Abstract
Background/Aim: Diffusion-weighted imaging (DWI) is a magnetic resonance imaging (MRI) technique based on measure of water diffusion that can provide information about tissue microstructure, especially about cell count. Increase of cell density induces restriction of water diffusion and decreases apparent diffusion coefficient (ADC). ADC can be divided into three sub-parameters: ADC minimum or ADCmin, mean ADC or ADCmean and ADC maximum or ADCmax. Some studies have suggested that ADCmin shows stronger correlations with cell count in comparison to other ADC fractions and may be used as a parameter for estimation of tumor cellularity. The aim of the present meta-analysis was to summarize correlation coefficients between ADCmin and cellularity in different tumors based on large patient data. Patients and Methods: For this analysis, MEDLINE database was screened for associations between ADC and cell count in different tumors up to September 2016. For this work, only data regarding ADCmin were included. Overall, 12 publications with 317 patients were identified. Spearman's correlation coefficient was used to analyze associations between ADCmin and cellularity. The reported Pearson correlation coefficients in some publications were converted into Spearman correlation coefficients. Results: The pooled correlation coefficient for all included studies was ρ=−0.59 (95% confidence interval (CI)=−0.72 to −0.45), heterogeneity Tau2=0.04 (p<0.0001), I2=73%, test for overall effect Z=8.67 (p<0.00001). Conclusion: ADCmin correlated moderately with tumor cellularity. The calculated correlation coefficient is not stronger in comparison to the reported coefficient for ADCmean and, therefore, ADCmin does not represent a better means to reflect cellularity.
Diffusion-weighted imaging (DWI) is a magnetic resonance imaging (MRI) technique based on measuring water diffusion in tissues (1). DWI can provide additional information about tissue microstructure, especially about cell count (1-5). Previously, some clinical and experimental studies investigated associations between apparent diffusion coefficient (ADC) and cellularity in several benign and malignant lesions (2-5). In most reports, statistically significant correlations between the parameters were identified (2-5). It has been shown that increase of cell density induced restriction of water diffusion and decreased ADC (2-5). Furthermore, according to the literature, ADC can be divided into three sub-parameters: ADC minimum or ADCmin, mean ADC or ADCmean and ADC maximum or ADCmax (6-9). Some studies have suggested that ADCmin shows stronger correlations with cell count in comparison to other ADC fractions and, therefore, may be used as a parameter for estimation of tumor cellularity (6, 8). Onishi et al. reported that, in breast cancer, the correlation coefficient for ADCmin and cellularity was −0.537 (p=0.022), whereas for ADCmean it was −0.412 (p=0.09) (8). However, other authors did not confirm these results (7, 9). For instance, in the study of Chen et al., investigated DWI findings in lung cancer demonstrated that the correlation coefficient between cellularity and ADCmin was −0.47 (p<0.01), and between ADCmean and cellularity −0.6 (p<0.01) (7).
The aim of the present meta-analysis was to estimate the correlation coefficient between ADCmin and cellularity in different tumors based on large patient data.
Patients and Methods
Data acquisition and proving. For this analysis, MEDLINE database was screened for associations between ADC and cell count in different tumors up to September 2016. The following search criteria were used: “DWI or diffusion-weighted imaging or diffusion-weighted imaging or ADC or apparent diffusion coefficient AND cellularity or cell density or cell count or cell number”. Secondary references were also recruited. We extracted only publications in English and used the Preferred Reporting Items for Systematic Reviews and Meta-Analyses statement (PRISMA) (10).
As a next step, duplicates and papers without information regarding associations between DWI and cellularity were excluded. Thereafter, 494 publications were involved into further analysis. For this work, only data regarding ADCmin were included. Exclusion criteria were as follows: Papers that did not contain correlation coefficients between ADCmin and cell count; Data retrieved from diffusion tensor imaging; Data regarding DWI parameters other than ADCmin, such as ADCmax and ADCmean; Experimental animals and in vitro studies.
Overall, 482 publications were excluded and, therefore, our analysis comprises 12 publications with 317 patients (7-9, 11-19). One study (8) contained two patient samples, therefore 13 patients samples were included. The following data were extracted from the literature: authors, year of publications, number of patients, tumor type and correlation coefficients (Table I).
The methodological quality of the 12 included studies was independently checked by two observers (A.S. and H.J.M.) using the Quality Assessment of Diagnostic Studies (QUADAS) instrument (20, 21). The results of QUADAS proving are shown in Table II.
Statistical analysis. Spearman's correlation coefficient was used to analyze associations between ADCmin and cellularity. The reported Pearson correlation coefficients in some publications were converted into Spearman correlation coefficients as reported previously (22).
The meta-analysis was undertaken by using software RevMan 5.3 (Computer program, version 5.3. Copenhagen: The Nordic Cochrane Centre, The Cochrane Collaboration, 2014). Heterogeneity was calculated by means of the inconsistency index I2 (23, 24). In a subgroup analysis, studies were stratified by tumor type. Furthermore, DerSimonian and Laird random-effects models with inverse-variance weights were used without any further correction (25).
Results
The pooled correlation coefficient for all included studies (Figure 1) was ρ=−0.59 (95% confidence interval (CI)=−0.72 to −0.45), heterogeneity Tau2=0.04, (p<0.0001), I2=73%, test for overall effect Z=8.67 (p<0.00001).
Discussion
The present analysis provided the correlation coefficient between ADCmin and cellularity in a large cohort.
The search for imaging parameters, which can reflect tissue composition of several tumors, has a high clinical relevance. They can be used as biomarkers for tumor cellularity, proliferation potential and, therefore, also predict tumor behavior. Previously, numerous studies investigated relationships between different imaging features and histopathology in benign and malignant lesions (26, 27). Especially ADC has been reported to have a great potential (7, 27). Furthermore, as mentioned above, ADC consists of different fractions: ADCmin, ADCmean and ADCmax, which may reflect different histopathological features (9, 19, 27). It has been reported that ADCmin correlated statistically significant with cell count but not with proliferation index Ki-67, whereas ADCmean correlated well with Ki-67 but not with cell count (27). Moreover, both parameters correlated well with total nucleic areas (27). In addition, ADCmax correlates slightly with cell count but not with Ki-67 and nucleic areas (27). However, other authors have indicated that none of ADC parameters correlated with cellularity (15). The main problem of the reported data was that they were based on small number of investigated lesions. This fact and controversial results question the use of ADC parameters in clinical practice and highlight the need for studies based on larger samples and/or systematic analysis of the published data. Recently, a meta-analysis regarding associations between ADCmean and cellularity in different tumors was reported (28). It has been shown that the cumulative correlation coefficient was −0.56 (28). Furthermore, it ranged significantly in different tumors (28). As seen, the cumulative correlation coefficient between cellularity and ADCmin calculated in the present analysis does not differ significantly from the reported coefficient for ADCmean. Therefore, in contrast to previous reports (6, 8, 18, 27), we postulate that ADCmin does not represent a better means to reflect cellularity. However, further studies are needed to investigate this association in larger groups and, more importantly, also in different tumors. It may be possible that, in some tumors, ADCmin correlates stronger with cell count.
In conclusion, ADCmin correlated moderately with tumor cellularity. The calculated correlation coefficient is not stronger in comparison to the reported coefficient for ADCmean and, therefore, ADCmin does not better reflect cellularity as expected.
Footnotes
All authors contributed equally to this study.
- Received May 1, 2017.
- Revision received May 25, 2017.
- Accepted May 26, 2017.
- Copyright© 2017, International Institute of Anticancer Research (Dr. George J. Delinasios), All rights reserved