CircRNAs in diagnosis, prognosis, and clinicopathological features of multiple myeloma; a systematic review and meta-analysis

Unlike improved treatment response in multiple myeloma (MM), the mortality rate in MM is still high. The study’s aim is to investigate the potential role of circRNAs as a new biomarker for diagnosis, prognosis, and clinicopathological features of MM. We identified studies through Web of Science, Scopus, PubMed and ProQuest databases, and Google Scholar to August 2022. The SEN, SPE, PLR, NLR, DOR, and AUC were combined to investigate the diagnostic performance of circRNAs in MM. Also, HR and RR were used for prognostic and clinicopathological indicators, respectively. 12 studies for prognosis, 9 studies about diagnosis, and 13 studies regarding clinicopathological features. The pooled SEN, SPE, DOR, and AUC were 0.82, 0.76, 14.70, and 0.86, respectively for the diagnostic performance of circRNAs. For the prognostic performance, oncogene circRNAs showed a poor prognosis for the patients (HR = 3.71) and tumor suppressor circRNAs indicated a good prognosis (HR = 0.31). Finally, we discovered that dysregulation of circRNAs is associated with poor clinical outcomes in beta-2-microglobulin (RR = 1.56), Durie-Salmon stage (RR = 1.36), and ISS stage (RR = 1.79). Furthermore, the presence of del(17p) and t(4;14) is associated with circRNA dysregulation (RR = 1.44 and 1.44, respectively). Our meta-analysis demonstrates that the expression analysis of circRNAs is valuable for MM’s diagnosis and prognosis determination. Also, dysregulation of circRNAs is associated with poor clinicopathological features and can be used as the applicable biomarkers for evaluating treatment effectiveness. Supplementary Information The online version contains supplementary material available at 10.1186/s12935-023-03028-z.


Introduction
Multiple myeloma (MM) is a type of plasma cell dyscrasia that may start with a monoclonal gammopathy of undetermined significance (MGUS) and progress to plasma cell leukemia and extramedullary myeloma [1].In MM patients increased secretion of nonfunctional intact immunoglobulins or light chains, can be detected in serum and/or urine [2][3][4].Currently, diagnosis, assessment of response to treatment, and minimal residual disease (MRD) in MM patients are made based on the IMWG group criteria [5][6][7].Improved treatment response and significantly increased survival have been observed in recent decades, resulting from the use of various therapies in patients with MM [2,3,8].In addition, increased attention must be paid to CRAB (hypercalcemia, renal failure, anemia, and lytic bone lesions) in the multiple myeloma treatment [5].In recent years, many studies have been done on epigenetic processes involved in the pathogenesis and development of MM, especially studies on diagnostic and prognostic biomarkers with high informative value.
Circular RNAs (circRNAs) are one of the newest types of non-coding RNAs [9].These single-stranded circular RNAs belong to the long non-coding RNAs, and unlike linear RNAs, they are covalently closed and lack 5' caps and 3' tails, which makes them resistant to digestion by RNase and thus more stable [10].CircRNAs are produced from precursor mRNAs by the back-splicing mechanism [11].Recent studies in various diseases, especially blood cancers, have shown that circRNAs can play a crucial role as oncogenes or tumor suppressors in intracellular processes by sponging with microRNAs [11][12][13].Several studies have investigated the association between circRNAs and pathogenesis, prognosis, diagnosis, and clinicopathological features in MM patients.For example, in 2021, Fan Zhou et al investigated the relationship between 10 circRNAs with high expression and 10 cir-cRNAs with low expression with the clinicopathological features, diagnosis, and prognosis of the disease using microarray analysis and qRT-PCR assays in MM samples [14].
Currently, several methods can be used to diagnose and evaluate the prognosis of MM patients, such as complete blood examination, serum/urine protein detection, bone marrow aspiration/biopsy, flow cytometry, skeletal examination (e.g., X-ray and CT scan), and the ISS and currently revised ISS (R-ISS) systems [5,15].Although bone marrow aspiration or biopsy is a well-known approach to confirm the diagnosis, both are quite invasive, expensive, and time-consuming [2].The using of flow cytometry has promoted the diagnosis of multiple myeloma, but the lack of specific markers and high expensive are limitations of this method [16].In addition, ISS is a highly accurate method for prognosis determination, but due to the need for systems like interphase fluorescence in situ hybridization and the complex interpretation of the results, these systems are difficult to use [17].Therefore, it is necessary to develop some minimally invasive and cost-effective methods and discover biomarkers to complement and improve the current strategies for the diagnosis and prognosis of MM.
The purpose of our article is to explore the role of cir-cRNAs in the pathogenesis, development, and response to treatment in patients with MM.A meta-analysis was also carried out using data from included studies to determine the diagnostic and prognostic value of circR-NAs for MM.The correlation between circRNAs and clinicopathological features in MM patients was also evaluated.

Eligibility criteria
We accomplished a systematic review, registered on PROSPERO (ID: CRD42022345468).This study was carried out based on PRISMA guidelines [18].The inclusion criteria were: (A) any sort of peer-reviewed study examining the function of circRNAs (including cellular, circular, and exosomal) in patients with MM, including cohort and case-control studies; (B) studies dealing with aspects of diagnosis, prognosis, progression, and response to treatment of MM.The exclusion criteria were: (A) studies without a complete paper, insufficient data, or just employing an in-silico methodology; (B) non-Englishlanguage articles and (C) studies on animals.

Information sources
The WOS, Scopus, PubMed, ProQuest databases and Google Scholar were searched for articles published through August 2022.Grey literature sources such as allconferences.com,conferencealerts.com,opengrey, and oatd.org were also searched.The reference lists of included articles were also examined.

Selection process
Two researchers (A.A and Y.M) screened the titles and abstracts of all retrieved studies to determine potentially relevant studies for this systematic review.In the next step, the studies' full text was independently assessed by two researchers to verify the qualified to be included according to the inclusion and exclusion criteria mentioned in Sect."Eligibility criteria".Any disagreement encountered was resolved by discussion, and if there were unresolvable disagreements, the final decision was made by the third researcher (M.R).Initial screening of the extracted articles was performed using the web-based software Rayyan [19].

Data collection process
Data extraction of the included articles was performed separately by three researchers (A.A, Y.M, and M.M) based on the data extraction checklist, and if there were unresolvable disagreements, the final decision was made by the fourth researcher (M.R).The WebPlotDigitizer 4.6 software was used to indirectly extract the data from the Kaplan-Meier and receiver operating characteristic (ROC) curves.The methods described by Tierney were used to calculate HR and 95% CI indirectly [20].However, before the indirect extraction of the data, the authors of the included studies were contacted three times (by email) to obtain information.

Data items
Three researchers extracted the data by using a pre-specified form.The extracted data included the first author's name; the name of the circRNA; the year; the number of patients; the number of the control group; changes in cir-cRNA expression; the type of sample; the methods for cir-cRNA analysis (techniques); the control gene; the effect of the circRNA on cell biology; microRNA sponging; and the effect of the circRNA on response to treatment.The required information extracted for the prognosis metaanalysis includes the following: HR with 95% CI for OS (if reported in the article), follow-up time, and survival outcome.Data extracted for the meta-analysis of diagnosis include the following: sensitivity SEN, specificity SPE, cutoff value point, AUC, true positive (TP), false positive (FP), false negative (FN), and true negative (TN).Finally, for the meta-analysis of clinicopathologic features, the data were extracted from the clinicopathologic characteristics tables that are as follows: Gender, B2-MG, albumin, hypercalcemia, renal insufficiency, bone lesions, Durie-Salmon (DS) stage, ISS, and cytogenetic abnormalities such as del (17p), t (4;14), and t (14;16).

Bias assessment of the studies included
The bias risk assessment was carried out using the Quality Assessment for Studies of Diagnostic Accuracy II (QUADAS II) checklist for diagnostic articles [21], and the Newcastle-Ottawa Scale (NOS) for cohort and casecontrol articles [22].The QUADAS II checklist Composed of four key scopes, including patient selection, index test, reference standards, and flow of patients.According to the QUADAS II tool, studies were rated ≥ 6 as high quality and < 6 as low quality (Additional file 2: Fig. S1).The NOS checklist evaluates selection categories, comparability, and outcome (cohort studies) /exposure (case-control studies) categories.articles scoring a 7 as good quality, 5-6 as fair quality, and < 5 as poor quality (Additional file 2: Table.S1).According to the QUADAS II tool, each article receives a maximum of 7 points, and according to the NOS checklist, each article receives a maximum of 9 points.

Statistical analysis
Extracted data that met the inclusion criteria was synthesized.For diagnostic analysis, the numbers of true positive (TP), false positive (FP), false negative (FN) and true negative (TN) were calculated, and finally the pooled sensitivity, specificity, AUC, PLR, NLR, DOR, 95% CIs, AUC, and heterogeneity were evaluated.The AUC values and their association with diagnostic accuracy are the following: 0.9 to 1.0: excellent, 0.8 to 0.9: very good, 0.7 to 0.8: good, 0.6 to 0.7: sufficient, 0.5 to 0.6: bad and < 0.5: test not useful, and also, good diagnostic tests have positive likelihood ratio (PLR) > 10 and negative likelihood ratio (NLR) < 0.1 [23,24].
For prognostic analysis, HR and 95% CIs were synthesized to examine the effect of circRNAs on OS.The RR and 95% CIs were used to analyze the clinical value of circRNAs' association with MM in terms of clinicopathological correlations.Due to methodological heterogeneity in the primary study, the Random Effects Model (REM) was used to combine HR and RR values [25].The magnitude of association between the study variables and the dysregulated expression of circRNAs and its interpretation areas for the prognostic index (HR) and clinicopathologic characteristics index (RR ) are as follows: 1 to 1.21: trivial (inconsiderable), 1.22 to 1:85: small, 1:86 to 2:99: moderate, 3 or more: large [26].The chi-square test and the I² statistic were utilized to assess the between-study heterogeneity.If an I² value was < 50%, it was considered to have no significant heterogeneity.To assess the potential source of heterogeneity, subgroup analysis were conducted according to similar features of the included studies, and also, a sensitivity analysis of all the included studies was carried out to find the effect of each article on the final effect of the meta-analysis results.Publication bias was examined quantitatively using the Deek's funnel plot, Egger's tests, and Trim and Fills tests.In this study, all meta-analysis was performed with STATA version 14.2 and Meta-Disc software.A p-value < 0.05 was considered statistically significant.

Study selection
The PRISMA flow diagram [18] of the studies' selection process is shown in Fig. 1.A total of 1041 studies were extracted via database searches.Prior to the initial screening, 168 articles were removed due to duplication.The title and abstract of 873 articles were initially screened by two researchers, and 841 of them were excluded due to incompatibility with the inclusion and  Chromosome; b 5-Aminoimidazole-4-Carboxamide Ribonucleotide Formyltransferase/IMP Cyclohydrolase exclusion criteria.32 studies were selected for full-text examine; 3 full-text studies were not retrieved, and 2 studies were excluded for the reasons described in Fig. 1.
Finally, the number of articles included in the qualitative synthesis was 27 [14, and the number of articles included in the quantitative synthesis meta-analysis was 15 [14, 27-29, 31-33, 35, 37-41, 48, 50].Of these, 9 articles were related to the meta-analysis of diagnosis, 12 articles were related to the meta-analysis of prognosis, and 13 articles were related to the meta-analysis of clinicopathological features.

Study characteristics
All the included articles were published between 2019 and 2022.The total number of patients was 1885, and the study population was exclusively Chinese.Changes in circRNA expression in the studies were measured by the qRT-PCR method.A total of 25 different cir-cRNAs were mentioned; in 10 articles, circRNAs had a tumor-suppressive role, and in 18 articles, circRNAs had an oncogenic role.Table 1 shows the role of circR-NAs in cell biology function and their relationship with various microRNAs, as well as the effect of circRNAs in response to treatment.The minimum follow-up period in cohort studies was 14 months, and the maximum was 60 months.In the study of Fan Zhou, 10 circRNAs with high and low expression were measured [14].To avoid multiplicity [53], one circRNA was selected to perform diagnostic and clinicopathologic features meta-analysis (circ-PTK2) and two circRNAs with oncogenic (circ-PTK2) and tumor suppressive (circ-AFF2) roles to perform prognostic meta-analysis.

Results of syntheses The prognostic performance of circRNAs in multiple myeloma
After reading the details of the I 2 included articles, the prognostic value of circRNAs was assessed.The main characteristics of prognostic studies are shown in Table 2. CircRNAs with an oncogenic role in MM patients were found in 7 studies and were negatively associated with the patients' prognosis.After metaanalysis, oncogene circRNAs showed poor prognosis for MM patients (high expression group vs. low expression group: HR = 3.71; 95% CI 2.89 to 4.76); also, I 2 = 0 showed that the results have low heterogeneity (Fig. 2A).Meanwhile, another 6 studies reported that circRNAs are tumor suppressors in MM patients and have a positive association with patient prognosis.
Tumor suppressor circRNAs indicated a good prognosis for MM patients (high expression group vs. low expression group: HR = 0.31; 95% CI 0.23 to 0.42) and I 2 = 0 indicated that the results have low heterogeneity (Fig. 2B).

Subgroup analysis
Due to significant heterogeneity, subgroup analyses were carried out according to the function of circRNAs (oncogenic or tumor suppressor) and quality studies based on QUADAS II (high or low) to evaluate the potential sources of heterogeneity.As shown in Table 4, oncogene circRNAs achieve a higher diagnostic performance than tumor suppressor circRNAs, with AUC values of 0.88 and 0.77, respectively.Moreover, a comparison of quality studies shows that the AUC (0.86 vs. 0.81) and the DOR (15.64 vs. 13.28) of high-quality studies were higher than those of low-quality studies (Forest plots of subgroup analysis are in the Additional file 3: Fig. S2).

The clinicopathological significance of circRNAs in multiple myeloma
Regarding the clinicopathological characteristics, 13 studies were included in our meta-analysis.We looked at the relationship between circRNA expression and clinicopathological features like gender, B2-MG, albumin, hypercalcemia, renal insufficiency, bone lesions, DS stages, ISS stages, and cytogenetic abnormalities like del(17p), t(4;14), and t(14;16) (at least five studies were looked at for each feature) (Table 5).Dysregulation of circRNAs has been associated with adverse clinical features DS stage; RR = 1.36, 95%CI

Related to prognosis
Low publication bias was found in the combined prognostic effects of two groups of oncogenes and tumor suppressors, as shown in Additional

Related to diagnosis
The sensitivity analysis showed that one included study (Shanshan Yu, 2020) had a big impact on the pooled results (Additional file 5: Fig. Fig.S3/Fig.3A).After removing this study, the I 2 value for the heterogeneity of DOR decreased from 98.61 to 92.89% (Table 4).Nonetheless, the pooled diagnostic values were comparable with those of the total studies (AUC: 0.86 vs. 0.85), showing that our results were relatively robust and not significantly to be affected by any individual of the included studies.

Discussion
CircRNAs play a role in a wide range of cell biology by sponging with various microRNAs in MM cells [51].As shown in Table 1, increasing or decreasing expression of circRNAs in MM cells ultimately affects the processes of proliferation, apoptosis, metastasis, cell cycle regulation, and response to treatment.Interestingly, in contrast to other studies, the study by Fang Chen [27] showed that circ-0069767, as a tumor suppressor, has increased expression in MM cells.The increased expression of this circRNA leads to a decrease in proliferation, migration, and invasion and an increase in apoptosis in MM cells.On the other hand, interestingly, some circRNAs have the ability to translate and produce proteins [54,55].CircRNAs through different mechanisms can be translated and produce proteins such as N6 methyladenosine modification or via the internal ribosome entry site (IRES), regions that elevate direct binding of initial factors to circular RNAs [56][57][58][59].Two studies by Xiaozhu Tang et al. have shown that circBUB1B and circ-HNRNPU have the ability to translate and produce circBUB1B_544aa and circHNRNPU_603aa proteins, respectively [43,44].
Several primary studies have demonstrated the prognostic value of circRNAs in MM.This prognostic metaanalysis included 12 studies and 1093 MM patients.MM patients with increased expression of oncogenic circR-NAs had a poorer OS and a nearly 4-fold higher risk of death than the control group (HR = 3.71); moreover, increased expression of tumor suppressor circRNAs are associated with a favorable OS, and almost 70% of the risk of death in this group is lower than the control group (HR = 0.31).So finally, According to the mentioned interpretation areas [26], a large correlation was observed between increased expression of oncogenic circRNAs and OS and a large correlation between increased expression of tumor suppressor circRNAs and OS.All these results indicate that circular RNAs play a role as novel biomarkers in predicting OS in patients with multiple myeloma.
Our results showed that circRNAs are diagnostic promising biomarkers for MM, with a combined AUC: 0.86 and DOR: 14.70, that larger AUC represents greater diagnostic value of each variable [23], and a higher DOR, as an important index used in meta-analysis of diagnostic studies, represents a more valuable indicator with better diagnostic efficacy (Fig. 3E).Moreover, the pooled sensitivity and specificity of circRNAs were 0.82 and 0.76, respectively, implying that circular RNAs represents good diagnostic accuracy.In addition, PLR values were 3.42, which means circRNA expression changes (positive results) happen 3.42 times more in a multiple myeloma patient than a patient without the multiple myeloma, and NLR values were 0.23, which means the probability of a negative test in a non-patient is 4.34 times greater than that of a negative test in an M.M patient.As circRNAs with diverse expression statuses may exert different functions in MM, we've performed subgroup analyses.Stratified analysis based on the function of circRNA showed better diagnostic accuracy for oncogene circRNAs than tumor suppressor circRNAs for MM.Moreover, based on quality subgrouping, it revealed that high-quality studies achieved a higher diagnostic performance than low-quality studies.
Heterogeneity is unavoidable in a meta-analysis and was therefore also evident in our meta-analysis.We also explored the potential factors responsible for heterogeneity using the sensitivity analysis and the subgroup analysis.The sensitivity analysis indicated that one study was an outlier, but further investigation revealed that the heterogeneity of our data was acceptable, and the combined effects were reliable.The subgroup analysis traced the different factors, such as circRNAs expression level, and showed that the function of circRNAs may be a major cause of heterogeneity.Aiding with clinical decision-making is one of the important key features of a novel biomarker.Therefore, likelihood ratios (negative and positive) and post-test probabilities are two useful parameters for medical professionals, because they provide information about the likelihood that a patient with a positive or negative test actually has MM or not.This study demonstrated the clinical applicability of two positive and negative likelihood ratio indices in the diagnosis of MM.PLR > 10 and negative likelihood ratio NLR < 0.1 indicate good diagnostic accuracy of test [23,24].In addition, the Fagan nomogram was used to describe the post-test probabilities of disease in the MM patients.If the prior probability of MM is 20%, the post-test probability of MM would reach 46% if the circRNA test is positive, and if the circRNA test is negative, this would mean that the post-test probability of MM would drop to 6%.
For the final interpretation of the clinicopathological features, the RR was chosen for the report because, if the odds ratio were reported, the association between cir-cRNAs and clinicopathological features would be exaggerated [60].Our results show a small but significant association between aberrant expression of circRNAs and elevated ISS and DS stages and B2-MG, which indirectly reflect the status of MM patients.Furthermore, the presence of del(17p) and t(4;14) has a small but significant association with abnormal circRNA expression.

Conclusion
According to the importance of MM diagnosis and the determination of the prognosis for effective management, our review suggests measuring the changes in the expression of circRNAs as a specific and valuable marker related to the prognosis and diagnosis of MM.Also, the changes in the expression of circRNAs can be associated with poor clinicopathological features and can be used as valuable markers for investigation of treatment effectiveness and clinical diagnosis.Through future studies, circRNAs can be considered important targets for the efficient treatment of MM.

Limitations of the review
However, our current meta-analysis still had the following limitations: First, the studies were all from China, which may circumscribe the generalization of these findings and lead to population selection bias.The second is the lack of access to the cutoff to check the threshold effect.Third, heterogeneity is still a vital issue in this meta-analysis, although we carried out subgroup analysis to explore possible sources.Fourth, the number of included studies is relatively small, which may give the appearance of bias.Fifth, due to the small number of studies, individual analysis of more subgroups was limited.The sixth reason is that articles with positive results are more likely to be published, which may increase overall diagnostic accuracy and the seventh, Due to the linguistic restrictions we only accepted articles in English (at least in the abstract), which may have influenced our results.
• fast, convenient online submission • thorough peer review by experienced researchers in your field • rapid publication on acceptance • support for research data, including large and complex data types • gold Open Access which fosters wider collaboration and increased citations maximum visibility for your research: over 100M website views per year

•
At BMC, research is always in progress.

Learn more biomedcentral.com/submissions
Ready to submit your research Ready to submit your research ?Choose BMC and benefit from: ? Choose BMC and benefit from:

Fig. 1
Fig. 1 The PRISMA flow diagram for the study selection process s nomogram (to describe the post-test probabilities of disease in MM patients) (Additional file 3: Fig S1), the likelihood ratio scattergram (Additional file 3: Fig. S1/Fig.1A), and the Probability Modifying Plot (Additional file 3: Fig. S2/Fig.1B) have been used in the clinical application of circRNAs.

Fig. 2 Table 3 Fig. 3
Fig. 2 Forest plots for the prognostic value of circRNAs in overall survival (OS) of MM patients.Oncogenic circRNAs (High-expressing) indicate worse prognosis (A) and tumor suppressor circRNAs (Low-expressing) indicate good prognosis in the MM patients (B)

file 5 :
Fig S1/ Fig. 1A C (Egger's test, P values of 0.752 and 0.505, respectively).The Trim and Fill method was used to better estimate the potential effects of publication bias, and like Egger's test, publication bias was not significant (Additional file 5: Fig. S1/Fig.1B, D).The sensitivity analysis was performed in the 2 subgroups of oncogene and tumor suppressor, and there was no outlier study (Additional file 5: Fig.S2/Fig.2A,B), indicating our results were not significantly to be affected by any individual of the included studies.

Table 1
The role of circRNAs in the development of MM and the impact of therapy response Author's

Table 2
Main characteristics of the prognostic studies a Multiple myeloma; b Overall survival; c Hazard ratio; d 95% confidence interval; e Newcastle-Ottawa Scale; f Peripheral blood * Months

Table 4
Subgroup analysis for diagnostic meta-analysis a Positive likelihood ratio; b Negative likelihood ratio; c Diagnostic odds ratio; d The area under the receiver operating characteristic curve

Table 5
Correlation between circRNAs and clinicopathological features of MM P values less than 0.05 are shown in bold a Beta 2 Microglobulin; b Durie-Salmon stage; c International Staging System; d Deletion