Neuro-Oncology Advances 1

2(1), 1–10, 2020 | doi:10.1093/noajnl/vdaa026 | Advance Access date 10 March 2020

A simple score to estimate the likelihood of

pseudoprogression vs. recurrence following stereotactic
radiosurgery for brain metastases: The Bergen Criteria
  

Bente Sandvei Skeie , Per Øyvind Enger, Jonathan Knisely, Paal-Henning Pedersen,
Jan Ingeman Heggdal, Geir Egil Eide and Geir Olve Skeie

Downloaded from https://academic.oup.com/noa/article/2/1/vdaa026/5802330 by guest on 04 February 2022

Department of Neurosurgery, Haukeland University Hospital, Bergen, Norway (B.S.S., P.-H.P.); Department of
Neurosurgery, Stavanger University Hospital, Stavanger, Norway (P.Ø.E.); Department of Radiation Oncology,
Weill Cornell Medical College, New York, NY, USA (J.K.); Department of Oncology and Medical Physics, Haukeland
University Hospital, Bergen, Norway (J.I.H.); Department of Global Public Health and Primary Care, University of
Bergen, Norway (G.E.E.); Centre for Clinical Research, Haukeland University Hospital, Bergen, Norway (G.E.E.);
Department of Neurology, Haukeland University Hospital, Bergen, Norway (G.O.S.)

Corresponding Author: Bente Sandvei Skeie, MD, PhD, Department of Neurosurgery, Haukeland University Hospital, N-5021 Bergen,
Norway (bsai@helse-bergen.no).

Abstract
Background. A major challenge in the follow-up of patients treated with stereotactic radiosurgery (SRS) for brain
metastases (BM) is to distinguish pseudoprogression (PP) from tumor recurrence (TR). The aim of the study was to
develop a clinical risk assessment score.
Methods. Follow-up images of 87 of 97 consecutive patients treated with SRS for 348 BM were analyzed. Of these,
100 (28.7%) BM in 48 (53.9%) patients responded with either TR (n = 53, 15%) or PP (n = 47, 14%). Differences be-
tween the 2 groups were analyzed and used to develop a risk assessment score (the Bergen Criteria).
Results. Factors associated with a higher incidence of PP vs. TR were as follows: prior radiation with whole brain
radiotherapy or SRS (P = .001), target cover ratio ≥98% (P = .048), BM volume ≤2 cm3 (P = .054), and primary lung
cancer vs. other cancer types (P = .084). Based on the presence (0) or absence (1) of these 5 characteristics, the
Bergen Criteria was established. A total score <2 points was associated with 100% PP, 2 points with 57% PP and
43% TR, 3 points with 57% TR and 43% PP, whereas >3 points were associated with 84% TR and 16% PP, P < .001.
Conclusion. Based on 5 characteristics at the time of SRS the Bergen Criteria could robustly differentiate between
PP vs. TR following SRS. The score is user-friendly and provides a useful tool to guide the decision making whether
to retreat or observe at appropriate follow-up intervals.

Key Points
• Treatment data predict the risk of pseudoprogression vs. recurrence post-SRS.
• The Bergen Criteria robustly assesses the risk of pseudoprogression vs. recurrence.

A major challenge in the management of patients with brain further treatment is needed. PP on the other hand is a sign
metastases (BM) treated with stereotactic radiosurgery (SRS) of successful SRS due to radiation induced damage to the
is to distinguish radiation necrosis or pseudoprogression (PP) blood–brain barrier3 and influx of inflammatory cells.4,5 These
from tumor recurrence (TR) for which treatment and prog- changes lead to increased T1-contrast enhancement on mag-
nosis are different1,2 (Figure 1A). TR is due to failed SRS and netic resonance imaging (MRI) mimicking TR, but subside

© The Author(s) 2020. Published by Oxford University Press, the Society for Neuro-Oncology and the European Association of Neuro-Oncology.
This is an Open Access article distributed under the terms of the Creative Commons Attribution License (http://creativecommons.org/licenses/by/4.0/),
which permits unrestricted reuse, distribution, and reproduction in any medium, provided the original work is properly cited.
2 Skeie et al. The Bergen Criteria—pseudoprogression vs. recurrence

Importance of the Study

A key task in the follow-up of patients with recurring tumors. The present study is the
brain metastases (BM) treated with stereo- first to evaluate clinical data to predict the
tactic radiosurgery (SRS) is to differentiate risk of developing PP vs. recurrence. Known
pseudoprogression (PP) from recurrence, for predictors for successful SRS (small volume,
which treatment and prognosis are different. radiosensitive histology, prior radiation, and
Nearly 1/3 of BM treated with SRS will display optimal dosimetry) intuitively predict higher
increased contrast enhancement on standard risk of PP vs. recurrence. Based on the presence
magnetic resonance imaging; either tempo- or absence of these 4 predictors the Bergen
rarily due to successful SRS or progressively Criteria was established. It estimates with high

Downloaded from https://academic.oup.com/noa/article/2/1/vdaa026/5802330 by guest on 04 February 2022

due to failed SRS. Previous studies have fo- accuracy the likelihood of PP vs. recurrence, is
cused on sophisticated imaging techniques to user-friendly and cost-free. Further validation
distinguish radiation induced changes from of the Bergen Criteria is warranted.

spontaneously without change in treatment.6 Ambiguous iSRS. In the present study, we included all BM that either
MRI changes during follow-up are often managed by con- pseudoprogressed (47 BM) or recurred (53 BM), Figure 1B.
ducting an additional, early MRI follow-up to see if the Ethical approval was obtained from the Regional Ethical
changes stabilize or subside7,8 and/or to add additional im- Committee for Medical Research (REK, 2010/801) and
aging techniques9 to try to differentiate between the two. written informed consent was obtained from all patients.
MRI perfusion10,11 may show high intratumoral blood per-
fusion in TR and Positron Emission Tomography (PET)12–14
may show increased uptake of glucose or amino acids in SRS Treatment: Radiation Doses Used for Various
TR compared with PP, respectively, but differentiation can Sizes of Metastases
still be difficult. In some cases a biopsy15 or resection of
BM were treated with a prescribed dose of 20–25 Gy
the lesion will provide a definite diagnosis but involves
(n = 340), and 16–18 Gy if prior radiation, large total tumor
the risks associated with a surgical procedure. Moreover,
volume (>10–15 cm3) and/or close proximity to critical
a wait and see strategy may lead to delayed treatment in
structures (n = 50). For 16 (4%) BM the prescription dose
cases with TR, whereas immediate treatment may turn out
was ≤15 Gy due to prior SRS (n = 9), prior whole brain ra-
to be unnecessary if PP and even harmful if the lesion is
diotherapy (WBRT) (n = 3), large tumor volume (n = 2), or
reirradiated.16
tumor location (n = 2).
It is well documented that the chance of successful
SRS is higher if the BM are small,17,18 radiosensitive,19
previously irradiated20 and treated with optimal dosim-
Follow-up Schedule and Volume Measurements
etry.17,21 Thus, we wanted to investigate whether base-
of SRS Treated BM
line tumor- and treatment-related data that are readily
available in a clinical setting could predict the likelihood The BM volumes on MRI T1-contrast enhanced imaging (MRI-
of PP vs. TR (when follow-up MRI shows increased con- T1-C) were measured utilizing the GammaPlan software
trast enhancement) and whether these data could be in- (Elekta, Stockholm, Sweden) at GKS and during follow-up
tegrated into a clinical risk assessment score. To achieve at 1 and 3 months, then every third month post-GKS until
this, we investigated the predictive value of various September 30, 2019 or death. Follow-up images were available
baseline characteristics at SRS to assess the likelihood for 348 (85.7%) out of 406 BM in 87 (89.7%) of the 97 patients
of PP vs. TR. (95.6% of the 91 patients alive at first MRI 1 month post-SRS).
Two patients (2.1%) are still alive in September 2019.

Methods Volumetric Tumor Response—Definition of PP

vs. TR
Volumetric tumor change was a secondary end point in a
prospective study evaluating quality of life changes fol- The volume changes over time on MRI-T1-C from SRS until
lowing Gamma Knife Surgery (GKS) for BM.22 Treatment LF leading to new treatment (surgery, WBRT, or re-SRS) or
and follow-up MRI of 97 consecutive patients treated for a until last follow-up/death was used to define PP from TR.
total of 406 BM with Gamma Knife Perfexion at Haukeland Local failures that were retreated with a second course of
University Hospital between 2009 and 2011 were pro- SRS were evaluated as separate lesions/treatments and
spectively collected and analyzed. Two hundred and volume changes from re-SRS were recorded.
twenty-five BM were treated initially (iSRS), 160 BM at sub- All BM followed 1 of 4 principal response patterns: ei-
sequent GKS (sSRS) as distant failures, and 21 BM were ther continuous reduction in size (n = 238, 68%), transient
treated with repeat-SRS due local failure (LF) following increase in volume defined as PP (n = 47, 14%), delayed
 Skeie et al. The Bergen Criteria—pseudoprogression vs. recurrence 3

Advances
Neuro-Oncology
  
A B

Tumor recurrence Pseudoprogression

60

40
Recurrence

Tumor volume (cm3)

Tumor volume (cm3) on MRI

20
8
0
T1 w/contrast

6
60
4

2 40
Pseudoprogression

Downloaded from https://academic.oup.com/noa/article/2/1/vdaa026/5802330 by guest on 04 February 2022

20
0
0 1 3 6 9 12 15 18 21 0
Time following SRS (months) 0 1 3 6 9 12 15 18 21 24
Time following SRS (months)

C D

Continuous tumor volume

reduction n = 37 (42.5 %) Prior WBRT: n = 4 (19.0 %)
Lung cancer: n = 9 (42.9 %)
Tumor Recurrence
21 patients with
Tumor Recurrence 28 BM
Prior WBRT: n = 4 (26.7 %)
(Delayed tumor growth) Lung cancer: n = 5 (33.3 %)
n = 21 (24.2 %)
Tumor Recurrence and Tumor Recurrence and
Pseudoprogression Pseudoprogression
n = 15 (17.2 %) 15 patients with
50 BM

Continuous
tumor growth Pseudoprogression
n = 2 (2.3 %) Pseudoprogression 12 patients with
(Temporary increase in tumor volume) 22 BM Prior WBRT: n = 4 (33.3 %)
No MRI follow-up, n = 10 n = 12 (13.8 %) Lung cancer: n = 7 (58.3 %)

Figure 1 (A) Illustration of 2 potential response patterns for brain metastases (BM) treated with stereotactic radiosurgery (SRS):
pseudoprogression (PP) and tumor recurrence (TR). (B) Individual tumor volume response curves on contrast enhanced T1-weighted MRI at ster-
eotactic radiosurgery (time: 0) and during follow-up for (A) 53 pseudoprogressing tumors and (B) 47 recurring tumors. (C) Venn diagram of 87 out of
97 patients with follow-up MRI of a total of 348 BM post-radiosurgery. The diagram illustrates the proportion of patients with BM responding with
the 4 distinct volumetric response patterns on contrast enhanced MRI: tumor recurrence (red), pseudoprogression (green), continuous tumor
volume reduction (blue), and continuous tumor growth (orange). (D) Venn diagram of patients (n = 48) included in the development of the Bergen
Criteria. Tumor type (lung cancer vs. other primary sites) and prior irradiation history for patients in 3 cohorts are illustrated: 21 (43.8%) patients
with 28 recurrences (red), 12 (25.0%) patients with 22 pseudoprogressions (green), and 15 (31.2%) patients with some recurrences (n = 25 BM) and
some pseudoprogressions (n = 25 BM) following initial and/or repeat-SRS. MRI, magnetic resonance imaging.
  

growth defined as TR (n = 53, 15%), or no response with due to TR. The mean age of the 48 included patients was
continuous growth (n = 10, 3%) (B.S.S. et al., unpublished 60 years (range 39–86 years) and 22 (45.8%) were males.
manuscript). Only 2 (2%) of the BM were resected after
SRS yielding biopsy material. Figure 1C illustrates the dis- Tumor Subtypes
tribution of the different volumetric response patterns for
all 87 patients (348 BM) with follow-up MRI. The primary cancer site was lung (n = 21 patients [43.8%]),
melanoma (n = 7 [14.6%]), renal (n = 5 [10.4%]), breast
(n = 4 [8.3%]), colorectal cancer (CRC, n = 9 [18.8%]), or un-
Patient Population Used to Develop a Risk known (n = 2 [4.2%]).
Assessment Score for the Likelihood of PP vs. TR
We included all 48 (55.1%) out of the 87 patients with fol- Prior Whole Brain Radiotherapy and
low-up images of at least 1 BM that either pseudoprogressed Immunotherapy
(47 BM) and/or recurred (53 BM), Figure 1D. Thus, 100
(28.7%) out of 348 BM were included of which 86 were Twelve (25%) of the 48 patients that had previously been
treated with iSRS and 14 of these with subsequent re-SRS irradiated with WBRT. Thirty of the included BM (30%)
4 Skeie et al. The Bergen Criteria—pseudoprogression vs. recurrence

occurred in the 12 patients with prior WBRT. None of the used Pearson’s χ 2 test to refine the score: prior SRS (yes
patients in this cohort were treated with immune check- or no), prior WBRT (yes or no), BM volume ≤2 cm3 (yes
point inhibitors at the time of SRS. or no), target cover ratio ≥98% (yes or no), primary lung
cancer (yes or no), and to assess the predictive value of
the Bergen Criteria. Overall survival (OS) was analyzed by
Repeat-SRS the Kaplan–Meier method. Date for end-of-follow-up was
September 30, 2019.
Eleven (22.9%) of the 48 patients underwent iSRS (n = 14
BM) and later re-SRS (n = 14 BM) for the same BM due
to recurrence. Post-re-SRS 12 of the 14 BM subsequently
pseudoprogressed and 2 re-recurred. The remaining 37 Results
(77.1%) patients only underwent iSRS for 72 BM; 12 pa-
tients had at least 1 BM that pseudoprogressed, 20 patients The BM in the 4 response groups described above dif-

Downloaded from https://academic.oup.com/noa/article/2/1/vdaa026/5802330 by guest on 04 February 2022

had at least 1 BM that recurred, and 5 patients had a mix- fered significantly in volume, prescribed dose, and target
ture of PP and TR. cover ratio at SRS (Table 1). There was a striking simi-
larity between BM with favorable outcomes (continuous
tumor volume reduction or PP) and BM with unfavorable
The Bergen Criteria outcomes (continuous tumor growth or TR). BM with a
favorable outcome were generally small (mean volume
All 47 BM responding with PP were compared with the 53
≤2.3 cm3) and treated with high target cover ratio (mean
BM responding with TR. Baseline tumor- and treatment-
target cover ratio ≥ 98%) at SRS. On the other hand, BM
related characteristics found to be associated with devel-
with an unfavorable outcome were large (mean volume
opment of PP compared with TR were used to design a risk
≥3.5 cm3) and less optimally covered at SRS (mean target
assessment score to estimate the likelihood of PP vs. TR
cover ratio <98%). We found significant differences when
(the Bergen Criteria). The score was refined by choosing
comparing BM in the PP group with the TR group: First,
predictive factor splits for maximizing practical use.
they were significantly smaller (mean BM volume 2.3 vs.
4.8 cm3, P = .015) and more completely covered with the
prescribed dose (mean target cover ratio 98.6% vs. 97.6%,
Statistics P = .026). These differences at baseline were used to pre-
dict the likelihood that increased MRI-T1-C enhancement
To analyze differences between BM responding with PP
following SRS was due to PP or recurrence.
vs. recurring BM we used Student’s t-test for continuous
variables (baseline BM volume, prescription dose, max-
imum dose, target cover ratio, 12 Gy normal brain volume, Baseline Tumor Volume as Predictor for PP vs. TR
total steroid dose, and time from SRS until increase in
BM volume on MRI-T1-C) and Pearson’s χ 2 test for catego- For practical reasons we used 2 cm3 as cutoff. It corres-
rical variables (primary cancer site, prior treatment with ponds to a tumor diameter of 1.5 cm that can be easily
SRS and/or WBRT, and steroid medication). For variables assessed on MRI. We found that BM less than 2 cm3 were
found to be significantly associated with PP vs. TR we more likely to develop PP than larger BM. Of the 47 BM

  
Table 1 Baseline Characteristics for Brain Metastases (BM) According to 4 Principle Tumor Volume Response Curves on Contrast Enhanced
T1-Weighted-MRI Following Stereotactic Radiosurgery (SRS) in 348 of 406 BM With Follow-up MRI (in 87 of 97 Consecutive Patients) at Haukeland
University Hospital in Bergen (Norway) Between 2009 and 2011

Characteristic Response Curve Group n (%) Mean P Value

Tumor volume at SRS Continuous tumor volume reduction 238 (68) 1.4 <.001
(cm3)
Temporary increase in tumor volume (PP) 53 (15) 2.3
Delayed growth (TR) 47 (14) 4.8
Continuous tumor growth 10 (3) 3.5
Prescription dose at SRS Continuous tumor volume reduction 238 (68) 20.5 <.001
(Gy)
Temporary increase in tumor volume (PP) 53 (15) 20.2
Delayed growth (TR) 47 (14) 19.3
Continuous tumor growth 10 (3) 15.9
Target cover ratio at SRS Continuous tumor volume reduction 238 (68) 98.9 <.001
(%)
Temporary increase in tumor volume (PP) 53 (15) 98.6
Delayed growth (TR) 47 (14) 97.6
Continuous tumor growth 10 (3) 96.0

MRI, magnetic resonance imaging; PP, pseudo-progression; TR, tumor recurrence.

  
 Skeie et al. The Bergen Criteria—pseudoprogression vs. recurrence 5

Advances
Neuro-Oncology
displaying PP, 32 (68.1%) vs. 15 (31.9%) were ≤2 cm3, or less The Bergen Criteria for Assessing the Risk of PP
than 1.5 cm in diameter vs. >2 cm3 (P = .054, χ 2 = 3.7, df = 1, vs. Recurrence
and Cramer’s V: 0.192). Similar numbers for recurring BM
were 26 (49.1%) vs. 27 (50.9%), Figure 2A. Factors associated with a higher incidence of PP vs.TR used
were as follows (1) prior SRS, (2) prior WBRT, (3) target
cover ratio ≥98%, (4) BM volume ≤2 cm3, and (5) primary
Target Cover Ratio at SRS as Predictor for PP lung cancer vs. other primaries. Based on the presence
vs. TR (0) or not (1) of these 5 parameters, the Bergen Criteria for
risk assessment of PP vs. TR was established (Table 2). The
BM responding with PP had significantly more often a total score ranges from 0 to 5 points. A score of 0 corres-
higher coverage with the prescribed dose at SRS com- ponds with high risk of PP vs. TR, whereas a score of 5 cor-
pared with recurring BM. Using 98% as cutoff we found responds with a low risk of PP vs. TR. For simplicity, the
a significantly higher incidence of PP vs. recurrence if the predictive value of 4 Bergen Criteria groups, total score <2,

Downloaded from https://academic.oup.com/noa/article/2/1/vdaa026/5802330 by guest on 04 February 2022

target cover ratio was ≥98%. Of 46 BM displaying PP, 37 2, 3, and >3 points, respectively, was calculated (Table 3).
(78.7%) vs. 10 (21.3%) had a target cover ratio of ≥98% A score of <2 point was associated with a 100% likeli-
vs. <98% and for recurring BM 32 (60.4%) vs. 21 (39.6%), hood of PP, 2 points with 57% risk of PP vs. a 43% risk of TR,
P = .048, χ 2 = 3.9, df = 1, and Cramer’s V: 0.198, Figure 2B. 3 points with a 57% risk of TR vs. 43% risk of PP whereas >3
points were associated with 84% likelihood for TR vs. 16%
for PP (χ 2 = 24.6, df = 3, P < .001, and Cramer’s V = 0.496).
Prior Radiation Treatment as Predictor for PP
All BM with Bergen Criteria score >3 originating from
vs. TR
primary CRC (n = 6), lung cancer (n = 3), or unknown pri-
Prior IR was a strong predictor of PP vs. TR (P = .001, mary (n = 2) belonged to the TR group vs. 71% (n = 10) of
χ 2 = 13.7, df = 2, and Cramer’s V: 0.370). In the PP group, 10 the 15 BM origination from other primaries (melanoma,
BM (21.3%) were previously treated with SRS, 17 (36.2%) renal, and breast cancer). All 7 BM with Bergen Criteria >3
with WBRT whereas 20 (42.6%) were not previously irradi- and contrast enhancement occurring >6 months after SRS
ated. On the contrary, in the TR group, only 1 BM (1.9%) belonged to the TR group vs. 14 of the 18 BM occurring
was previously treated with SRS, 13 (24.5%) with WBRT, ≤6 months after SRS.
and 39 (73.6%) were not previously irradiated, Figure 2C. The Bergen Criteria is applicable also when excluding
the 13 retreated BM. For the 87 BM that were not retreated
Bergen Criteria 0–1 were associated with 100% PP and
Primary Cancer Site as Predictor for PP vs. TR Bergen Criteria of 2, 3, and 4–5 with 52%, 42%, and 12% PP,
respectively, P < .001).
We found that BM in the PP group more often originated The mean scores for the Bergen Criteria for the 4 re-
from primary lung cancer than from other primaries com- sponse patterns, continuous reduction in size, PP, delayed
pared with BM in the TR group (P = .084, χ 2 = 3.0, df = 1, and growth (TR), and continuous growth, are 2.7 (SD: 0.72,
Cramer’s V: 0.173). The primary cancer site was lung for 24 range 1–4), 2.3 (SD: 0.95, range: 1–4), 3.2 (SD: 0.78, range
(57.1%) of the 47 BM in the PP group and 18 (34.0%) of the 2–4), and 3.1 (SD: 0.74, range 2–4), respectively (χ 2: 20.5, df:
53 BM in the TR group, Figure 2D. 3, P < .001).

Time Frame for Detection of PP vs. TR Treatment at PP and TR Following SRS

The mean time from SRS until increased contrast en- None of the BM responding with PP were resected, 43
hancement on MRI was detected was significantly shorter (91.5)% were asymptomatic, and 4 (8.5%) were managed
for PP than TR, 5.1 months (SD 4.1, range 1–18 months) with a short course of dexamethasone treatment.
vs. 8.4 months (SD 6.5, range 3–36 months), respectively, Twenty-nine (52.7%) out of 53 TRs were treated conserv-
P = .003. The mean time for the contrast enhancement to atively with observation, 18 (32.7%) were retreated with
subside in PP was 3.2 months (SD 1.4, range 1–6 months). SRS, 2 (3.6%) were resected with biopsy confirmed TR, and
PP occurred within 6 months of SRS for 37 (78.7%) of 47 4 (7.3%) were treated with WBRT.
BM and more than 6 months post-SRS for 10 (21.3%).
Similar numbers for TR are 31 (58.5%) and 22 (41.5%) of 53
BM, respectively (P = .030, χ 2 = 4.7, df = 1, and Cramer’s V: Overall Survival
0.216).
The OS for patients with TR, PP, or both was significantly
longer than for the 49 patients with MRI follow-up that ex-
Steroid Treatment as Predictor for PP vs. TR perience neither TR nor PP, P < .001. The median OS for the
last group was only 4.5 months (95% CI 3.8–5.1) compared
Neither the use of steroids at SRS (yes vs. no; P = .897) nor with 11.7 months (95% CI 6.6–16.8) for the 21 TR only pa-
the steroid dose used at baseline (categorical variables: 0 tients, 10.9 months (95% CI 3.9–17.9) for the 12 PP only pa-
vs. 1–4 vs. >4 mg dexamethasone; P = .748/continuous var- tients, and 15.4 months (95% CI 12.4–18.5) for the mixed
iable: total steroid dose; P = .930) was significantly associ- response group, respectively. The OS for the 10 patients
ated with development of PP (mean 4.0 mg, SD 6.0, range without MRI follow-up was 0.83 months (95% CI 0.4–1.2),
0–16 mg) vs. TR (mean 5.0 mg, SD 5.2, range 0–16 mg). Figure 2E.
6 Skeie et al. The Bergen Criteria—pseudoprogression vs. recurrence

  
A Baseline
B Target
40 Tumor 40 Cover
volume ratio
≤ 2 cm3 > 98%
Number of brain metastases

Number of brain metastases

> 2 cm3 ≤ 98 %
30 30
P = .039 P = .031

20 20

Downloaded from https://academic.oup.com/noa/article/2/1/vdaa026/5802330 by guest on 04 February 2022

10 10

0 0
Tumor recurrence Pseudoprogression Tumor recurrence Pseudoprogression

C Prior D Primary
radiation cancer
40 40
therapy site
Prior SRS Lung
Number of brain metastases

Number of brain metastases

Prior WBRT ± Other
30 30
SRS
P = .084
None

20 20
P = .001

10 10

0 0
Tumor recurrence Pseudoprogression Tumor recurrence Pseudoprogression

E
1.0
At least 1 TR post-SRS
At least 1 PP post-SRS
0.8 At least 1 TR and 1 PP
post-SRS
No TR nor PP post-SRS
Overall Survival

No MRI post-SRS
0.6 ...

0.4 P < .001

0.2

0.0

0 20 40 60 80
Time following SRS (months)

Figure 2 The number of brain metastases (BM) that responded to stereotactic radiosurgery (SRS) with pseudoprogression or tumor recurrence
with: (A) Baseline volume ≤2 cm3 vs. >2 cm3. (B) Target cover ratio at SRS <98% vs. ≥98%. (C) Prior treatment with SRS, whole brain radiotherapy ±
SRS vs. no prior radiation treatment. (D) Primary lung cancer origin vs. other origin than lung cancer. (E) Overall survival curves for patients with
(a) at least 1 BM that recurred (TR) (red), (b) at least 1 BM that pseudoprogressed (PP) (green), (c) a mixture of TR and PP (violet), (d) no TR nor PP
(blue), and (e) no follow-up images (black).
  
 Skeie et al. The Bergen Criteria—pseudoprogression vs. recurrence 7

Advances
Neuro-Oncology
than 3, 84% belonged to the TR group. The Bergen Criteria
is based on easily accessible tumor- and treatment-related
Discussion characteristics available in the patient’s medical record and
Differentiating TR from PP is a daily clinical challenge and does not require special training or additional advanced
of utmost importance to avoid both unnecessary treat- imaging techniques. Furthermore, the Bergen Criteria is
ments and treatment delays. More than 50% of the pa- based purely on known predictors2 for successful SRS
tients in our material experienced either TR or PP of at least and is therefore intuitive. We know when we treat a large
one of their BM, which is in line with previous studies.23 radioresistant brain metastasis with an incomplete target
One could argue that PP is due to successful or even cover ratio that the chance of success is lower than if we
“overtreatment” with SRS while TR reflects failed SRS. treat a small, previously irradiated radiosensitive brain me-
Utilizing treatment data is a novel way of distinguishing PP tastasis with optimal target cover ratio. The Bergen Criteria
from TR. From our data we found 4 distinct baseline char- works as follows. If a brain metastasis ≤2 cm3 originating

Downloaded from https://academic.oup.com/noa/article/2/1/vdaa026/5802330 by guest on 04 February 2022

acteristics at SRS distinguishing BM responding with PP from lung cancer that was irradiated with WBRT prior to
from recurring BM. They were more often previously ir- SRS starts to enlarge on MRI-T1-C, we can be 100% certain
radiated, significantly better covered with the prescribed that this is due to PP if the brain metastasis was completely
dose at SRS, smaller and more often originating from pri- covered with the prescribed dose (Bergen Criteria score 0).
mary lung cancer. Using this information we were able Similarly, if a brain metastasis >2 cm3 at SRS originating
to create a simple score which robustly distinguished PP from melanoma and not previously irradiated starts to
and TR in a large number of patients. All BM in our study enlarge following SRS, we can be 84% confident that the
responding either with PP or TR with a score less than 2 BM are recurring if the BM were not fully covered at SRS
belonged to the PP group. For the BM with a score more (Bergen Criteria score 5). The likelihood might further be
estimated to 100% if the primary cancer is CRC, unknown
or lung or the contrast enhancement is detected more than
6 months post-SRS, but this needs to be confirmed in a
  
Table 2 The Pseudoprogression (PP) Risk Assessment Score (the larger cohort of patients.
Bergen Criteria), With Total Range: 0 (Low Risk of PP) to 5 (High Risk of The Bergen Criteria can classify all BM enlarging fol-
PP), Is Defined as the Sum of the Scores on 5 Baseline Characteristics lowing SRS either as most likely due to PP or most likely
Based on 348 of 406 BM With Follow-up MRI (in 87 of 97 Consecutive due to TR. Depending on the total score the risk of TR varies
Patients) at Haukeland University Hospital in Bergen (Norway)
Between 2009 and 2011 from 0% to 84% and the risk of PP from 16% to 100%. If TR is
most likely it may be an advantage to confirm whether an
Baseline Characteristics Bergen enlarging tumor on MRI-T1-C follow-up is due to PP or TR
Criteria with additional imaging such as the T1/T2 mismatch8,24 be-
Scores fore initiating treatment. If the tumor size is comparable on
0 1 contrast enhanced T1- and T2-weighted MRI images, this is
Primary lung cancer Yes No most likely to represent the true tumor size and hence TR.
3
BM volume ≤2 cm (≤1.5 cm in diameter) Yes No
If however, the BM are smaller on the T2- than T1-contrast
series, the difference in size on the T1 vs. T2 is most likely
Target cover ratio >98% Yes No
due to changes in the normal brain–blood circulation sur-
Prior SRS to the same BM Yes No rounding the tumor and hence reflect PP. MRI perfusion
Prior WBRT Yes No and PET may be of value as reduced blood flow and uptake
of glucose or amino acids strengthens the suspicion of PP
BM, brain metastases; MRI, magnetic resonance imaging; SRS, stere- vs. TR. Advanced imaging methods and biopsies evaluate
otactic surgery; WBRT, whole brain radiotherapy.
the tumor at the time of progression or PP. However, the
   Bergen Criteria can estimate the chance of success based

  
Table 3 The Likelihood of Pseudoprogression (PP) vs. Tumor Recurrence (TR) as a Cause of Tumor Enlargement on Contrast Enhanced
T1-Weighted-MRI Following Stereotactic Radiosurgery (SRS) (n = 100 Brain Metastases in 48 Patients) According to the Bergen Criteria at
Haukeland University Hospital in Bergen (Norway) Between 2009 and 2011

The Bergen Criteria Total BM With PP or TR Likelihood of PP (%) Likelihood of TR (%)

(N = 100) (n = 47) (n = 53)
N (%) % n (% of PPs) % n (% of TRs)
0–1 12 (12.0) 100 12 (25.5) 0 0 (0.0)
2 28 (28.0) 57 16 (34.0) 43 12 (22.6)
3 35 (35.0) 43 15 (31.9) 57 20 (37.7)
4–5 25 (25.0) 16 4 (8.5) 84 21 (39.6)

BM, brain metastases; MRI, magnetic resonance imaging. The Bergen Criteria: pseudoprogression risk assessment score, ie, the sum of 5 baseline
characteristics (Table 2).
  
8 Skeie et al. The Bergen Criteria—pseudoprogression vs. recurrence

on treatment-related factors alone and thus adds value to Tumor size is the main limitation for SRS. Smaller BM
the existing methods by combining information at the time respond better to SRS and therefore also more often re-
of SRS with information at the time the BM volume starts spond with PP than larger BM. However, BM from different
to increase on MRI-T1-C post-SRS. primaries represent different diseases. Lung cancer BM
PP is considered a good prognostic sign.23 PP is often are radiosensitive and respond well to SRS with a corre-
asymptomatic (>90% in our study) and symptoms usually sponding high rate of PP compared with BM from other
respond to treatment with a short course of steroids. In primary cancer sites. Moreover, lung cancer patients have
rare cases, surgery may be necessary to relieve symptoms. a high-risk developing BM and are thus routinely screened
Nevertheless, surgery has an inherent risk of complica- for BM. Consequently, BM in lung cancer patients are likely
tions, thus a wait and see strategy is preferred if one sus- to be detected when they are smaller compared with BM in
pects PP.2 We suggest that the Bergen Criteria may be used patients with more radioresistant primaries.
to tailor appropriate follow-up intervals. If a BM enlarges The significantly longer survival observed for patients ex-

Downloaded from https://academic.oup.com/noa/article/2/1/vdaa026/5802330 by guest on 04 February 2022

on MRI-T1-C following SRS with a Bergen Criteria score periencing TR, PP, or both compared with patients that do not
<2, the patient may continue with standard 3 months fol- suggests that the Bergen Criteria may be valid also for long-
low-up intervals due to a high likelihood of PP. Conversely, term survivors. However, as only 2 of 97 patients have more
a Bergen Criteria score >3 calls for an early MRI follow-up than 5 years OS in the present study, longer-term data on large
or immediate retreatment. If the score is 2 or 3, the patient datasets will need to be scrutinized to see if the Bergen Criteria
may be followed at intermediate intervals (6–8 weeks) or are as infallible in patients who are in long-term follow-up as
the patient should be offered additional imaging with per- they appear to be in the short-term follow-up cohorts.
fusion MRI or PET. The strength of the study is the prospective design, the
Radiation is the main cause of PP in BM patients. For a national referral area and the high compliance rate with
brain metastasis to achieve a Bergen Criteria score <2 it complete MRI follow-up until death for more than 95%
must have been previously irradiated either with SRS, of the patients. Still, the relatively low number of BM in-
WBRT, or both. There are few reports on PP following cluded is a limitation. Secondly, the Bergen Criteria is de-
WBRT alone. PP seems to occur when WBRT is combined rived in a cohort treated with SRS before the introduction
with SRS25 and when high-dose SRS is used alone or re- of immunotherapy. This is a potential shortcoming that re-
peated for the same brain metastasis. Thus, the total ac- quires cautioning for those who might employ the Bergen
cumulated dose to the tumor and normal brain is clearly Criteria in her/his own clinic to predict the future for indi-
related to the development of PP. In the present study, the vidual patients. Thirdly, temporary volume increase on
likelihood of PP increased if the brain metastasis was com- MRI-T1-C was used as definition for PP in the present study
pletely covered with the prescribed dose independent of and delayed growth as definition for TR; without histolog-
the dose itself. This implies that the minimum dose to the ical confirmation except for 2 cases with TR. Nevertheless,
tumor margins, rather than the maximum (or prescribed) the volumetric curves used represent the final outcome for
dose, may be the most important dosimetric factor for all BM as all patients are followed for more than 5 years or
successful SRS and thereby risk of TR vs. PP. Radiation until death. Importantly, none of the BM responding with
damage to the tumor induces an inflammatory response temporary volume increase (PP) recurred at a later stage.
which seems to play a role in the development of PP. The Finally, in the present study, BM are used as the observa-
more complete the target cover ratio is at SRS, the more tion unit and thus assessed as independent of the patient
tumor cell death is induced by radiation which again might even though some patients had more than 1 BM and a
induce a stronger antitumor immune response than if the dependency between BM in the same patient therefore is
target coverage ratio is lower. likely. Independent assessment of observations of BM in
PP is often observed in patients treated with immuno- the same patient may lead to excessive statistical signif-
therapy.5 Of note, none of the patients in the cohort used icance. Moreover, a prognostic index will always be ad-
to derive the Bergen Criteria were treated with immuno- justed to the material from which it has been developed
therapy. With their increasingly common use in a wide va- with an excessively good predictability.
riety of malignancies, there may be some differences in the The Bergen Criteria is user-friendly, intuitive, and cost-
PP and TR percentages, depending on whether or not these free. It estimates with high accuracy the likelihood of
drugs are being employed as systemic therapy. A retro- PP, a sign of successful treatment, vs. the risk of failed
spective study showed that immunotherapy in combination treatment based on 4 simple and readily available treat-
with SRS increased the local control rate and risk of radia- ment characteristics. The Bergen Criteria may be of help
tion necrosis (RN) when immunotherapy was administered when deciding whether to retreat or observe, informing
during or after SRS compared with when it was adminis- patient and choosing an appropriate follow-up interval.
tered before SRS.26 Recent reviews found inconclusive dif- Ultimately, it may lead to reduced use of expensive PET/
ferences in LC rates and risk of RN when they compared specialized MRI sequences and reduce the number of
combined SRS and immunotherapy with SRS alone.27,28 treatment delays and unnecessary surgeries. Larger
Nevertheless, due to the potential synergistic effect of SRS studies are however needed to validate our results.
in combination with checkpoint inhibitors, the predictive The Bergen Criteria may be extended to additional data
value of the Bergen Criteria needs to be confirmed in a co- sources, which may lead to a refined score for subtypes of
hort receiving a combination of these 2 treatments. BM from different primary cancers.
 Skeie et al. The Bergen Criteria—pseudoprogression vs. recurrence 9

Advances
Neuro-Oncology
7. Kano H, Kondziolka D, Lobato-Polo J, Zorro O, Flickinger JC, Lunsford LD. T1/
T2 matching to differentiate tumor growth from radiation effects after stere-
Keywords otactic radiosurgery. Neurosurgery. 2010;66(3):486–491; discussion 491.
brain metastases | pseudoprogression | radiation necrosis | 8. Dequesada IM, Quisling RG, Yachnis A, Friedman WA. Can standard
stereotactic radiosurgery | tumor recurrence. magnetic resonance imaging reliably distinguish recurrent tumor from ra-
diation necrosis after radiosurgery for brain metastases? A radiographic-
pathological study. Neurosurgery. 2008;63(5):898–903; discussion 904.
9. Delgado-López PD, Riñones-Mena E, Corrales-García EM. Treatment-
related changes in glioblastoma: a review on the controversies in re-
Funding sponse assessment criteria and the concepts of true progression,
pseudoprogression, pseudoresponse and radionecrosis. Clin Transl
This study was supported by postdoctoral grant from Helse
Oncol. 2018;20(8):939–953.
Vest (RHF) Regionalt samarbeidsorgan (Project number:
10.

Downloaded from https://academic.oup.com/noa/article/2/1/vdaa026/5802330 by guest on 04 February 2022

Hatzoglou V, Yang TJ, Omuro A, et al. A prospective trial of dynamic
912042; Project title: Gamma Knife Surgery for Brain Cancer –
contrast-enhanced MRI perfusion and fluorine-18 FDG PET-CT in
Radiosensitizers and Imaging techniques to improve treatment
differentiating brain tumor progression from radiation injury after cranial
Efficacy - Experimental and Clinical research.).
irradiation. Neuro Oncol. 2016;18(6):873–880.
11. Muto M, Frauenfelder G, Senese R, et al. Dynamic susceptibility contrast
(DSC) perfusion MRI in differential diagnosis between radionecrosis and
neoangiogenesis in cerebral metastases using rCBV, rCBF and K2. Radiol
Conflict of interest statement. None declared. Med. 2018;123(7):545–552.
12. Heinzel A, Müller D, Yekta-Michael SS, et al. O-(2-18F-fluoroethyl)-l-
tyrosine PET for evaluation of brain metastasis recurrence after radio-
therapy: an effectiveness and cost-effectiveness analysis. Neuro Oncol.
Previous presentations 2017;19(9):1271–1278.
13. Tsuyuguchi N, Sunada I, Iwai Y, et al. Methionine positron emission to-
Portions of this work were presented in abstract form at the mography of recurrent metastatic brain tumor and radiation necrosis
SNO meeting in New Orleans in November 2018, the ISRS after stereotactic radiosurgery: is a differential diagnosis possible? J
(International Stereotactic Radiosurgery Society) meeting in Neurosurg. 2003;98(5):1056–1064.
Rio in June 2019, and the SNO Inaugural Conference on Brain 14. Glaudemans AW, Enting RH, Heesters MA, et al. Value of
Metastases in New York in August 2019. 11C-methionine PET in imaging brain tumours and metastases. Eur J
Nucl Med Mol Imaging. 2013;40(4):615–635.
15. Narloch JL, Farber SH, Sammons S, et al. Biopsy of enlarging lesions
after stereotactic radiosurgery for brain metastases frequently reveals
Authorship Statement. Experimental design: BSS, GOS; im- radiation necrosis. Neuro Oncol. 2017;19(10):1391–1397.
plementation of the data: BSS, GOS, PØE, P-HP, JIH; anal- 16. Ruzevick J, Kleinberg L, Rigamonti D. Imaging changes following stere-
ysis and interpretation of the data: BSS, GOS, PØE, P-HP, JIH, otactic radiosurgery for metastatic intracranial tumors: differentiating
GEE, JK. pseudoprogression from tumor progression and its effect on clinical
practice. Neurosurg Rev. 2014;37(2):193–201; discussion 201.
17. Vogelbaum MA, Angelov L, Lee SY, Li L, Barnett GH, Suh JH. Local con-
trol of brain metastases by stereotactic radiosurgery in relation to dose
to the tumor margin. J Neurosurg. 2006;104(6):907–912.
References 18. Skeie BS, Skeie GO, Enger PO, et al. Gamma knife surgery in brain
melanomas: absence of extracranial metastases and tumor volume
strongest indicators of prolonged survival. World Neurosurg. 2011;75(5–
1. Song YP, Colaco RJ. Radiation necrosis—a growing problem in a case 6):684–691; discussion 598–603.
of brain metastases following whole brain radiotherapy and stereotactic 19. Iyer A, Harrison G, Kano H, et al. Volumetric response to radiosurgery for
radiosurgery. Cureus. 2018;10(1):e2037. brain metastasis varies by cell of origin. J Neurosurg. 2014;121(3):564–569.
2. Vellayappan B, Tan CL, Yong C, et al. Diagnosis and management of 20. Andrews DW, Scott CB, Sperduto PW, et al. Whole brain radiation
radiation necrosis in patients with brain metastases. Front Oncol. therapy with or without stereotactic radiosurgery boost for patients with
2018;8:395. one to three brain metastases: phase III results of the RTOG 9508 ran-
3. Yoshii Y. Pathological review of late cerebral radionecrosis. Brain Tumor domised trial. Lancet. 2004;363(9422):1665–1672.
Pathol. 2008;25(2):51–58. 21. Skeie BS, Enger PØ, Ganz JC, et al. Gamma knife surgery of colorectal
4. Simard JL, Smith M, Chandra S. Pseudoprogression of melanoma brain brain metastases: a high prescription dose of 25 Gy may improve growth
metastases. Curr Oncol Rep. 2018;20(11):91. control. World Neurosurg. 2013;79(3–4):525–536.
5. Cohen JV, Alomari AK, Vortmeyer AO, et al. Melanoma brain metastasis 22. Skeie BS, Eide GE, Flatebø M, et al. Quality of life is maintained using
pseudoprogression after pembrolizumab treatment. Cancer Immunol Gamma Knife radiosurgery: a prospective study of a brain metastases
Res. 2016;4(3):179–182. patient cohort. J Neurosurg. 2017;126(3):708–725.
6. Wen PY, Chang SM, Van den Bent MJ, Vogelbaum MA, Macdonald DR, 23. Patel TR, McHugh BJ, Bi WL, Minja FJ, Knisely JP, Chiang VL. A compre-
Lee EQ. Response assessment in neuro-oncology clinical trials. J Clin hensive review of MR imaging changes following radiosurgery to 500
Oncol. 2017;35(21):2439–2449. brain metastases. AJNR Am J Neuroradiol. 2011;32(10):1885–1892.
10 Skeie et al. The Bergen Criteria—pseudoprogression vs. recurrence

24. Kano H, Kondziolka D, Lobato-Polo J, Zorro O, Flickinger JC, melanoma brain metastases treated with ipilimumab. J Neurosurg.
Lunsford LD. Differentiating radiation effect from tumor progression 2017;127(5):1007–1014.
after stereotactic radiosurgery: T1/T2 matching. Clin Neurosurg. 27. Lehrer EJ, Peterson J, Brown PD, et al. Treatment of brain metastases
2010;57:160–165. with stereotactic radiosurgery and immune checkpoint inhibitors: an in-
25. Sneed PK, Mendez J, Vemer-van den Hoek JG, et al. Adverse radiation ternational meta-analysis of individual patient data. Radiother Oncol.
effect after stereotactic radiosurgery for brain metastases: incidence, 2019;130:104–112.
time course, and risk factors. J Neurosurg. 2015;123(2):373–386. 28. Kamath SD, Kumthekar PU. Immune checkpoint inhibitors for the treat-
26. Cohen-Inbar O, Shih HH, Xu Z, Schlesinger D, Sheehan JP. ment of central nervous system (CNS) metastatic disease. Front Oncol.
The effect of timing of stereotactic radiosurgery treatment of 2018;8:414.

Downloaded from https://academic.oup.com/noa/article/2/1/vdaa026/5802330 by guest on 04 February 2022