Risk of Fatal Cerebrovascular Accidents After External Beam Radiotherapy for Early Stage Glottic Larynx Cancer

Reporter: Abigail Berman, MD
The Abramson Cancer Center of the University of Pennsylvania
Last Modified: October 30, 2012

Presenting Author: Justin E. Bekelman, MD (on behalf of Samuel Swisher-McClure, MD)
Affiliation: Hospital of the University of Pennsylvania, Department of Radiation Oncology, Philadelphia, PA


  • Early-stage glottic cancer can be managed by external beam radiotherapy directed at the glottis, or by surgery removing the part of the vocal cord involved by cancer (cordectomy or partial laryngectomy). Both of these therapies provide high tumor control rates. Definitive radiotherapy is often preferred over surgery because failures can be salvaged with partial or total laryngectomy.
  • The optimal treatment for early-stage glottic larynx cancer should provide high rates of tumor control and larynx preservation. Because these patients have a high cure rate and a life expectancy often of decades, the best treatment should take into account the long-term morbidity from radiotherapy.
  • There is increasing evidence to suggest that head and neck radiotherapy may be associated with an increased risk of vascular stenosis and subsequent cerebrovascular accidents (CVA).
  • Prior studies have not explored long term CVA risk in the early stage patient population or have been limited by small sample size.
  • This study was designed as a retrospective observational cohort study of the Surveillance, Epidemiology, and End Results (SEER) database to compare the incidence of fatal CVA among patients with early-stage glottic larynx cancer receiving either surgery or external beam radiation therapy (EBRT).

Materials and Methods

  • The authors included a study population of patients diagnosed with pathologically confirmed squamous cell carcinoma of the glottic larynx between January 1, 1983 and December 31, 2008. They excluded patients with supraglottic/subglottic extension, impaired vocal cord mobility, nodal or distant metastases according to disease extent variables in SEER.
  • The primary outcome variable of interest was death due to CVA as defined in SEER from death certificates. The secondary outcome was death due to heart disease.
  • The authors took the following potential confounding variables into account:
    • Patient Characteristics: Age, Gender, Race, Hispanic Ethnicity, Marital status
  • Demographic Characteristics: Year of Diagnosis, SEER registry, population size, county level median household income
    • Statistical analysis included calculation of survival and cumulative incidence functions using a competing risks approach. Non CVA related deaths considered competing risks. Cumulative incidence functions were compared using k-sample test statistics. A multivariable competing risks regression models was used to measure the association of treatment with the risk of death due to CVA adjusting for measured confounders and competing causes of death.
    • Secondary analysis also measured the association of treatment with the risk of fatal heart disease through a multivariable competing risks regression


  • There were 7,237 (17%) EBRT patients and 1,484 (83%) surgery patients.
  • Regarding patient characteristics, the mean age at diagnosis was higher in the EBRT arm (64.5 vs 64.5, p=0.01). There was no difference in the proportion of males and females in each arm. There were significantly more white (86.6 vs 85.3%) and fewer black (8.7 vs 10%) and other (3.5 vs 4.4%) races who received surgery (p<0.001).
  • The unadjusted cumulative Incidence of fatal CVA was higher in the EBRT arm at 5, 10, 15, and 20 years at 0.6 vs. 1.0%; 1.4 vs. 2.0%; 1.5 vs. 2.8%; and 1.5 vs 3.7%, respectively.
  • There was no difference in the unadjusted cumulative incidence of heart disease by treatment at 5, 10, 15, or 20 years. The risks were 5.7 vs. 5.2%; 11.2 vs. 10.2%; 14.8 vs. 14.3%; and 19.0 vs 17.7%, respectively.
  • The cumulative incidence of fatal CVA after EBRT was significantly higher than after surgery (p=0.03). The risk of death from other causes was not different in the EBRT and surgery arms.
  • The cumulative incidence of fatal heart disease after EBRT was not significantly higher than after surgery. The risk of death from other causes was not different in the EBRT and surgery arms.
  • Looking at the association between treatment and risk of fatal CVA and heart disease, the risk of fatal CVA using both an unadjusted and multivariable competing risk model was significantly higher in the EBRT arm with HR 1.72 (1.02-2.89, p=0.04) and 1.75 (1.04-2.9, p=0.04), respectively.
  • There was no significant difference in risk of fatal heart disease or overall survival with EBRT versus surgery.

Author's Conclusions

  • The authors conclude that the treatment of early-stage glottic larynx cancer with EBRT is associated with a small but statistically significant increase in the risk of fatal CVA when compared to surgery.
  • These findings suggest the morbidity of EBRT may be limited by reduction of dose to the carotid arteries through advanced conformal radiation treatment techniques. In addition, further evaluation of modern surgical techniques is warranted.
  • All patients treated for early-stage glottis cancer with EBRT should undergo evaluation carotid arteries as survivors for assessment of individual risk of CVA and prevention if needed.

Clinical Implications

  • This is a highly statistically-sound, well-performed retrospective cohort study from SEER demonstrating concerning findings that patients who underwent EBRT for early-stage glottis cancer may be at increased risk for CVA.
  • In addition to its cohort size, one of the study's strengths is its use of a competing risks analysis, which, unlike a Kaplan-Meier analysis, accounts for death from glottic cancer.
  • It can also be seen that there is very little confounding in this analysis as the unadjusted and adjusted cumulative incidences are very similar.
  • The potential limitations for this study include the following:
    • Residual bias after adjusting for measured confounders. In addition, because the SEER database and not the SEER-Medicare database was utilized, there is no data on comorbidities and therefore this could not be accounted for in the analysis.
    • The analysis only includes fatal CVA outcomes and cannot compare the incidence of non-fatal CVA events.
    • This analysis includes many years of treatment spanning older and newer EBRT techniques, which we do not have any information regarding. Theoretically, there many have been a recent decrease in dose to the carotids with improved 3D planning. However, radiation technique for early-stage glottic cancer has remained fairly unchanged over the years and therefore these results likely represent the present risk to patients.
    • This study does not address local control and if there is a relationship between local control and CVA risk. Likewise, a potential confounder is that patients who fail EBRT may receive salvage surgery, leading to a crossover between the arms of this study.
  • An active area of research is utilizing more conformal techniques to reduce dose to the carotids. Chera et al. (IJROBP 2010 Aug 1;77) recently reported a planning study on 5 patients and found that IMRT was able to reduce the dose to the carotid arteries from a median dose of 19 Gy to 4 Gy. It is unclear, however, what the threshold dose for development of atherosclerosis to the carotid arteries is. In addition, this dose will likely be variable based on a patient's other comorbidities. Therefore, the question remains as to 1) what dose should the carotid be reduced to in order to reduce the CVA risk and 2) will excellent local control be maintained with more conformal radiation given the risk of contouring errors and organ motion.