1. Extent of resection in the treatment of gliomas Fred G. Barker II, M.D. Neurosurgical Grand Rounds Massachusetts General Hospital February 28, 2008

2. EOR questions GTR vs biopsy? Debulking vs biopsy? GTR vs. near-GTR? GTR vs. GTR plus margin of “normal” tissue (lobectomy)?

3. Resection of malignant glioma Cochrane Review, Metcalfe & Grant, 2001 “The electronic database search yielded 2100 citations. Of these, 2 articles were identified for possible inclusion, however both were excluded. The hand search and personal communication were similarly unproductive. No studies were included in the review and no data were synthesized.”

4. Goals of surgery for malignant gliomas 1. Providing diagnosis 2. Relieving symptomatic mass effect 3. “Setting up” postoperative externally delivered therapies 4. Prolonging survival through cytoreduction 5. Applying locally-delivered therapies

5. 1. Providing a diagnosis Main differentials: stroke – DWI, vascular distribution; “infection” (i.e. encephalitis) – susceptibility, location; demyelinating disease – incomplete ring sign T1 with gad FLAIRT1 with gad post 6 wk

6. Providing a diagnosis Nonenhancing tumors are not always low-grade Chance of anaplasia increases in older patients (50% by mid-40’s) Barker et al. (UCSF), Cancer 1997

7. Providing a diagnosis Gliomas are notoriously heterogeneous More extensive resections more frequently provide higher grade diagnosis Glantz et al., Neurology, 1991 higher grade diagnosis more likely as extent of resection increased

8. Stereotactic bx: grade 2; resection grade 3 Providing a diagnosis

9. Perry et al., Cancer 1999 identification of oligo component in grade 3 tumors was more likely as extent of resection increased (p = 0.01) % containing oligo components: Bx3% STR29% GTR43%

10. Providing a diagnosis Aghi et al., unpublished MGH data (grade 2) % containing oligo components: Bx46% ( more specimens -> more oligo) GTR89% p < 0.001 Carter et al., unpublished SEER data (grade 2) % containing oligo components: Bx32% Resection62% p < 0.001

11. 100% Astro Year of diagnosis 19952000 Probability 0% 20% 40% 60% 19901985 Oligo 80% Oligoastro p < 0.001 SEER data; Carter BS et al unpublished Probability of oligo-containing histology has increased significantly with time

12. Probability of oligo-containing histology vs. literature on chemo for oligos 40% PubMed # articles on chemo for oligos Year of diagnosis 19952000 Probability of oligo diagnosis 0% 10% 20% 30% 19901985 SEER % oligos 20 10 0 Number of articles on chemo for oligos SEER, PubMed data; Carter BS et al unpublished

13. Providing a diagnosis Jackson et al., Neuro-Oncol 2001 81 pts referred to MD Anderson after stereo bx elsewhere who were accepted for resection >95% resection in 57% of patients 38% had different pathology even after central review of outside slides; 26% would affect treatment Mortality/major morbidity in 17%

14. Providing a diagnosis MR spectroscopy can assist in choosing a biopsy target

15. 2. Relieve mass effect Obvious and frequent success in most neurosurgeons’ experience in relieving neurological symptoms from mass effect Possible effect in increasing KPS in malignant glioma Low grade glioma: relieving medically intractable seizures

16. Relieve mass effect Ciric et al., Neurosurgery 1987 neurologic condition at discharge EORImproveSameWorse Partial0%60%40% Near GT30%70%0% GT41%55%4%

17. Relieve mass effect Sawaya et al., Neurosurgery 1998 30% of all patients had improved KPS after resection 8% had decreased KPS % “major neurologic complications”: Partial12% Subtotal15% Total7%

18. Relieve mass effect Fadul et al., Neurology 1988 Neurologic deterioration (1 week postop) Bx29% Partial39% Subtotal30% Total20% most hemorrhages and herniations occurred after bx or partial resection

19. 3. “Setting up” postoperative therapies Is response to postoperative adjuvant radiation in newly- diagnosed glioblastoma improved by prior resection? Neurosurgery 49:1288, 2001

20. Response to radiation after surgery 301 GM pts treated using two prospective UCSF clinical protocols age, KPS, extent of resection, radiation response recorded prospectively radiation response assessed by imaging criteria (postop image compared with post-XRT image)

21. Results More extensive surgical resection predicted better imaging-assessed response to postoperative adjuvant radiation in both univariate and multivariate analyses (adjusted for age, KPS)

22. “Setting up” postoperative therapies Resection and TMZ – EORTC 26981 (Stupp) 2-year survivalmedian survival +TMZ -TMZ +TMZ -TMZ GTR37% 14%18m 14m STR23% 9%14m 12m Bx10% 5% 9m 8m van den Bent et al., Eur J Cancer 2005 [abstr]

23. Subset Analysis – Overall Survival

24. “Setting up” postoperative therapies Keles et al. J Neurosurg 2004 119 pts with recurrent GM Reoperation -> TMZ Volume of disease at start of chemotherapy was a significant predictor of time to progression and survival (progression risk doubled for 10-15cc residual mass c/w GTR, quadrupled for >15 cc)

25. 4. Prolong survival Cushing believed that resection extended survival in malignant gliomas but recognized shorter and shorter intervals between operation as the disease progressed – “ideally all should be operative mortalities” McKenzie first to replace bone flap after resection to limit prognosis, but generally benefit of resection was not questioned

26. Review Article

27. 4. Prolong survival Nazzaro and Neuwelt, 1990 Reviewed neurosurgical literature 1940 - 1990 (184 refs) “This analysis shows that there is little justification for dogmatic statements concerning the relationship between increasing patient survival times and aggressive surgical management …”

28. Nazzaro and Neuwelt, 1990 Failure to adjust for other prognostic factors such as age Failure to adjust for difference in postop treatments* Failure to use “any form of statistical analysis” All studies had retrospective design Failure to adjust for resectability *logical fallacy – resection does influence chance of tolerating XRT as well as chance of reoperation

29. Prolong survival Considering the more than 30 years of experience and apparent failure, does it not seem that the ostensible myth of the benefit of cytoreduction for the “isolated sake of cytoreduction” needs to placed on the intellectual scrap heap? -- Michael L.J. Apuzzo

30. Prolong survival Many multivariate analyses of survival after resection of GM (nonrandomized) now provide evidence that extent of resection is an independent prognostic factor for survival (independent of age and KPS)

31. Prolong survival Laws et al., JNS 2003 (GO project) 788 patients (1997-2001) Resection was favorable prognostic factor (compared to biopsy) after correction for age, KPS, and after omission of patients with multifocal disease

32. GTR STR Bx Years after diagnosis Proportion surviving 0.00.51.01.52.0 0.0 0.2 0.4 0.6 0.8 1.0 p < 0.001 Barker et al. (UCSF), JNS 1996 Survival in GM stratified by extent of resection (nonrandomized)

33. Extent of MST1-yr2-yr5-yr resection (months) Gross total 1772%31%12% Subtotal 1147%15% 2% Biopsy 723% 2% 0% Barker et al. (UCSF), JNS 1996 Survival in GM stratified by extent of resection (nonrandomized)

34. Survival after biopsy or resection of supratentorial lobar glioblastoma: a population-based study Manish K. Aghi, M.D., Ph.D., William T. Curry Jr., M.D., Bob S. Carter, M.D., Ph.D. and Fred G. Barker II, M.D. Neurosurgical Service Massachusetts General Hospital Harvard Medical School Harvard Medical School

35. Results 11,134 glioblastoma patients diagnosed 1988 to 2001 11,108 intracranial tumors 7,423 supratentorial lobar tumors 6021 tumors did not cross the midline or tentorial notch, had not spread outside brain, through CSF or to spine, had no contraindications to surgery, and had a surgical procedure specified (size known for 3520 tumors)

36. Results Factors predicting biopsy over resection: Older age (odds ratio 1.38 per decade)Older age (odds ratio 1.38 per decade) Smaller tumor size (odds ratio 0.84 per cm)Smaller tumor size (odds ratio 0.84 per cm) Tumor location (parietal – highest chance of biopsy, temporal lowest)Tumor location (parietal – highest chance of biopsy, temporal lowest) Histology (glioblastoma – highest chance of biopsy, giant cell glioblastoma lowest)Histology (glioblastoma – highest chance of biopsy, giant cell glioblastoma lowest) Unmarried statusUnmarried status SEER registrySEER registry

37. Prob (biopsy) Tumor size (mm) Age Probability of biopsy rather than resection: Relation to age and tumor size

38. Results: survival 0.00.51.01.52.02.53.0 Years after diagnosis 0.0 0.2 0.4 0.6 0.8 1.0 Survival RESECTION BIOPSY Median survival – 12 months for resection 7 months for biopsy All pts had postop XRT

39. Results Biopsy rates vary by SEER registry: 27% 4% 23% 13% 15% 13% 12% 15% 15% 22% 22%

40. Resection Tumor size (mm) 406080 Probability of postop XRT 60% 65% 70% 90% 80% 75% 200 Biopsy 85% Probability of XRT vs. tumor size for biopsied & resected patients

41. Cancer 2006;106:1358 Survival Rates and Patterns

42. Prolong survival - subgroups Do all glioblastoma patients benefit equally from extensive resection? Potentially important subgroups: patients with mass effect younger patients

43. Mass effect and GM resection Kreth et al., Cancer 1999 Stereo bx+XRT c/w resection+XRT 225 patients, supratentorial GM Tumor resection effective only in group with midline shift (P < 0.01)

44. Graph

45. GTR STR Bx Years after diagnosis Proportion surviving Survival in GM stratified by extent of resection 0.51.01.52.0 0.0 0.2 0.4 0.6 0.8 1.0 Age 16 to 39 N = 47 p = 0.01 Age and GM resection

46. GTR STR Bx Years after diagnosis Proportion surviving 0.51.01.52.0 0.0 0.2 0.4 0.6 0.8 1.0 Age 40 to 49 N = 58 p = 0.01 Survival in GM stratified by extent of resection

47. GTR STR Bx Years after diagnosis Proportion surviving 0.51.01.52.0 0.0 0.2 0.4 0.6 0.8 1.0 Age 50 to 64 N = 114 p = 0.12 Survival in GM stratified by extent of resection

48. GTR STR Bx Years after diagnosis Proportion surviving 0.51.01.52.0 0.0 0.2 0.4 0.6 0.8 1.0 Age 65 to 79 N = 80 p = 0.04 Survival in GM stratified by extent of resection

49. Survival after GM resection: subgroups Subgroup Subgroup Hazard ratio Hazard ratio biopsy vs. resection Age ≤ 40 1.97 Age > 40 1.66 Size < 3 cm 1.54 Size 3-5 cm 1.72 Size > 5 cm 1.73 Source: Aghi et al, unpub (SEER)

50. Grade II astro / oligo / oligoastro Surgery Total N = 599

51. Nonrandomized trials Most trials in neurosurgery are not randomized Some topics can only be studied using nonrandomized methods (such as volume-outcome relationship) Recent studies have shown that when nonrandomized trials use concurrent controls eligible for the experimental treatment, results are often reasonably similar to randomized trial results

52. Solid: nonrandomized Graph

53. RCT of resection for glioma Design of RCT for resection in glioma Enrollment and randomization based on imaging diagnosis (sometimes bx is required in non-resected group) All tumors must be considered resectable Postop deaths and severe morbidities should not be excluded (i.e. no loss to followup based on results of surgery)

54. Loss to followup Of 28 malignant glioma trials included in NS review paper, 24 were either cooperative group trials (with explicit performance status criteria) or single-center trials (often with frank exclusion of poor results) 15/24 showed benefit of EOR – compared with 1 of 4 population based studies

55. Better Study Design #1 - RCT Only RCT of resection for malignant glioma: Vuorinen et al, Acta Nchir 2003 Age > 65, KPS > 60 30 pts randomized; 10/14 had resection (refusal, hematoma, lymphoma, infarct), 13/16 had biopsy (met x 2, hematoma) 9/10 resected pts, 10/13 biopsied pts started XRT (remainder: poor clinical condition)

56. Malignant glioma resection - survival Vuorinen, Acta Neurochir 2003 P = 0.035 Hazard ratio 2.7 Favors resection

57. Malignant glioma resection – time to failure Vuorinen, Acta Neurochir 2003 P = 0.057 Favors resection

58. RCT RCT of bx vs resection for elderly with presumed malignant glioma was supposed to open in France in 2007

59. Adjustment for resectability While many nonrandomized studies of EOR in glioma have included “adjustment for location”, none have included adjustment for resectability Barker et al., JNS 96

60. Resectability In fact, resectability is a complex concept that is a frequent subject of disagreement between surgeons Factors influencing rates of resection include pt-related (age, KPS, marital status); tumor- related (size, location, fuzziness of borders) and provider-related (specialist status, volume of practice, training and experience, economic and professional incentives) Resectable and nonresectable tumors may well have different molecular pathology

61. Publication

62. Resectability in glioma Two published attempts at defining a scale of resectability in gliomas Vorster and Barnett (Nsurg Focus 98) Eloquent / noneloquent Eloquent areas: sensorimotor, visual, language cortices, internal capsule, basal ganglia

63. Resectability in glioma Sawaya Nsurg 98

64. Resectability in glioma

65. Better study design – # 2 Nonrandomized study in which all tumors were eligible for resection, but heterogeneity of surgeons’ practices pseudorandomizes patients to bx/resection (or STR/GTR) Propensity score for balancing (as in Barker Nsurg 98)

66. An example – Shaw et al. RTOG 9802 – surgeon-determined GTR of LGG, then observed (single-arm phase II trial) 111 pts entered 1998-2002

67. RTOG 9802 PFS < 1 cm residual (59%)74% 1 – 2 cm residual (32%)32% > 2 cm residual (9%)11% 82% relapsed <2cm from resection cavity; 2% distant Shaw et al., JNS in revision 2008

68. Better Study Design #3 While it may be difficult to randomize pts between biopsy and resection, at least two RCTs of surgical adjuncts to improve EOR have been completed and reported For an effective adjunct (with no indempendent treatment effect) this would in effect randomize between different EORs Neuronavigation trial was negative (Willems JNS 104:360, 2006)

69. Prolonging survival Stummer et al. RCT of fluorescence-guided GM resection Pts for resection of presumed GM randomized to optimal white-light resection or fluorescence-guided (5-ALA) 5-ALA group had 65% GTR c/w 36% for WL group 6-mo PFS doubled in 5-ALA group (41% vs. 21%) Stummer et al., Lancet Oncol 2006

70. Malignant glioma resection - PFS Stummer et al., Lancet Oncol 2006 65% GTR 5-ALA vs 36% GTR WL

71. Prolonging survival Could an unexpected photodynamic treatment effect of 5-ALA at low light intensity account for results? Unpublished subgroup analyses – no difference in survival between groups after GTR; minimal improvement in PFS in STR group with 5-ALA Suggests benefits are a direct result of improved resection Stummer et al., Lancet Oncol 2006

72. Balance against risk Apples-oranges problem Common coin – KPS as prognostic factor for survival GTR/STR or STR/bx difference roughly worth ~10 to 30 KPS points

73. Table

74. Univariate P < 0.001 Age 304050607080 Probability 0.0 0.1 0.2 0.5 0.4 0.3 Barker et al., NeuroOnc 2005 Outcome vs. age s/p resection of primary brain tumor

75. Conclusions Surgical resection is an independent prognostic factor for survival in GM and LGG in nonrandomized trials Whether this would persist after adjustment for resectability has been controversial; single RCT now available with confirmatory results Burden of proof increasingly rests on nihilists; no likelihood of new LGG randomized trial anytime soon, new trial in MG is possibly in planning stages

76. Logo

77. Reoperation Who benefits from reoperation in glioblastoma?

78. Reoperation for GM: selection factors 223 patients with documented failure before death; multivariate logistic regression to predict patients who will undergo reoperation age< 40 34% p = 0.0240-6023% 60+11% extent of initial surgeryGTR 32% p = 0.02STR22% Bx4%

79. Reoperation for GM: relieving symptoms KPS after second resection: 28% improved 49% stable 23% declined (by 10-30 points) more likely benefit when recurrence was symptomatic

80. Reoperation for GM: comparison to unoperated patients Median survival after first failure (KPS known) reoperated (36 pts)41 wk not reoperated (136 pts)23 wk univariate hazard ratio 0.71p = 0.036 Cox mvhazard ratio 0.66p = 0.08 after adj for age, KPS at failure extent of initial resection & interval until failure: NS

81. Survival after first relapse Reoperation No reoperation

82. Reoperation for GM: survival benefit Cox simulation for “typical” patient 55 years old KPS at recurrence of 80 initial resection STR or GTR recurrence at/near original tumor site predicted difference in median survival between reoperated and nonreoperated groups: 8 weeks Barker et al. (UCSF), NS 1998

83. 4. Prolong survival Kowalczuk et al., Neurosurg 1997 75 patients with malignant gliomas significant: age, KPS, tumor grade, complications within 30 days after surgery, XRT dose administered not significant: 17 other factors, including extent of resection (volumetric analysis)