Our use of the Fazekas visual rating scale to assess WMH before RT is novel. The simplicity and speed of the Fazekas rating scale would make its use by radiation oncologists much more practical than volumetric segmentation in a clinic setting when trying to decide whether WBRT would be appropriate for a given patient.27,28 This study found that age (p = 0.003), WMH volume (p = 0.02) and Fazekas score (p = 0.007) are all significant predictors of cognitive decline after WBRT. In multivariable analyses age remained the only significant predictor of cognitive decline in the model with Fazekas score. Hence, our results confirm the findings of Brown et al. that older age was an important predictor of cognitive decline in their multivariable analysis of the CC001 study. It is noteworthy that every patient over 70 in our study experienced cognitive decline after WBRT.
To our knowledge, this the second study to demonstrate that the volume of pre-treatment WMH is a risk factor for cognitive decline after WBRT. WMH is an MRI biomarker for age-related microvascular injury in the brain, so we were not surprised to find that age had a significant linear relationship with log transformed WMH volume (p < 0.0001) and Fazekas score (p < 0.0001). We also found that our two metrics of WMH burden in the brain: WMH volume and Fazekas score, were highly correlated. Our volumetric WMH findings substantiate those from a secondary analysis of a subset of 33 patients treated with hippocampal the RTOG 0933 clinical trial. As we did, Bovi et al. found that a significant correlation between greater baseline WMH volume and older age (rho = 0.38, p = 0.03) could be detected in patients with brain metastases despite the presence of potentially confounding peritumoural edema. They also found that the change in the Hopkins Verbal Learning Test – Revised immediate recognition memory at 4 months after WBRT was correlated with baseline WMH volume (rho = 0.54, p = 0.001), consistent with a greater burden of WMH being associated with a greater decline in a single subtest of memory. In contrast, we used the MoCA test, which assesses eight neurocognitive domains, to study change in cognition. We believe that use of a global cognitive measure increases the robustness of our finding that WMH burden predicts cognitive decline after RT.
A counter-intuitive strength of this study is that it was performed at a single institution: we had full access to the patients’ clinical records, follow-up imaging and attending oncologists. As opposed to a multi-institutional study, in which analyses can only be performed with the data collected on case-report forms and case-report forms can only be completed for patients who attend appointments, we could find out what happened to patients who would have otherwise have been lost to follow-up. It is clear from Fig. 3 that older patients have higher WMH volumes and Fazekas scores than younger patients. However, this plot also shows that there is a subset of older patients with very high WMH volumes and high Fazekas scores. Our study showed that this subset of patients have particularly bad clinical outcomes. After WBRT, the three patients with a Fazekas Score of 6 lived, respectively, for 18 months with full-time home care nursing, 28 months in a nursing home and 6 months in a hospice. Our observation that a Fazekas score of 6 was associated with profound cognitive impairment after WBRT is based on only three patients. Yet, we feel it is important to alert the radiation oncology community about this observation because it is biologically plausible that WBRT could cause precipitous cognitive impairment in patients with extensive microvascular damage to the brain and minimal cognitive reserve. It should also be noted that while studies have suggested that patients with underlying demyelinating disease (e.g. multiple sclerosis) are at risk for severe neurotoxicity after brain radiotherapy, no patients in our cohort had clinical histories or imaging consistent such pathology.37 Based on our results, we suggest that a Fazekas score of 6 should be considered a relative contraindication to WBRT and that alternatives such as surgery, SRS, systemic therapy, or observation should be strongly considered. The randomized QUARTZ trial showed no difference in quality of life between WBRT and best supportive care in unselected lung cancer patients who were ineligible for surgery or SRS.14 However, our findings suggest that for patients with elevated Fazekas scores, WBRT may cause worse clinical outcomes than observation. Our findings also suggest that clinicians should also assess the WMH burden on pretreatment MRI when considering prophylactic cranial irradiation (e.g. small cell lung cancer). We hope that our report will prompt more research in this area.
There are advantages and disadvantages to the two methods we used to measure WMH burden. Segmentation can achieve a more rigorous, quantitative determination of WMH burden; however, due to a lack of well-defined contouring guidelines in the literature, there was uncertainty in some instances. Also, segmentation underestimates the age-related WMH burden when extensive peri-metastatic edema obscures age-related WMHs. According to our three neuro-radiologists, the use of the Fazekas visual rating scale overcame these contouring uncertainties by allowing them to “see past” the generally asymmetric peri-tumoural edema and focus instead on the distribution of contralateral age-related WMH. In Fazekas scoring, there was difficulty in differentiating a normal, thin periventricular increase in T2 signal (pvWMH Fazekas = 0) from a thin, but abnormal, increase in T2 signal along the lateral ventricle (pvWMH Fazekas = 1), due to variation in FLAIR-sequence intensity and slice thicknesses between scans.38,39 It was also sometimes difficult to distinguish an irregular periventricular WMH cap at the end of the lateral ventricle from confluent punctate dWMH in the parenchyma. These challenges in rating low volumes of WMH led us to consider Fazekas score = 0–2 as a “no to low” WMH category, distinct from Fazekas score = 3–6, which we considered as a “radiologically certain” WMH category.
Using the Fazekas score, Mayinger et al. measured the development of post-treatment WMH in patients receiving WBRT and hippocampal-avoidance WBRT (HA-WBRT) for prophylactic cranial irradiation.40 The nine patients in each cohort had the same Fazekas score at baseline, but they found that, while the median Fazekas score was 2 after HA-WBRT, it was only 1 after conventional WBRT. They also observed that there was a significant Fazekas score increase over time after HA-WBRT (p = 0.001). They hypothesized that dose inhomogeneity in the HA-WBRT plans, which had a higher Dmax and greater volumes receiving more than the prescription dose might be responsible for these differences. In our study, patients received WBRT and SRS, so there were small volumes of normal brain that received a high dose and more dose inhomogeneity in our patient’s plans than in conventional WBRT plans. However, in one–year survivors after radiotherapy for brain metastases, Monaco et al. showed that only 3% (1/33) developed leukoencephalopathy after SRS compared to 97.7% (36/37) after WBRT. With this in mind, we believe that the small volumes of brain exposed to SRS doses in our study would have had little influence our main study endpoint. Had this been influential, its effect would have been to hasten and worsen cognitive decline in patients with more and larger brain metastases, but we found no association between number of metastases and MoCA decline.
Our study must be interpreted within the context of its strengths and limitations. Due to our small sample size, we could not control for potential confounders such as education level, number of brain metastases and intracranial volume (as a surrogate for brain volume and cognitive reserve). Also, we could not correlate changes in WMH or edema after radiotherapy with MoCA scoring because the patients enrolled in our phase II clinical trial were followed for local progression and new brain metastases with only gadolinium-enhanced T1 sequences. Although we accounted for local progression, radionecrosis and new brain metastases by censoring patients with these events, we did not measure metastases after WBRT. Hence, we may have missed some cognitive decline events from progressive WMH that were masked by improving cognition from shrinking metastases and peri-metastatic edema. An analysis with a larger cohort should investigate whether the overall radiological response of the brain metastases is associated with improved cognition within cohorts of patients of matched baseline WMH burden. In our phase II clinical trial, using the MoCA was a practical decision based on the need to test cognition in 10 minutes during brief follow-up appointments. Although it tests eight cognitive domains, it may have missed subtle cognitive decline that lengthier neurocognitive testing would have detected, causing our study to underestimate the effects of baseline WMH on cognitive decline. A larger sample size with more detailed neurocognitive testing would allow us to confirm our findings. A secondary analysis of two large phase 3 trials is planned.2,3