radiation Therapy Oncology Group (RTOG), which provides prolonged follow-up from patients treated at multiple centers throughout the United States and Canada.

PATIENTS AND METHODS

Patient Population

Patients entered onto RTOG trials for biopsy-proven GBM constitute the study group for this article. These trials accrued 3,052 patients between 1974 and 2003. Primary treatment outcome reports from these trials have been previously published.^{3–5,17–21} Eligibility criteria were consistent in all of the studies: histologically confirmed supratentorial GBM; age of at least 18 years; normal hepatic, renal, and bone marrow function; and an interval of 6 weeks or less from surgery to initiation of radiotherapy. Ineligibility criteria included prior malignancies (except skin carcinomas), prior chemotherapy, or head and neck irradiation.

Protocol Summaries

The data for the present secondary analysis were culled from 16 studies. The treatment regimens of the trials included in this analysis are concisely described in Table 1. Additional information is included on the official Web site of the RTOG (www.rtog.org).

Table 1.

RTOG Study Details

Study No.	Phase	Treatments	Total No. of Patients	No. of Patients With GBM
RTOG 7401/	III	1. 60 Gy whole-brain RT	639	449
ECOG 1374		2. 60 Gy whole-brain RT plus a 10 Gy RT boost dose
		3. 60 Gy whole-brain RT plus BCNU 80 mg/m2/d × 1 every 8 weeks
		4. 60 Gy whole-brain RT plus MeCCNU (125 mg/m2/d × 1 every 8 weeks) and DTIC (150 mg/m2/d × 5 every 4 weeks)
RTOG 7918	III	1. 60 Gy whole-brain RT and BCNU	318	247
		2. 60 Gy whole-brain RT and BCNU with radiosensitizer misonidazole at a dose of 2.5 mg/m2 before RT each Monday
		BCNU 80 mg/m2 IV days 3, 4, and 5 of first week of radiotherapy, then BCNU 80 mg/m2 IV × 3 days every 8 weeks beginning day 64
RTOG 8302	I/II	A dose-escalation trial of hyperfractionated partial-brain RT and accelerated hyperfractionated partial-brain RT with carmustine. Four RT dose levels of hyperfractionated partial brain RT were studied in 1.2 Gy twice-daily fractionation with an interfraction interval of 4to 8 hours. These dose levels were 64.8, 72.0, 76.8, and 81.6 Gy. The final portion of the study was a randomization between the total accelerated hyperfractionated partial-brain RT dose of 48.0 and 54.4 Gy in 1.6 Gy twice-daily fractionation with the same interfraction interval requirements	786	570
RTOG 8409	I/II	Combined conventional doses of RT with the quinone AZQ (15 mg/m2 once weekly for 4 weeks)	54	46
RTOG 9006	III	1. Conventional RT plus BCNU	712	534
		2. Hyperfractionated RT (72 Gy in 1.2 Gy fractions administered twice daily) plus BCNU
		BCNU 80 mg/m2 IV days 1, 2, and 3 of RT then every 8 weeks for a total of 6 cycles
RTOG 9305	III	1. Conventional RT plus BCNU	203	203
		2. Conventional RT plus BCNU with upfront radiosurgical boost (tumor size-dependent dosing raging from 15 Gy to 24 Gy)
		BCNU 80 mg/m2 IV days 1, 2, and 3 of RT then every 8 weeks for a total of 6 cycles
RTOG 9411	II	Accelerated hyperfractionated RT (64.0 or 70.4 Gy) plus BCNU (80 mg/m2 days 1, 2, and 3 of RT and repeated on days 56, 57, and 58 then every 8 weeks for 4 cycles for a total of 6 cycles)	108	108
RTOG 9417	II	Conventional RT with intravenously administered tirapazamine (159 mg/m2 or 260 mg/m2)	124	124
RTOG 9513	II	Cranial RT plus topotecan (1.5 mg/m2 per day IV for 3 days/wk every 3 weeks for 3 cycles)	87	87
RTOG 9602	II	Conventional RT plus weekly paclitaxel (225 mg/m2/3 hours/wk × 6)	62	62
RTOG 9710	II	Conventional RT followed by recombinant β-interferon (6 million U intramuscularly administered 3 times per week; 3 weeks on drug, 1 week off drug)	109	109
RTOG 9803	I/II	Conformal RT to doses of 66 Gy, 72 Gy, 78 Gy, or 84 Gy with BCNU (80 mg/m2 days 1, 2, and 3 of RT and repeated on days 56, 57, and 58 then every 8 weeks for 4 cycles for a total of 6 cycles)	209	209
RTOG 9806	II	Conventional doses of RT with incremental increases of thalidomide starting at 200 mg/d with escalations to a maximal dose of 1,200 mg	128	128
RTOG 0013	II	Using conventional RT followed by intra-tumoral bleomycin that was delivered with a refillable sustained device	19	19
RTOG 0021	II	Conventional RT plus high dose tamoxifen (escalated from 20 mg/d to 80 mg/d)	77	77
RTOG 0023	II	50 Gy of external-beam RT in 2 Gy fractions followed by a stereotactic RT boost (4 treatments of 5 Gy for tumors > 40 mm or 7 Gy for tumors ≤ 40 mm once per week during weeks 3-6) along with BCNU (80 mg/m2 IV for 3 days, beginning within 1 month after completion of RT then every 8 weeks for a total of 6 cycles)	80	80

Abbreviations: RTOG, Radiation Therapy Oncology Group; GBM, glioblastoma multiforme; ECOG, Eastern Cooperative Oncology Group; RT, radiation therapy; BCNU, carmustine; MeCCNU, 1-(2-Chloroethyl)-3-(4-Methylcyclohexyl)-1-Nitrosourea; IV, intravenously; AZQ, diaziquone.

Statistical Methods

The analyses are based on the data used for manuscripts or presentations. Because a new treatment standard for GBM (as manifest by a statistically significant survival advantage attributable to any given therapeutic arm) did not emerge from the respective studies, there was justification in pooling the data. Survival was measured from the date of study registration to the date of death or last follow-up, and survival rates were estimated using the Kaplan-Meier method.²² The log-rank test was used to compare survival between the interval groups.²³ Outcome was assessed with the aid of the recursive partitioning technique (a method of building decision trees to model predictors) that was previously published by the RTOG.²⁴ The cutoff points used to define the groups (ie, intervals of ≤ 2 weeks, > 2 to 3 weeks, > 3 to 4 weeks, > 4 weeks) were selected based on the distribution of intervals using percentiles (25%, 50%, 75%). A Cox proportional hazards model was also performed on this database using the variables common to all of the clinical trials that were included in the analysis.²⁵

RESULTS

Patient Characteristics

Within the RTOG database for GBM, 3,052 patients were deemed suitable for the current analysis. Of these, 78 were rendered ineligible because of incomplete outcome data. In 119 cases, the chronology of treatment could not be determined (ie, the date of surgery was unknown for 70 patients; the initiation date of radiotherapy was unknown for 49 patients). Accordingly, 2,855 patients were included in this analysis.

Table 2 lists the pretreatment characteristics for the patients studied as a function of the interval from surgery to the start of radiotherapy. Relationships between pretreatment characteristics (performance status [Karnofsky Score/Zubrod scale], neurologic function, mental status, type of surgical procedure [biopsy, partial resection, total resection], and recursive partitioning analysis [RPA]class), and the time interval were assessed with χ² tests and revealed statistically significant associations between them (P < .0001).

Table 2.

Pretreatment Characteristics

Characteristic	Interval From Surgery to Start of Radiation
	≤ 2 Weeks (n = 756)		> 2 to 3 Weeks(n = 805)		> 3 to 4 Weeks (n = 757)		> 4 Weeks (n = 537)
	No.	%	No.	%	No.	%	No.	%
Days from surgery to start of radiation
Median	12		19		26		33
Range	1-14		15-21		22-28		29-73
Age, years
Median	57		56		57		55
Range	14-83		19-83		18-86		17-82
Dose, Gy
No. of patients with dose	494		676		672		424
Median	60		60		60		60
Range	1.7-82.87		6-85.68		2-86		2-84.32
Performance score*
Karnofsky
20-50	77	11	36	4	21	2	4	< 1
60	82	11	77	10	65	9	30	6
70	120	16	132	16	97	13	69	13
80	170	22	185	23	174	23	115	21
90	176	23	248	31	263	35	189	35
100	31	4	68	8	72	10	62	12
Zubrod
0	12	2	16	2	35	5	36	7
1	36	5	29	4	21	3	26	5
2	31	4	9	1	4	< 1	2	< 1
3	16	2	5	1	4	< 1	4	< 1
4	5	< 1	0	0	1	< 1	0	0
Prior surgery
Biopsy	223	29	178	22	124	16	94	18
Partial resection	399	53	436	54	426	56	290	54
Total resection	114	15	167	21	190	25	140	26
Other	15	2	26	2	9	1	10	2
Unknown	5	< 1	8	1	8	1	3	< 1
Neurologic impairment
None/minor	290	38	421	52	470	62	375	70
Moderate	351	46	318	40	257	34	151	28
Severe	107	14	61	8	29	4	9	2
Unknown/missing	8	1	5	1	1	0	2	0
Mental status
Normal function	395	52	495	61	486	64	348	65
Minor mental confusion	294	39	248	31	199	26	125	23
Gross confusion	41	5	26	3	21	3	7	1
Rousable with difficulty	4	1	2	< 1	0	0	0	0
Unknown/missing	22	3	34	4	51	7	57	11
RPA stage
III	83	11	129	16	119	16	113	21
IV	246	32	317	39	366	48	271	50
V	298	39	274	34	215	28	124	23
VI	127	17	82	10	53	7	28	5
Unknown	2	< 1	3	< 1	4	< 1	1	< 1

Abbreviation: RPA, recursive partitioning analysis.

^*Karnofsky was collected on studies 9806, 9803, 9710, 9602, 9513, 9417, 9411, 9305, 9006, 8409, 8302, and 7918; Zubrod was collected on studies 0023, 0021, and 0013; both Karnofsky and Zubrod were collected for study 7401 (for that study, Karnofsky performance score is reported where available).

Outcome Data

Figure 1 displays the overall survival with regard to the four interval (ie, time between surgery and initiation of radiotherapy) groups. The median survival time of those with the longest interval (ie, > 4 weeks) was significantly greater than the median survival time of the group with the shortest interval (ie, ≤ 2 weeks; 12.5 months v 9.2 months; P < .0001). When this was examined by RPA stage, only RPA V showed the same trend.

Fig 1.

Overall survival by surgery-to-radiation therapy time interval groups.

Twenty-one patients (0.7%) were enrolled at more than 6 weeks from time of surgery (average, 7.2 weeks; range, 6.1 to 10.4 weeks; median, 6.7 weeks). This number was too small to calculate any significant difference or trend as a separate interval group; however, interestingly, the average survival time for this group was higher than expected at 15 months.

Multivariate analysis of overall survival including time intervals (interval of > 4 weeks) and RPA stage showed that both variables are statistically significant factors with respect to overall survival (Table 3). Treatment effects (whole-brain radiation therapy arms v radiation therapy arm and radiation therapy alone arms v radiation therapy plus radiosensitizers or chemotherapy arms), interaction terms with intervals, and RPA classes were also added to the multivariate analysis, and no significant effects were found.

Table 3.

Cox Proportional Hazards Model for Overall Survival

Covariate	HR	95% CI	P
Time from surgery to start of RT, weeks
≤ 2	RL		—
> 2-3	0.97	0.87 to 1.07	.49
> 3-4	0.91	0.82 to 1.01	.07
> 4	0.84	0.75 to 0.95	.004
RPA
III	RL		—
IV	1.72	1.53 to 1.93	< .0001
V	2.74	2.42 to 3.10	< .0001
VI	4.39	3.74 to 5.14	< .0001

Abbreviations: HR, hazard ratio; RT, radiation therapy; RL, reference level; RPA, recursive partitioning analysis.

Table 4 was constructed to determine whether there was a disproportionate representation of progression during treatment or shortly thereafter among those who initiated radiotherapy within shorter intervals. No statistically significant differences in the rates of progression during treatment or in the month immediately after completion of radiation were observed across the four intervals (≤ 2 weeks, > 2 to 3 weeks, > 3 to 4 weeks, and > 4 weeks) studied.

Table 4.

Progression Period if Progressed

Progression	Interval From Surgery to Start of Radiation
	≤ 2 Weeks (n = 488)		> 2 to 3 Weeks (n = 596)		> 3 to 4 Weeks (n = 587)		> 4 Weeks (n = 415)
	No.	%	No.	%	No.	%	No.	%
During radiation	45	9	47	8	35	6	22	5
Within 30 days of radiation end	82	17	70	12	89	15	63	15
After 30 days of radiation end	361	74	479	80	463	79	330	80

DISCUSSION

The issue of prolonged waiting time for radiotherapy has been underscored in the literature for nearly two decades. The problem has now reached a state of crisis even in certain modern countries that enjoy a high quality of life by Western standards, such as Canada and Australia.^26,27 Longer waiting times are a source of anxiety among patients and health care professionals because of the presumed deleterious effect of delay on tumor control.

The impact of delayed initiation of radiotherapy after surgery has been extensively studied in two settings: carcinoma of the breast as well as head and neck cancer.²⁸ In the case of breast cancer, the most common interval studied is 8 weeks between surgery (usually lumpectomy) and the first administration of radiotherapy. In most of the reported studies,^29–33 5-year locoregional recurrence rates in women treated with postoperative irradiation begun more than 8 weeks after surgery (approximately 9%) are significantly higher than those women treated within 8 weeks of surgery (approximately 6%). In several studies,^32,33 there also seemed to be an increased rate of distant metastases among women who received postoperative irradiation initiated more than 8 weeks after surgery. Among patients with unresected cancers of the head and neck, 1-month delays in the initiation of radiotherapy tended to increase the risk of local recurrence at 5 years.^12,13,34 For head and neck cancers managed with primary surgery followed by postoperative irradiation,^35–38 there was a higher probability of locoregional recurrence in patients treated by irradiation begun more than 6 weeks after the operation. In one study,¹⁴ actuarial 5-year survival rates were 61%, 46%, and 30% for patients with non–small-cell lung cancer who underwent radiation at 1 to 6 weeks, at 7 to 8 weeks, and at more than 8 weeks after surgery, respectively (P = .046). Although the impact of delayed thoracic radiation has been studied in the context of small-cell and non–small-cell lung cancer, it is difficult to draw conclusions from the literature as a result of the common use of sequential regimens that interpose chemotherapy between surgery and radiotherapy.³⁹

To date, only one report has specifically examined the impact of delayed radiotherapy among patients with high-grade glioma.¹⁶ In that study, three variables were identified that had a negative impact on survival: older age, reduced radiation dose, and delayed time from presentation to the radiation department until initiation of radiotherapy, where the hazard of death increased by 2% per day. Several caveats are worth underscoring. First, this report studied grade 3 as well as grade 4 gliomas. Although it was customary in the past to combine these two entities (even in some RTOG trials),^3,5,17,19 it is likely that they behave as distinctly separate diseases. Second, the authors did not find a detrimental effect on survival with increased time interval between surgery and commencement of radiation; however, interestingly, they did find a significant effect of decreased survival in patients with longer waiting intervals between presentation to the radiation department and the initiation of radiation treatment. The RTOG database did not track the data point of presentation to the RT department. Third, these observations were made at a single institution. Part of the robustness of the RTOG database is derivative of the vast numbers of patients with retrievable follow-up information and the wide cross-section of medical centers participating, theoretically creating a better reflection of the broader reality. Finally, the RTOG mandated the relatively prompt initiation of treatment as an eligibility criterion for study participation (ie, maximal 6-week wait). The much wider window tolerated by Do et al (range, 1 to 62 days) may have had a negative impact on survival in certain cases.

Notwithstanding, it is noteworthy that our results do not comport with trends recognized in the experience with carcinoma of the breast and head and neck cancer that are outlined above. As well, a recent report that assessed neoadjuvant temozolomide before radiotherapy for newly diagnosed patients with GBM was, by admission of the authors, inferior to standard concomitant radiotherapy plus temozolomide.⁴⁰ Whether this poor result by Chinot et al reflects a suboptimal radiation–drug interaction, a cohort that had an overrepresentation of patients expressing high levels of methylguanine methyltransferase,⁴¹ or simply the consequence of introducing up to a 4-month delay secondary to the induction regimen is unclear.

Although 16 trials comprise the current report, none of these studies were designed a priori to address the specific issue of interval to initiation of radiation therapy after neurosurgical intervention. In fact, the ethical legitimacy of such a protocol design may be dubious at best, and it is unlikely that physicians would marshal the equipoise to conduct such a study. Thus despite the methodologic appeal of designing a prospective randomized trial to directly assess the impact of timing of radiotherapy, it is unlikely that such an effort will ever be mounted. As such, caution must be exercised in assessing whether a true effect of interval prolongation (either beneficial or detrimental) actually exists. Furthermore, although approximately 3,000 patients were entered into the current analysis, it is unclear that the retrospective nature of the review allowed sufficient power to exclude the possibility that delays in initiating radiotherapy may have had small but clinically meaningful deleterious effects.

The danger inherent in the delay of radiotherapy for patients with GBM seemed axiomatic. GBM is known to have a short doubling time.⁴² In addition, like many tumors, GBM tends to invade locally, and the probability of achieving control is expected to decline as the tumor size increases.⁷ Accordingly, we were surprised to see that our data did not lend support to this expectation and even yielded a result that was counterintuitive.

It is difficult to propose a plausible mechanism for an association between delayed therapy and improved survival in the treatment of GBM. We explored an alternative explanation, which posits that we were detecting a pragmatic epiphenomenon rather than a true biologic reality. Physician's intuition may have lead to expedited treatment for those patients who looked particularly fragile. If indeed, such patients went on to experience treatment failure quickly then we have simply used exotic statistical techniques to validate the astute judgment of clinicians who selected patients for prompt treatment on the basis of their clinical judgment that therapeutic intervention was required expeditiously. In other words, given the finite resources in many systems, physicians may have chosen to hasten the initiation of treatment for patients with the most advanced tumors when circumstances did not allow all patients to be treated with equal immediacy. As seen in Table 2, there is a larger number of patients with Karnofsky Performance Score of 70 or less/Zubrod score of 3 to 4 in the group radiated earlier; likewise, there is an overrepresentation of patients undergoing biopsy in the groups of patients who underwent radiation earlier.

A surrogate means of checking the hypothesis that physicians treated patients with poorer prognostic factors differently was probed, as displayed in Table 4, by evaluating for early tumor progression as correlating with earlier initiation of radiation therapy. However, in assessing these data, it is impossible to conclude that the more aggressive tumors (as manifest by a propensity to recur during therapy or within a month after completion) were overrepresented in the groups with the shortest intervals between surgery and initiation of radiotherapy.

Finally, there may be a detrimental effect to the injured organ (the brain) when treated with radiation too soon after the primary insult of surgery. Hypoxia from surgical manipulation and edema in the immediate postoperative period may diminish radiosensitivity.

Data from a study using rat models to examine brain surgery followed by radiation at differing onsets (also using controls without radiation) suggests that early initiation of radiation (within 1 to 2 weeks) after surgery, compared with 3 weeks or more, may result in higher levels of tissue damage.⁴³

In summary, within the relatively narrow temporal limits (6 weeks) of initiating radiotherapy after surgery that were permitted by the RTOG in the trials reviewed, we were unable to uncover a disadvantage associated with delayed radiotherapy. However, it is difficult to construct a rationale for the conscious implementation of such delays. Additionally, treatment initiated beyond 6 weeks postoperatively may well be detrimental, but it is beyond the scope of our observations, and it is not feasible to design a prospective trial to test this hypothesis.

Although we are disinclined to recommend deliberately forestalling radiotherapy among patients suffering from GBM, physicians may be able to reassure those patients who are waiting for treatment to commence that cancer control is unlikely to be compromised so long as the guidelines proposed by the RTOG investigators are respected.

AUTHORS' DISCLOSURES OF POTENTIAL CONFLICTS OF INTEREST

Although all authors completed the disclosure declaration, the following author(s) indicated a financial or other interest that is relevant to the subject matter under consideration in this article. Certain relationships marked with a “U” are those for which no compensation was received; those relationships marked with a “C” were compensated. For a detailed description of the disclosure categories, or for more information about ASCO's conflict of interest policy, please refer to the Author Disclosure Declaration and the Disclosures of Potential Conflicts of Interest section in Information for Contributors.

Employment or Leadership Position: None Consultant or Advisory Role: Deborah T. Blumenthal, Schering-Plough (C), Eisai (C) Stock Ownership: None Honoraria: Deborah T. Blumenthal, Schering-Plough Educational Forum Research Funding: None Expert Testimony: None Other Remuneration: None

AUTHOR CONTRIBUTIONS

Conception and design: Deborah T. Blumenthal, Jeff M. Michalski, C. Leland Rogers, Benjamin W. Corn

Financial support: Walter J. Curran

Administrative support: Minesh P. Mehta, Walter J. Curran, Jeff M. Michalski

Provision of study materials or patients: Walter J. Curran, Luis Souhami, Jeff M. Michalski, Benjamin W. Corn

Collection and assembly of data: Minhee Won, Walter J. Curran, Jeff M. Michalski

Data analysis and interpretation: Deborah T. Blumenthal, Minhee Won, Minesh P. Mehta, Jeff M. Michalski, C. Leland Rogers, Benjamin W. Corn

Manuscript writing: Deborah T. Blumenthal, Minhee Won, Minesh P. Mehta, C. Leland Rogers, Benjamin W. Corn

Final approval of manuscript: Deborah T. Blumenthal, Minhee Won, Minesh P. Mehta, Walter J. Curran, Luis Souhami, Jeff M. Michalski, C. Leland Rogers, Benjamin W. Corn