6660: Whole-Body MRI in the Evaluation of Pediatric Malignancies

1
6660 Whole-Body MRI in the Evaluation of
Pediatric Malignancies

Marilyn J. Siegel, MD (Principle
Investigator) Fredric Hoffer, MD Suddhasatta
Acharyya, PhD Brad Wyly, MD Berry Siegel, MD
Alison Friedmann, MD, MSc
2
ACRIN 6660 Whole-Body MRI in the Evaluation of
Pediatric Malignancies

• Biostatisticians
• Brad Snyder, MS
• Vincent Girardi, MS
• Lead Data Manager
• Jamie Downs

3
6660 Primary Aim

• Establish non-inferior diagnostic accuracy of
  whole body MRI compared with conventional imaging
  studies for detecting metastatic disease for use
  in staging of common pediatric tumors

4
ACRIN 6660 Study Overview

• Patients to undergo conventional studies
  including
• Scintigraphy (Bone, MIBG or gallium)
• Abdominal/Pelvic CT or MRI
• Chest CT (optional for neuroblastoma)
• FDG-PET (optional)
• Experimental Studies
• Whole-Body Fast MRI

5
Inclusion Criteria

• Male or female gender
• 21 years or younger
• Suspected soft tissue sarcoma, Ewings sarcoma
  family of tumors, neuroblastoma, Hodgkins
  disease, and non-Hodgkins lymphoma.
• Initial imaging completed in a timely fashion
• Final analysis only included above proven tumor
  types

6
Materials

• 192 patients enrolled
• 140 evaluable
• 51 with distant disease (or stage IV)
• 89 with lesser stage
• 70 selected for multi-reader study (35 with stage
  4 solid tumor or advanced stage lymphoma)
• 21 FDG PETs were included in conventional
  imaging of the reader study

7
Fast WBMRI Techniques

• Whole Body Imaging
• Vertex to toes
• Coronal plane images
• Body Coil
• Scans performed on a 1.5 T
• STIR

8
STIR MR-Ewing Sarcoma
9
Rhabdomyosarcoma
MRI
CT
Mass
Mass
Renal Metastasis
10
Central Image Interpretation

• 10 readers for pretreatment conventional CT/MRI,
  experimental WBMRI
• 10 readers for pretreatment scintigraphy
• FDG-PET, bone scans, MIBG
• Readers initially blinded to had washout period
  between conventional experimental imaging
• 10 pairs of readers of conventional scintigraphy
  cross sectional imaging had combined reports
• All pretreatment studies assessed for distant
  tumor extent

11
Proof of truth committee determined distant
disease (usually stage 4)

• 4 oncologists, 1 pediatric radiologist
• Conventional imaging findings from primary
  readers
• Bone marrow biopsy
• CSF aspirates when available
• Additional confirmatory imaging
• Additional confirmatory biopsy
• 6 months of data to determine initial stage

12
Determining the Primary Aim

• Reader study accuracy as area under ROC curve
  (AUC)
• Average AUC whole-body MRI (WBMRI) vs. average
  AUC conventional imaging
• To declare non-inferiority of WBMRI vs.
  conventional imaging for detecting distant
  disease, the expected 95 lower bound of the
  confidence interval for AUC (WBMRI minus
  conventional imaging) must be above 0.03 (closer
  to zero)

13
AUC for ROC analysis

• 70 cases (35-, 35)
• WBMRI
• 0.8291 empirical, 0.8436 parametric
• Conventional imaging
• 0.8676 empirical, 0.8896 parametric

14
Results for primary aim

• The difference in AUC between the modalities The
  95 confidence interval (CI) for WBMRI
  Conventional Imaging
• -0.0384 -0.1091, 0.0323 empirically
• -0.0461 -0.1195, 0.0274 parametrically
• The lower bound CI was not above -0.03
• WBMRI could not be declared non-inferior to
  conventional imaging

15
Search for reasons that WBMRI failed to achieve
non-inferiority

• WBMRI reader experience
• Lymphoma vs. other tumors
• Sensitivity
• Specificity
• Patient age
• Distant tumor size locations

16
WBMRI Reader experience Empirical AUC
MRI reader pairs (5 each) WBMRI AUC Convent-ional AUC WBMRI Conventional 95 CI
Experienced 0.8441 0.8895 -0.0455 (-0.1452, 0.0542)
Less experienced 0.8142 0.8456 -0.0314 (-0.1357, 0.0730)

• The experienced MRI readers did better on both
  WBMRI and conventional imaging readings.

17
AUC as per tumor type
Empirical AUC WBMRI Average AUC Conventional Average AUC WBMRI Conven-tional 95 CI
Lymphoma 31 0.7060 0.8177 -0.1117 (-0.2305, 0.0070)
Solid tumor 39 0.9116 0.9078 0.0038 (-0.0694, 0.0772)

• There was a trend toward non-inferiority for
  WBMRI to detect stage 4 solid tumors but the
  sample size was not sufficient for statistical
  significance.

18
Average sensitivity
Tumor type Imaging Estimate 95 CI
Lymphoma WBMRI 0.5156 (0.4158, 0.6142)
Conv. 0.6308 (0.5444, 0.7095)
Solid tumor WBMRI 0.8652 (0.8055, 0.9087)
Conv. 0.8864 (0.8370, 0.9222)

• The average sensitivity for advanced stage
  lymphoma was lower than stage 4 solid tumors for
  both WBMRI conventional imaging (plt0.0001).

19
Average specificity
Tumor type Imaging Estimate 95 CI
Lymphoma WBMRI 0.8323 (0.7523, 0.8902)
Conv. 0.8673 (0.7711, 0.9270)
Solid tumor WBMRI 0.8746 (0.8023, 0.9229)
Conv. 0.8588 (0.7578, 0.9220)

• Average specificity of WBMRI tended to be better
  than conventional imaging for solid tumors but
  not significantly

20
Age of patient vs. WBMRI reading
Age of patient of patients Average WBMRI Sensitivity Average WBMRI Specificity
Age lt 2 yrs 17 0.7673 0.8856
Age gt 2 yrs 53 0.7282 0.8378
Significance of age difference (p0.5331) (p0.2404)

• This trend failed to suggested that the red
  marrow of patients under age 2 was confused with
  bone marrow tumor on WBMRI.

21
Weighted averages of false negatives (FN) for
WBMRI reading (all missed lesions were lt1 cm)
FN location Lymphoma Solid tumor
Lung 52.38 36.67
Liver 12.69 33.33
Lymph node 7.94 43.33
Other 53.97 20.00
22
False negative (FN) cases missed by gt5 of 10
readers by WBMRI or conventional imaging among 35
positive cases
Advanced Stage Location FN Cases gt5 WBMRI readers FN Cases gt5 conventional
Lung 5 2
Pleura 1 0
Chest wall 2 2
Lymph nodes 2 1
Liver 2 2
Skeletal only 1 3
CSF by LP 2 1
23
False negative (FN) WBMRI conventional readings
in a subset of 13 of 35 cases missed by gt5
readers
Advanced Stage Location WBMRI FN reads Conventional FN reads
Lung 39 18
Pleura 6 1
Chest wall 16 14
Lymph nodes 14 12
Liver 17 17
Skeletal only 13 22
24
Conclusion

• ACRIN 6660 failed to demonstrate that WBMRI with
  STIR coronal imaging is not inferior to
  conventional imaging for determining metastatic
  pediatric malignancy
• WBMRI had more false negatives than conventional
  imaging due to lung metastases other lesions lt
  1 cm
• WBMRI trended to be as accurate more specific
  than conventional imaging for determining solid
  tumor metastases (but not for advanced lymphoma)

25
Image Gently

• www.imagegently.org

• CT is the major source of radiation in diagnostic
  radiology
• Children are more susceptible than adults to
  cancer after radiation exposure

26
Diffusion weighted Whole Body MRI

• Whole-body diffusion-weighted imaging for staging
  malignant lymphoma in children. Kwee TC, Takahara
  T, Vermoolen MA, Bierings MB, Mali WP,
  Nievelstein RA. Pediatr Radiol. 2010
  Oct40(10)1592-602.
• Whole-body MR imaging, bone diffusion imaging
  how and why? Jaramillo D. Pediatr Radiol. 2010
  Jun40(6)978-84.