Key-words:
Apparent diffusion coefficient - glioblastoma - magnetic resonance imaging
Introduction
Gliomas are tumors of the white matter that are diffusely infiltrative and constitute
approximately 80% of all brain tumors.[[1]],[[2]] Oligodendrogliomas (OD) are a subtype of gliomas (according to the World Health
Organization [WHO] 2016 classification of brain tumors), that are associated with
a better prognosis, especially if associated with chromosome 1p/19q co-deletion.[[3]],[[4]],[[5]],[[6]],[[7]] The prevalence of this mutation is similar in both anaplastic (WHO grade 3) and
low-grade OD[[8]] and the presence of this co-deletion is an independent predictor of remarkably
better progression-free survival and overall survival.[[9]] The gold standard for diagnosis of OD and 1p/19q deletion status remains a biopsy
for histopathology and gene analysis. However, more recently, several radiological
features on magnetic resonance imaging (MRI) scans have been reported that are suggestive
of diagnosis of OD and the co-deletion status.[[4]],[[10]] This radiological differentiation is important for several reasons; it provides
the treating physician a high index of suspicion prior to a formal biopsy, as despite
significant advances in intraoperative aids and operative techniques, around 5% of
biopsies yield tissue specimen that is insufficient in size or quality to demonstrate
1p/19q loss.[[10]] It may assist in the choice of tissue for sampling in case of difficult-to-access
tumors. It may allow us a better understanding of the morphological differences between
various types of OD, allowing a more elaborate classification system. Finally, in
low-middle income countries, where genetic analysis is not widely available (there
is only one center in our country of more than 200 million that performs genetic analysis
on brain tumor specimen), an MRI scan may be the only aid to histopathology for subclassifying
these tumors.
This, however, may only be possible if we can establish reliable radiological features
that can predict the 1p/19q status of the OD, with high sensitivity and specificity.
Megyesi et al. studied 33 patients and concluded that 1p/19q co-deletion OD stratification
based on grading is possible through MRI alone.[[11]] This reflected that 1p/19q co-deleted tumors have distinct radiological features
and this finding was further validated by Preusser et al. in their study of 67 patients
with histological and molecular diagnosis of gliomas.[[12]] Apparent diffusion coefficient (ADC) was found to be the best differentiating characteristic
between the different genetic subtypes of gliomas as also outlined in the 2016 WHO
guidelines.[[6]],[[7]] ADC is a measure of the magnitude of diffusion of water molecules found within
the tissue. Values of ADC are automatically calculated, using a diffusion-weighted
image (DWI), by software and displayed in the form of a parametric map highlighting
the degree of diffusion of molecules of water through different types of tissues.
Regions of Interest (ROIs) are used to define the region for which ADC values are
being calculated. In addition, new modalities such as diffusion tensor magnetic resonance-derived
metrics and arterial spin labeling are rapidly increasing the options available for
precise, non-invasive testing based on radiological parameters.[[13]],[[14]],[[15]],[[16]],[[17]]
Based on these studies, we hypothesize that OD with 1p/19q chromosomal deletion is
radiologically distinct from OD without 1p/19q loss. In this study, we attempted to
explore whether 1p/19q codeletion, Ki-67, and tumor grade in OD can be reliably predicted
based on mean ADC volumes on MRI scans in conjunction with other radiological features.
Methodology
Since the study was retrospective in nature, hospital records were reviewed and patients
with biopsy-proven ODs were selected. The cohort was further narrowed by isolating
patients with pre-surgery MRI scan, done using the in-house 1.5 Tesla scanner, and
a biopsy specimen confirming the diagnosis of OD. The selected patients were then
divided into those with 1p19q co-deletion and those without 1p19q co-deletion. The
World Health Organization 2016 central nervous system tumor grading criteria were
used to determine the grade of tumors.
Apart from ADC and grade, six additional parameters were investigated. Homogeneity
within the tumors, regularity of tumor edge, necrosis, hemorrhage, calcification,
and contrast enhancement were also reviewed in the MRIs. These were used to identify
aggressive features in high-grade gliomas, aggressive features were classified as
extensive necrosis, indeterminate edges, presence of hemorrhage, calcification, and
strong contrast enhancement. Tumor location was decided to be set as the region or
lobe containing the bulk (more than 80% of mass) of tumor. The corpus callosum was
set as the point to distinguish unilateral and bilateral tumors, with OD traversing
the corpus callosum considered to be bilateral. Allelic loss of chromosome was determined
through loss of heterozygosity assays in tumor DNA pair. Microsatellite markers were
used on chromosomes 1p36 and 19q13.
MRI imaging was done using a 1.5 Tesla (T) clinical MR imaging system (General Electric
Signa HD MRI systems) using an eight-channel phased-array breast coil. A sample of
these images is shown in [[Figure 1]] and [[Figure 2]]. A T2-weighted transverse pulse sequence was performed with 60/5600/180 (echo time/repetition
time/inversion time) ms, 4 mm thickness, a field of view of 36 cm × 36 cm, and a matrix
of 316 × 320. An axial plane was used to acquire DWI images and ADC maps were automatically
created by the system using the trace-weighted images with b values of 0 and 1000.
ADC values were calculated using the following formula: ADC = -(1/b) ln (S2/S1), where
S2 and S1 are the intensity of signals at a b value of 1000 and 0, respectively. A
sample of the imaging and some data generated are present in [[Figure 1]] and [[Figure 2]].
Figure 1: Sample of magnetic resonance imaging in a patient without 1p/19q codeletion
Figure 2: Sample of magnetic resonance imaging in a patient with 1p/19q codeletion
The study underwent ethical review by the Aga Khan University Hospital Ethical Review
Committee before data collection and analysis was done. No human interventions were
involved throughout the study; no financial compensations were made to study participants.
Data were stored in a secure password-protected folder on an encrypted computer accessible
by the primary investigator.
Data analysis was divided into the three steps. First, we generated descriptive statistics
(mean and standard deviation for continuous variables and frequency and proportions
for categorical variables). Second, we evaluated normality for variables such as ADC
value, Age, and Ki-67 tumor proliferation index (Histogram, Smirnov test of hypothesis
for normality). Lastly, since the data were not normally distributed, we applied nonparametric
tests to assess the relationship between predictors of ADC inferential statistics
with a P value of alpha <0.05 set as a level of significance.
For our primary objectives, ADC was correlated with 1p19q codeletion, grade of tumor,
and Ki-67 value. For our secondary objectives, MRI features were also correlated with
tumor grade and 1p19q co-deletion separately. Details of each step are discussed in
the results section below.
Results
Demographic and clinical characteristics of participants
A total of 73 patients participated in this study. The mean age of the patients was
38.86 years. The average value of ADC was found to be 1286.0. The study revealed that
the majority of 49 (67.1%) patients were male; whereas, 24 (32.9%) were female. On
histopathology, 39 (53.4%) were Grade II and 34 (46.6%) OD were grade III. Around
34 (46.6%) were co-deleted tumors and 39 (53.4%) were non co-deleted tumors in 1p19q
genetic testing. [summarized in [[Table 1]]]
Table 1: Descriptive Statistics of ADC among patients with Oligodendroglioma
Forty-three (59%) tumors were located predominantly in the frontal lobe, 20 (27.3%)
in the temporal lobe and 10 (13.7%) in the parietal lobe. Among Grade III OD, 2 (6.9%)
showed no contrast enhancement, 23 (79.3%) showed some contrast enhancement and 4
(13.8%) showed homogenous contrast enhancement. Eighteen Grade II OD (72%) had shown
no contrast enhancement, 6 (24%) showed some contrast enhancement, and 1 (4%) showed
homogenous contrast enhancement.
Forty (54.8%) of the tumors had irregular margins, with 4 (5.5%) having smooth margins
and 29 (39.7%) having indeterminate margins. Thirty-three (45.2%) of the tumors had
cystic or necrotic changes visible while 40 (54.8%) did not.
Ki 67, an alternative assay of cellular proliferation, was identified for all 73 tumors
and was found to be high in 33 (45.2%) tumors, low index in 25 (34.2%) and it was
not available for 15 (20.5%) samples.
Sixty-five (89%) of the tumors were unilateral and 8 (11%) were bilateral. Forty-three
(58.9%) tumors were located predominantly in the frontal lobe, 20 (27.4%) in the temporal
lobe, and 10 (13.7%) in the parietal lobe as shown in [[Table 1]].
Inferential statistics
We determined the predictors of ADC among patients with oligodendroglioma. The normality
test showed that the data are not normal as shown in [[Table 2]]; therefore, we applied non-parametric tests.
Table 2: Kolmogorov.Smirnov test statistics for normality testing
In order to assess the relationship between predictors of ADC, inferential statistics
such as Mann –Whitney test, Kruskal Wallis, and Spearman rank Correlation tests were
run to analyze the dataset as shown in [[Table 3]]. P value of alpha less than 0.05 was set as level of significance.
Table 3: Inferential Test statistics according to types of variables
[[Table 4]] shows that only one variable, Ki-67 tumor proliferation index, was statistically
associated with ADC level among patients with oligodendroglioma. Since P value (0.048)
was <0.05 for Ki-67 tumor proliferation index and ADC. We conclude that there is a
significant relationship between these two variables.
Table 4: Inferential Statistics of ADC among patients with Oligodendroglioma
Linear regression analysis
Since the outcome variables were continuous in nature, linear regression analysis
was carried out for the inferential statistics. Beta coefficient and their confidence
intervals were computed as mean estimated change in ADC with one-unit change in the
predictor variable.
Univariate analysis
Univariate analysis was run to analyze the independent effect of each independent
variable with ADC. Each individual variable was individually regressed against ADC
level. The results are shown in [[Table 5]].
Table 5: Univariate Analysis of Determinants
Discussion
Our results highlight that ADC is a good indicator of Ki-67. However, ADC alone is
a weak indicator of 1p/19q co-deletion and tumor grade. Limitations based on population
characteristics, sample size, and using one radiological parameter may be a possible
reason for our results.
1p/19q co-deletion and apparent diffusion coefficient
The mean value of ADC of 1p/19q co-deleted tumors was not statistically significant
(as P value was >0.05) from the mean ADC of non-co-deleted tumors in our study. This
reflects literature which shows increased cellularity in 1p/19q co-deleted tumors
in some cases but no significant difference in mean value of ADC alone.[[10]],[[18]],[[19]],[[20]],[[21]] This finding corresponds with Fellah et al. who reported a similar finding in 270
patients in their study on a single-center experience.[[22]] However, Jenkinson et al. report that tumors with 1p/19q loss are more likely to
have a smaller mean ADC compared with tumors that do not have 1p/19q chromosomal deletion.[[23]] In our results, we see a trend similar to that found by Fellah et al. which concluded
that mean ADC alone is not sufficient as a modality to distinguish between 1p/19q
co-deleted and non-co-deleted tumors. Other modalities such as cerebral blood volume
might serve as an adjunct parameter to introduce significance for a radiological testing
modality.[[22]] Moreover, Abdel Razzak et al. have demonstrated the advent of new modalities such
as arterial spin imaging, diffusion tensor imaging, and their respectively derived
metrics as viable alternatives to conventional DWI based metrics.[[14]],[[15]],[[17]]
Ki-67 and apparent diffusion coefficient
In a study by Preusser et al. in 2012, Ki-67 was evaluated as a clinical and prognostic
tool and was found to have a strong prognostic impact.[[12]] In our study, the OD with the lowest Ki-67 index were all non 1p/19q co-deleted
tumors, while the 1p/19q co-deleted tumors had a higher ratio of tumors with a high
Ki-67 index. Ki-67 values were also significantly different (with P = 0.048) between
different tumor grades making Ki-67 a viable predictor for tumor grades. Studies by
Pouget et al. and Duregon et al. both investigate the prognostic impact of Ki-67 in
OD with 1p/19q codeletion and both conclude that Ki-67 is a strong predictor of prognosis.[[24]],[[25]] This conclusion correlates with other literature, pointing towards the idea that
Ki-67 could be used to support prognostic and therapeutic clinical decisions.[[26]],[[27]]
Tumor grade and apparent diffusion coefficient
Mean ADC for grade III OD, irrespective of 1p/19q co-deletion, was 1273, while mean
ADC for grade II OD was 1300. A lower mean ADC is a marker of higher cellularity that
corresponds with the histological grade of the tumors. Latysheva et al. reported in
2019 that Histogram derived ADC parameters could successfully distinguish between
OD and oligoastrocytoma.[[28]] Similarly, Anwar et al. report a significant difference in ADC mapping for different
tumors.[[29]] However, data in our study does not demonstrate a significant (P < 0.05) difference
between mean ADC of OD Grade II and III. These findings are reflected in literature
by Hilario et al. and Fellah et al. who also found no significant correlation between
grade of OD and the mean value of ADC.[[18]],[[22]] Naveed et al. who further explored grading of OD through MRI concluded that ADC
values when combined with relative cerebral blood volume and MR spectroscopy is sufficient
in differentiating groups of OD.[[30]] Thus, a study combining mean ADC with other radiological parameters[[19]],[[31]] such as MR spectroscopy[[32]] or rCBV may yield a significant result in distinguishing grades of OD based purely
on radiological characteristics.
Other radiological characteristics and 1p/19q co-deletion
Co-deletion of 1p19q in OD is likely due to recurring translocation and may be a marker
of therapeutic response to chemotherapy and overall long-term survival. Some basis
for radiological distinction between OD based on 1p/19q loss has already been reported
in literature. Megyesi et al. showed that OD with 1p/19q loss showed indistinct borders,
paramagnetic susceptibility, and calcification more commonly than their counterparts.[[11]] In our data, almost half (54.7%) of the tumors had irregular borders, while 39.4%
had indeterminate borders. None of the tumors with 1p/19q co-deletion were found to
have smooth borders, the majority having indeterminate borders. Homogenous signal
intensity was also found more commonly in tumors with higher ADC, and tumors with
noncircumscribed borders were found only above a mean ADC of 1000 mm2/s.
A study published as far back as 2001 showed that bi-hemispherical growth patterns
and peripheral tumor location had a significant correlation with chromosomal 1p/19q
codeletion.[[33]] Our data showed that 59% of our subjects had the tumor present in the frontal lobe,
with the second most common site being the temporal lobe. 1p/19q co-deleted OD had
twenty-three (67.6%) tumors present in the frontal lobe, seven (20.6%) in the parietal
lobe and four (11.8%) in the temporal lobe. In contrast, 1p/19q non-co-deleted OD
had 18 (50%) tumors in the frontal lobe, 12 (33.3%) in the parietal lobe, and six
(16.7%) tumors in the temporal lobe.
Other radiological characteristics and grade of tumor
Grade III tumors in our study showed significant partial or homogenous contrast enhancement
when compared with Grade II OD. Similarly, Grade III tumors showed a greater ratio
of indeterminate edges, necrosis, and cystic changes. In Grade III OD, 1p/19q codeletion
has been associated with distinct radiological characteristics,[[20]] particularly blurred tumor borders, frontal lobe location, and intra-tumoral signal
heterogeneity.[[34]]
Quantitative analysis of MRI has demonstrated a high sensitivity and specificity to
1p/19q chromosomal codeletion.[[10]] Another research has shown that angiogenic subtypes of OD that are distinguishable
on MRI can also be indicative of 1p/19q loss and other chromosomal deletions.[[35]] This research is aided by the fact that sufficient advances in MRI techniques and
analytical tools have also allowed for more introspective analysis and quantitative
logging of features extracted from images of tumors.[[17]],[[21]]
Currently, diagnostic testing of OD tumors for chromosomal deletions such as 1p/19q
is not widely available even in more developed regions. In a low-resource and population
intensive region, there are very limited centers where chromosomal testing is available,
and even where available the cost is prohibitive. The technical expertise required
to analyze and correctly interpret the data are often lacking. Fluorescence in situ
hybridization and loss of heterozygosity studies present an appealing alternative
but hold their own drawbacks. Inherent sampling errors and difficulty in residual
tumor evaluation are two major ones.
An alternative method of predicting molecular and pathological patterns in OD is emerging
with the evolution of radiological techniques. MRI studies, which are done for the
diagnosis of all ODs, can serve as an analytical tool based on software analysis and
mapping techniques. This would further aid clinical and surgical decision-making in
providing a patient-oriented approach with each case.
Conclusion
We conclude that mean ADC is a useful diagnostic tool in predicting Ki-67 values of
OD and thus overall prognosis. However, mean ADC alone would not improve prebiopsy
discrimination of 1p/19q co-deleted tumors from 1p/19q nonco-deleted tumors at the
moment. Histopathological examination remains the gold standard for classification
and grading of brain tumors. Even though radiological features may be suggestive of
certain grades and genotypes, with the technology currently available in developing
countries, it continues to show low sensitivity and specificity.
