Radiation choice for brain metastases requires ‘transparent discussion of tradeoffs’

Editor’s note: An abbreviated version of this column appeared in HemOnc Today’s Feb. 10 issue as a Point/Counter. To read a response to this commentary, click here.

The Choosing Wisely recommendations from the American Society for Radiation Oncology state to not “routinely add adjuvant whole brain radiation therapy [WBRT] to stereotactic radiosurgery for limited brain metastases.”

This statement has been interpreted by many as the demise of WBRT in limited metastatic disease, which has been reinforced by the publication of the QUARTZ trial, which showed no OS or quality-of-life benefit with WBRT compared with supportive care in a predominantly hospice-eligible patient population. However, we believe that this is an extreme viewpoint, and the decision to offer WBRT should entail considerable personalization that requires a very balanced approach on the part of the radiation oncologist with an honest and open risk-benefit discussion with the patient.

Randomized trials have consistently demonstrated the benefits of WBRT in combination with locally aggressive treatment (ie, surgery or stereotactic radiosurgery [SRS]), typically in the form of improved local control, decreased intracranial progression and reduction of neurologic deaths. The benefits of WBRT are now perceived and presented in a dilutionary manner due to the lack of OS benefit, the high local control offered by SRS and the neurocognitive toxicity associated with it. Despite publication of randomized trials, the exact clinically relevant magnitude of this toxicity remains poorly elucidated, and the various strategies available to mitigate cognitive dysfunction described in prospective studies are used relatively infrequently in standard practice.

Historical data on WBRT

WBRT has been integral in the management of brain metastases for over half a century primarily because of its ability to produce symptomatic relief, decrease the likelihood of intracranial relapse and because of the possibility of enhancing survival among well-selected patient subsets.

Trials consistently and reproducibly show that WBRT in combination with local treatment — surgery or SRS — improves local control and decreases intracranial progression.

For instance, Sahgal and colleagues reported on a combined meta-analysis of three randomized trials that included a total of 364 patients (WBRT with SRS, 49%; SRS alone, 51%). Results recapitulated the well-proven dictum: Patients with WBRT had improved local control compared with SRS alone, with an overwhelmingly massive HR in favor of WBRT (HR = 2.56; 95% CI, 1.54-4.26). The risk for distant failure also was lower with WBRT, although limited to patients aged older than 50 years, a possible statistical quirk, given the relatively small number of patients.

The majority of the patients recruited to participate in these randomized clinical trials had excellent performance status with active, but limited, extracranial disease. On the other hand, more than 70% of patients enrolled in the QUARTZ trial had a woefully low graded prognostic assessment score (GPA) of 0 to 2 and, thus, had extremely poor prognosis before commencement of any treatment. Consequently, it is not surprising that WBRT had no advantage in terms of quality of life or even survival in a group whose predicted median life expectancy ranged from 2 to 4 months; no therapy can demonstrate survival gain in this clinical setting.

Despite the categorical benefits of WBRT in terms of reducing, delaying and preventing intracranial progression and relapse when combined with local treatment, a common argument against it is the absence of an OS benefit.

Patchell and colleagues showed the importance of surgery in combination with WBRT among patients with a single metastatic lesion to the brain. Local occurrence decreased (20%; P < .01) and OS improved (40 weeks vs. 15 weeks; P < .02).

In RTOG 9508, researchers randomly assigned 331 patients with one to three newly diagnosed brain metastases to receive WBRT (n = 167) or WBRT plus SRS (n = 164). OS served as the primary endpoint. Patients with a single metastatic lesion had an improved median survival with WBRT plus SRS (6.5 months vs. 4.9 months; P = .0393), but SRS did not improve survival for patients with more than one lesion.

A lack of survival benefit from SRS has not precluded its widespread use among patients with multiple brain metastases, with the contention that it improves local control without significant neurotoxicity and dramatically escalates intracranial relapse — HRs of 2.5 or more. Deliberately withholding WBRT is deemed an acceptable consequence by the physician with the contention to the patient any recurrences can be salvaged by performing multiple sequential follow-up MRI scans with multiple repeat SRS applications, and ultimately WBRT.

Reducing intracranial relapse risk

Other than for patients with a single metastatic brain lesion, improvement in OS has rarely been shown with any modality.

In a secondary analysis of RTOG 9508, researchers stratified outcomes based on GPA score. In a subset of patients with a high GPA score (3.5-4), researchers observed a trend toward an OS benefit with the addition of SRS to WBRT (21 months vs. 10.3 months; P = .05), regardless of the number of metastatic lesions present. This provides the hypothesis that, among patients less likely to die of extracranial disease progression (a consequence of high GPA score), improved local control provided by SRS could perhaps translate to improved OS, even among patients with more than one lesion; of course, this hypothesis would apply to any treatment capable of improving intracranial control, even WBRT, in the correct clinical scenario.

In fact, this was borne out in a similar secondary analysis of the JROSG 99-1 trial by Aoyama and colleagues. When researchers stratified patients with non-small lung cancer by their diagnosis-specific GPA score, those with a more favorable score (2.5-4) had a significantly improved median OS with WBRT plus SRS compared with SRS alone (16.7 months vs. 10.6 months; P = .04). These data for WBRT — similar to the data for SRS from RTOG 9508 — suggest that decreasing the risk for intracranial relapse with any modality, in this case WBRT, may play a more critical role in improving survival among patients who have a lower likelihood of dying of extracranial disease progression.

Pirzkall and colleagues further corroborated this hypothesis in a study where they observed a trend toward survival improvement among patients who had controlled extracranial disease and who received the addition of WBRT to SRS (15.4 months vs. 8.3 months, P = .08). A similar analysis of the Alliance trial failed to corroborate this observation, but the bulk of the published data support this biologically rational hypothesis.

Columbia University’s experience further reinforces the observation that prognosis based on GPA and treatment modality can affect survival outcomes. In their 15-year institutional analysis of 528 patients treated with stereotactic radiosurgery, researchers showed a 40% reduction in mortality in a multivariate analysis among patients who received WBRT plus SRS compared with SRS alone (HR = 0.63; 95% CI, 0.47-0.84). Once again, this benefit appeared more relevant among patients who had a better performance status, controlled extracranial disease and higher GPA scores.

Role of systemic therapy

This issue is becoming significantly more relevant as targeted agents become more readily available for patients with metastatic disease and the efficacy of systemic agents, including their intracranial activity, continues to improve.

The role of intracranial control with WBRT in improving survival becomes more critical for these patients as their risk for extracranial death diminishes. The best example of this phenomenon comes from improved systemic treatment among patients with metastatic/locally advanced NSCLC in the form of EGFR tyrosine kinase inhibitors, ALK translocation or ROS1 inhibitors, and immunotherapeutic agents.

In a multi-institutional analysis by Magnuson and colleagues, the use of SRS and WBRT in conjunction with EGFR TKIs both significantly improved OS (P < .001), despite that the WBRT cohort had a less favorable prognosis. With the availability of the highly active agent osimertinib (Tagrisso, AstraZeneca) — a next-generation EGFR TKI — researchers are launching new combinatorial trials evaluating radiation, including WBRT.

Neurocognitive toxicity

Emerging prospective data have underscored the adverse neurocognitive effects of WBRT. This cognitive detriment has led many providers to not offer WBRT to patients with limited intracranial disease.

Researchers from The University of Texas MD Anderson Cancer Center conducted a prospective study to test whether the addition of WBRT to SRS worsened learning and memory functions 4 months after completion of treatment using the Hopkins Verbal Learning Test total recall assessment. Through the use of Bayesian probability, researchers determined that the addition of WBRT to SRS had a significant detriment in terms of total recall (52% vs. 24%), delayed recall (22% vs. 6%) and delayed recognition (11% vs. 0%) among only 31 patients (WBRT plus SRS, n = 11; SRS, n = 20). The authors concluded that SRS is the preferred treatment strategy.

The shortcomings of this small trial have already been described. For example, the single time point of cognitive dysfunction 4 months after treatment deserves some attention. Previous studies have shown that WBRT can cause a transient effect on memory that nadirs at 4 months and improves later, thus contributing tremendous bias to their outcomes. Further, researchers did not stratify patients based on baseline neurocognitive function. Their data suggest that patients randomly assigned to WBRT had worse baseline neurocognitive function.

Saito and colleagues have shown how critical baseline cognitive testing — also influenced by patient age and performance status — can be in terms of impacting post-WBRT cognitive testing. Their trial showed a survival difference between the two arms, and several published studies demonstrated that patients with cancer experienced intense neurocognitive dysfunction before death, implying possible interarm prognostic imbalance.

The Alliance study was a larger (n = 213) randomized trial investigating the addition of WBRT to SRS on cognitive dysfunction among patients with one to three brain metastases, assessed by a battery of seven neurocognitive tests. Cognitive deterioration — defined as a decline of greater than one standard deviation from baseline in any one or more of the seven cognitive tests at 3 months — served as the primary endpoint.

Researchers designed the trial to detect “a change in the anticipated 3-month cognitive deterioration rate of 0.65 for the SRS plus WBRT group, based on a clinical trial of patients with brain metastases treated with WBRT who were prospectively assessed at baseline and over time with a cognitive battery, to 0.4 or lower for the SRS group.” This baseline hypothesis can be restated as follows: When added to SRS, WBRT provides meaningful and substantial improvement in intracranial disease control — but does not improve survival — and can produce deterioration in neurocognitive test results among 65% of patients, which is unacceptably high, and SRS alone will cause such neurocognitive decline among “only” 40% of patients.

Yet, results from the Alliance trial by Brown and colleagues showed an unacceptably high 3-month rate of cognitive decline in the SRS alone arm at 63.5% — close to the 65% “unacceptable” rate from WBRT that was speculated as the trial hypothesis — which was, however, statistically better than the 91.7% rate observed for WBRT plus SRS.

A simple explanation for this high rate of cognitive impairment in the SRS-alone cohort could be the high rate of intracranial failures at 3 months (24.7% vs. 6.4%; P < .001), cognitive dysfunction from disease progression, the need for repetitive salvage brain treatments or the effects of SRS itself.

Brown and colleagues concluded that SRS should be the preferred strategy among patients with limited metastases to the brain given that WBRT does not provide an OS benefit and is associated with worse cognition. Another way to interpret the data could be that SRS also does not improve OS among most patients and results in extremely high disease failure in the brain that requires further, and often invasive, interventions that also result in unacceptably high rates of neurocognitive impairment.

Lim and colleagues evaluated patients with one to four asymptomatic oligometastases from NSCLC randomly assigned to SRS with sequential chemotherapy or chemotherapy alone. The addition of SRS to chemotherapy failed to improve OS. Although there was an observed increase in intracranial progression without SRS, this did not reach statistical significance.

If we take into consideration the Alliance data — in which 63% of patients receiving SRS had acute cognitive deterioration in one or more domains at 3 months, with 50% intracranial disease progression at 1 year and no OS benefit — would the next logical step be to exclude SRS in favor of chemotherapy alone for this patient population? Of course not, because most SRS proponents would contend that the rate of clinically significant cognitive neurotoxicity associated with SRS treatment is nowhere near 63%. However, we also know that the 91.7% rate ascribed to WBRT also is nowhere near clinical reality!

Assessing, addressing impact on cognition

Another possible explanation of the high neurocognitive decline reported in the Alliance trial could be due to the use of exquisitely sensitive tests in addition to using a relatively low threshold — only one standard deviation to score any one of several possible events — which may pick up a lot of “noise” and possibly cognition “changes” from baseline that may not be truly relevant to the patient’s real-life experience.

In fact, we, like most others, are hard-pressed to describe with clinical precision what a one standard deviation change in one of several tests at one point in time actually means to a patient.

Alternatively, a more robust two-to-three standard deviation change with use of an actual Bonferroni correction for multiple observations (ie, treating each test as an independent observation and dividing the necessary P value by the number of tests administered before arriving at a summative score), may result in more clinically relevant outcomes.

Indeed, if a 1.5 standard deviation threshold is used in combination with a drop in two tests, the 3-month deterioration drops to 19% for SRS alone and to 46% for WBRT.

Although this may be more reflective of what truly occurs in this patient population as assessed by these neurocognitive tests, which threshold is fundamentally meaningful in a clinical setting? Unless we have an answer to this question, let us not throw away the baby with the bathwater.

The use of cruder measures — like the mini-mental status examination (MMSE) — might seem retrogressive until we take into consideration that the majority of patients with metastatic brain disease still measure their survival in months. In that regard, a change in a test score, unless it can be correlated to a meaningful clinical outcome, probably does not mean much for this patient cohort.

The MMSE is a crude test, but it is fairly sensitive in detecting severe drops in cognitive dysfunction, which is precisely what it was designed to measure and detect.

In the Aoyama trial, researchers used a 3-point drop as a measure of severe cognitive decline.

Likelihood of a 3-point drop was 23.9% at 12 months and 31.5% at 24 months for SRS with WBRT compared with 40.7% at 12 months and 48.1% at 24 months for SRS alone. These data indicate that, in the first 24 months or so, the severe decline in cognition induced by disease progression is worse than the effect of WBRT.

It is somewhat of a lost opportunity that the Alliance trial did not include the simple and easy-to-use MMSE in addition to the battery of cognitive tests.

Another concern with the use of neurocognitive batteries in this cohort is that patients may decline on one test at one time point and another test at some other time. There also is the possibility of “score recovery” and subsequent decline in certain domains, creating a dynamic picture of cognitive test score change that cannot be accurately represented in a one-dimensional reporting of change in any one test at one point in time. We should really be questioning not only the tools and methodology used to measure neurocognitive function among this group of patients, but the validity, interpretation and applicability of these results in a clinical setting.

Various methods have been prospectively evaluated to mitigate cognitive detriment associated with WBRT.

Researchers of the phase 3 randomized RTOG 0614 trial investigated whether memantine — used for the treatment of Alzheimer’s disease — in combination with WBRT decreased the extent of cognitive dysfunction. Memantine conferred an 11% reduction in the probability of cognitive dysfunction at 6 months and better preserved executive function, processing speed and delayed recognition.

Another phase 3 randomized trial evaluated Alzheimer’s drug donepezil given 6 or more months after WBRT. Although it did not significantly improve the composite cognitive battery score, donepezil resulted in observable improvement in several cognitive functions, particularly among patients who had greater initial impairment before treatment.

Hippocampal-avoidance WBRT has emerged as an alternative method for delivering WBRT.

RTOG 0933 — a phase 2 multi-institutional trial of hippocampal-avoidance WBRT among 42 patients with brain metastases compared with historic controls — showed only a 7% decline in Hopkins Verbal Learning Test delayed recall testing at 4 months (P < .001). This intervention is being investigated in two randomized trials: NRG-CC001, among patients with brain metastases, and NRG-CC003, as prophylactic cranial irradiation among patients with small cell lung cancer.

Limitations of SRS

Accessibility to high-quality SRS can be limited, especially in rural areas in the United States and worldwide.

More importantly, SRS is a technically challenging multidisciplinary therapeutic strategy, requiring high-quality continuous interaction among members of a specialized highly experienced team, a luxury that is simply unavailable in a number of smaller U.S. centers.

ASTRO strongly encourages against routine WBRT in combination with local therapy for patients with limited metastatic disease. Yet, in places where SRS is not accessible, WBRT remains a “wise” option as opposed to completely eradicating this treatment paradigm or attempting SRS as an occasional strategy, which is probably riskier than using WBRT routinely.

To achieve durable local control with SRS alone, data from Cleveland Clinic suggested using a higher radiation dose via a two-step SRS method, which may also result in a modest increase in radiation necrosis and other adverse events. A multidisciplinary team approach is needed to appropriately manage radiation necrosis, and access to appropriate care teams remains critical, which could further limit the availability of SRS.

Need for open discussions

Radiation oncologists are counseled to strongly consider omitting WBRT in combination with SRS among patients with limited metastatic disease to the brain, despite the dramatic increase in intracranial progression and the possibility of an OS benefit in appropriately selected patients — ie, those with a favorable GPA score and well-controlled extracranial disease. With improvements in systemic treatments for patients with metastatic cancers, intracranial control becomes even more crucial and has a greater likelihood of impacting survival.

Although the increased risk for neurocognitive impairment with WBRT is well known, whether the rates of clinically meaningful decline are as high as the test performance changes reported in the Alliance trial is highly questionable. If one is to accept those rates at face value, we would need to start counseling our patients undergoing SRS that they face a two-in-three chance of cognitive decline within 3 months — something that almost no practitioners of radiosurgery do, and something that was completely omitted from ASTRO’s Choosing Wisely recommendations.

We are of the opinion that both risks and benefits need to be thoroughly discussed in any decision affecting a patient, and if one is to abandon WBRT based on the Alliance cognitive test score change results, then the extremely high rates of cognitive decline associated with SRS in that trial must be part of the informed written medical procedure consent form. We posit that advances in radiation delivery through hippocampal-avoidance WBRT and pharmacological interventions will likely result in reconsideration of this blanket exclusion of WBRT.

Further clinical trials are needed — and are underway — to discern the specific subset of patients who are likely to benefit from WBRT in addition to SRS. In the meantime, treatment selection should include a comprehensive, unbiased and transparent discussion of tradeoffs with patients.

References:

Angelov L, et al. J Neurosurg. 2017;doi:10.3171/2017.3.JNS162532.

Andrews DW, et al. Lancet. 2004;doi:10.1016/S0140-6736(04)16250-8.

Antonia SJ, et al. N Engl J Med. 2017;doi:10.1056/NEJMoa1709937.

Aoyama H, et al. JAMA. 2014;doi:10.1001/jama.295.21.2483.

Aoyama H, et al. JAMA Oncol. 2015;doi:10.1001/jamaoncol.2015.1145.

Armstrong CL, et al. Neuropsychiatry Neuropsychol Behav Neurol. 2000;13:101-111.

Borghaei H, et al. N Engl J Med. 2015;doi:10.1056/NEJMoa1507643.

Brahmer J, et al. N Engl J Med. 2015;doi:10.1056/NEJMoa1504627.

Brown PD, et al. JAMA. 2016;doi:10.1001/jama.2016.9839.

Brown PD, et al. Neuro Oncol. 2013;doi:10.1093/neuonc/not114.

Chang EL, et al. Lancet Oncol. 2009;doi:10.1016/S1470-2045(09)70263-3.

Chao JH, et al. Cancer. 1954;doi:10.1002/1097-0142(195407)7:4<682::AID-CNCR2820070409>3.0.CO;2-S.

Chao ST, et al. Int J Radiat Oncol Biol Phys. 2013;doi:10.1016/j.ijrobp.2013.05.015.

Folstein MF, et al. J Psychiatr Res. 1975;doi:10.1016/0022-3956(75)90026-6.

Fukuoka M, et al. J Clin Oncol. 2011;doi:10.1200/JCO.2010.33.4235.

Gondi V, et al. Int J Radiat Oncol Biol Phys. 2013;doi:10.1016/j.ijrobp.2012.11.031.

Gondi V, et al. J Clin Oncol. 2014;doi:10.1200/JCO.2014.57.2909.

Herbst RS, et al. Lancet. 2016;doi:10.1016/S0140-6736(15)01281-7.

Kocher M, et al. J Clin Oncol. 2011;doi:10.1200/JCO.2010.30.1655.

Lawlor PG, et al. Arch Intern Med. 2000;doi:10.1001/archinte.160.6.786.

Lim SH, et al. Ann Oncol. 2015;doi:10.1093/annonc/mdu584.

Magnuson WJ, et al. J Clin Oncol. 2017;doi:10.1200/JCO.2016.69.7144.

Mahmood U, et al. Lancet Oncol. 2010;doi:10.1016/S1470-2045(09)70389-4.

McMillan MT, et al. Radiat Oncol Biol. 2017;doi:10.1016/j.ijrobp.2017.04.004.

Mitchell AJ. J Psychiatr Res. 2009;doi:10.1016/J.JPSYCHIRES.2008.04.014.

Mok TS, et al. N Engl J Med. 2009;doi:10.1056/NEJMoa1208410.

Mulvenna P, et al. Lancet. 2016;doi:10.1016/S0140-6736(16)30825-X.

Noble J, et al. J Thorac Oncol. 2006;doi:10.1016/S1556-0864(15)31641-5.

Patchell RA, et al. N Engl J Med. 1991;325:303-310.

Pereira J, et al. Cancer. 1997;79:835-842.

Pirzkall A, et al. J Clin Oncol. 1998;doi:10.1200/jco.1998.16.11.3563.

Rapp SR, et al. J Clin Oncol. 2015;doi:10.1200/JCO.2014.58.4508.

Reck M, et al. N Engl J Med. 2016;doi:10.1056/NEJMoa1606774.

Regine WF, et al. Int J Radiat Oncol Biol Phys. 2002;doi:10.1016/S0360-3016(01)02645-1.

Rosell R, et al. Lancet Oncol. 2012;doi:10.1016/S1470-2045(11)70393-X.

Sahgal A, et al. Int J Radiat Oncol Biol Phys. 2015;doi:10.1016/j.ijrobp.2014.10.024.

Saito H, et al. Int J Mol Sci. 2016;doi:10.3390/ijms17111834.

Sequist LV, et al. J Clin Oncol. 2013;doi:10.1200/JCO.2012.44.2806.

Shaw AT, et al. N Engl J Med. 2013;doi:10.1056/NEJMoa1214886.

Simone II CB, et al. Transl Lung Cancer Res. 2015;doi:10.3978/j.issn.2218-6751.2015.10.05.

Soria J-C, et al. N Engl J Med. 2017;doi:10.1056/NEJMoa1713137.

Sperduto PW, et al. Int J Radiat Oncol Biol Phys. 2014;doi:10.1016/j.ijrobp.2014.07.002.

Sperduto PW, et al. J Clin Oncol. 2012;doi:10.1200/JCO.2011.38.0527.

Sperduto PW, et al. JAMA Oncol. 2017;doi:10.1001/jamaoncol.2016.3834.

Wang TJ, et al. J Neurooncol. 2015;doi:10.1007/s11060-015-1728-y.

Yap ML, et al. J Glob Oncol. 2016;doi:10.1200/JGO.2015.001545.

For more information:

Andrew B. Lassman, MD, is the John Harris associate professor of neurology and chief of neuro-oncology at Columbia University. He can be reached at abl7@cumc.columbia.edu.

Minesh Mehta, MD, is deputy director and chief of radiation oncology at Miami Cancer Institute. He can be reached at mineshm@baptisthealth.net.

Melissa A.L. Vyfhuis, MD, PhD, is chief resident in the department of radiation oncology at University of Maryland Medical Center. She can be reached at mvyfhuis@umm.edu.

Disclosures: Lassman reports honoraria from AbbVie, Celgene, Cortice, Kadmon, Novocure, prIME Oncology, Sapience and WebMD. Mehta reports consultant roles with AbbVie, Astra-Zeneca, Celgene, Oncoceutics, Tocagen and Varian; and a data and safety monitoring board role with Monteris. Vyfhuis reports no relevant financial disclosures.