Cell-Free Circulating DNA Integrity Based on Peripheral Blood as a Biomarker for Diagnosis of Cancer: A Systematic Review

Cancer is a major health problem all over the world, with more than 1,685,000 new cancer cases estimated to be diagnosed in the United States in 2016 alone (1). Cancer has become the first leading cause of death in developed countries and the second leading cause of death in developing countries with projected 595,690 deaths from cancer deaths to be occurred in the United States just in 2016 (1, 2).

Early detection of cancer is of great importance, as the cancer detection in early stage is directly related with the increased survival rates (3). Many diagnostic methods have been used to detect cancer, such as protein tumor markers, imaging techniques, and pathology biopsy. These techniques have been widely used in clinic but still have certain limitations. Proteins or peptides markers, such as AFP, CEA, and CA125, which are always adjunct markers, have disadvantages, as they have high false-positive rates among patients with early stages of cancer (4). Traditional imaging techniques, such as ultrasound, CT, MRI, or PET-CT, are based on the tumor morphologic features that would not detectable at early stages of cancer and are costly (5, 6). Tissue biopsy is the gold standard for diagnosing cancer. However, it has disadvantages as an invasive procedure that would lead to certain medical complications and limit its usage (7). Thus, the reliable methods for cancer diagnosis at its early stage are urgently needed.

In recent years, the investigation of circulating molecular markers in peripheral blood (“liquid biopsies”) is of great importance because of the advantages, like easily accessible, reliable, reproducible, and early detectable in cancer (8). Many blood-based biomarkers, such as circulating DNA, miRNAs, and circulating tumor cells, have been investigated for the diagnosis of cancer (9–14).

Circulating DNA is described as cell-free DNA (cfDNA), or circulating tumor DNA present in serum or plasma (15, 16). Elevated cfDNA concentrations have been observed in many cancer types (8). In general, the circulating cell-free DNA integrity (cfDI) is calculated as the ratio of longer DNA fragment concentration to shorter ones from the same specific genetic locus (direct cfDNA concentration ratio). Another calculation method was described by Wang and colleagues in 2003 (17). Each fragment got ΔCp values between the Cp value of a standard human genomic DNA and the Cp value of each fragment of each sample. Two ΔCp values of each fragments were subtracted to obtain ΔΔCp. DNA integrity was calculated using the formula e ^{(-ΔΔCp × ln (2))}. Many studies have been conducted to evaluate the potential application of cfDI as a diagnostic biomarker in different types of cancer.

However, the performance of cfDI as a biomarker for cancer diagnosis varied a lot in different studies. Therefore, we conducted a systematic analysis to better clarify the diagnostic value of cfDI. The aim of this systematic review is to explore the value of cfDI as a biomarker of cancer in peripheral blood, which, to the best of our knowledge, has not been performed before.

A comprehensive systematic literature search strategy was used to identify the articles studying the utility of cfDI for cancer diagnosis up to February 28, 2017. Databases included PubMed and Web of Science. The following keywords employed for literature retrieval were adopted: “cell-free DNA Integrity” or “circulating DNA Integrity” or “DNA Integrity” or “cfDNA Integrity” and “cancer” OR “tumor” OR “carcinoma” OR “adenocarcinoma” and “diagnosis” or “ROC curve” or “sensitivity” or “specificity.” For further relevant studies, the references of selected articles were also identified through manual search.

All studies included in this review were confirmed to be investigating cell-free DNA integrity as a biomarker for cancer detection with high quality. All publications were identified by two reviewers (J. Cheng and Q. Tang). Eligible studies were included if they fulfilled the following criteria: (i) studies enrolled more than 10 patients and reported the utility of cfDI for diagnosis in cancer; (ii) sufficient data to support sensitivity and specificity or true-positive (TP), false-positive (FP), true-negative (TN), and false-negative (FN) data; (iii) all studies based on blood-based biomarkers in human cancers. Exclusion criteria were the following: (i) publications not related to the diagnosis of cfDI; (ii) studies unrelated with cfDI or without valid data; (iii) studies published in following formats: editorials, letters, case reports, or reviews. On the basis of the actual situation of cfDI, we calculated the diagnostic ability of cfDI of two different directions separately. All of the literature in line with the above criteria was considered to be analyzed as qualified studies. All of the studies were approved by the institutional review board, and all participants gave their informed consent.

All the included studies were carefully reviewed by two reviewers (J. Cheng and Q. Tang). The details of all studies, such as authors, publication year, country of the study, ethnicity of the subjects, sample size of the study, source of control, mean age of patients and controls, type of cancer, stage of cancer, DNA extraction kits, type of gene, qPCR results validation methods, cfDI calculation methods, cfDI index results, types of specimen, sensitivity and specificity, TP, FP, TN, and FN data were extracted.

We used the statistical software Meta-Disc to perform all statistical analyses. The sensitivity and specificity data of cfDI associated with the diagnostic value of cancer were extracted from each study with the corresponding true-positive (TP), false-positive (FP), true-negative (TN), and false-negative (FN) data. The following measures with their 95% confidence intervals for each study, the positive likelihood ratio (PLR), negative likelihood ratio (NLR), and diagnostic odds ratio (DOR), were calculated (18, 19). The pooled PLR is defined as the ratio of positive outcome in cases with cancer while the pooled NLR indicates the ratio of positive outcome in those without cancer. Simultaneously, DOR, the odds of PLR to NLR, ranges from zero to infinity, which reflects a discriminatory performance. The sensitivity and specificity of each included study were used to plot the summary receiver operator characteristic (SROC) curve and the area under the SROC curve (AUC) was also calculated to evaluate perfect discriminatory ability to differentiate patients from normal controls (20, 21).

The selection flowchart of the literature research is presented in Fig. 1. The initial search returned a total of 214 articles, of which 67 duplicated publications were excluded. Of remaining 147 articles, 102 articles were exclude because of reviews (7), meeting abstracts (4), book chapter (1), patent declaration (1), and not related to our definition of cfDI (94). The remaining 41 articles were subject to the next evaluation, among which 4 articles were not diagnostic markers, 7 articles were not peripheral blood based, and 5 articles were related to mitochondrial DNA. Finally, the remaining high-quality 25 articles were appropriate for our system review (11, 17, 22–44). With sufficient sensitivity and specificity data of high quality, we selected 20 studies from 15 articles into further systematic review (11, 17, 22–24, 26–28, 31–33, 36, 37, 42, 43).

Main characteristics of the 40 studies from the 25 articles are shown in Table 1 by the order of the publication year. The publication year ranged from 2003 to 2017. In this review, all studies analyzed contained 2,803 participants of which cancer patients were 1,594 and controls were 1,395 (967 matched healthy individuals and 418 patients with other benign diseases). Of the 40 studies included, 8 were conducted in Asian populations and 32 in Caucasian populations. Different gene fragments were used to calculate DNA integrity (16 of ALU elements, 12 of β-actin, and 4 of APP). The lengths of base pair in long and short fragments were also different like 400/100 bp of β-actin, 384/106 bp of β-actin, 247/115 bp of ALU, 260/111 bp of ALU, and so on. All of the included studies employed the quantitative real-time PCR (qRT-PCR) method to detect the value of cfDI while 19 of them tested in serum and 21 in plasma. cfDI was calculated as the ratio of the concentration of longer fragments to the ratio of shorter fragments. According to the definition of cfDI calculation, as the short sequences are represented within the annealing sites of longer fragments, cfDI should theoretically vary between 0 and 1. Nevertheless, some studies observed a cfDI higher than 1 that were not reasonable.

Most of the studies found significantly elevated cfDIs in cancer patients compared with healthy individuals or benign controls. In the first article about diagnostic value of cfDI, Wang and colleagues in 2003 demonstrated a significant higher median cfDI index of 0.66 in the neoplastic group compared with 0.14 in non-neoplastic group (P < 0.0001; ref. 17). Since then, similar results were observed in other types of cancer like colorectal cancer (24), breast cancer (11, 39), head and neck cancer (22), acute leukemia (28), renal cell cancer (29), melanoma (31, 34), hepatocellular cancer (27, 32), bladder cancer (33), glioma (35), and prostate cancer (36, 41).

Interestingly, there were also some studies that observed a totally opposite direction of the results. In colorectal cancer (30), testicular germ cell cancer (26), breast cancer (11, 44), and hepatocellular cancer patients (42), cfDI was found to be significantly lower compared with healthy controls. Those results are labeled in bold in Table 1.

As preanalytic considerations of cfDNA, the process of specimen is of extremely vital in further cfDI analysis (45). Most of the studies here reported the detailed blood sample process; however, some are missing. Most studies indicated that blood was processed within 2 hours of withdrawal. Two-round centrifugation process was applied in 18 studies. Pinzani and colleagues applied the centrifugation protocol of 1,600 rcf for 10 minutes and then 14,000 rcf for 10 minutes (31). Seventeen studies showed one-round centrifugation at 1,000–3,000 × g for 10–20 minutes. We also summarized the different extraction kit of cfDNA from blood of every article. Most of the research chose QIAamp DNA Blood Mini Kit as the extraction kit. To verify the cfDI results, we analyzed the verification methods of qPCR results. We observed that some studies evaluated the validity of cfDI by gel electrophoresis, melting curve analysis, and amplification efficiency. The overall methodologic qualities of the selected studies were generally good and are summarized in Table 1 and Supplementary Table S1.

As further analysis to the diagnostic accuracy of cfDI in cancer, we selected studies with high quality into a pooled analysis. Because of the opposite directions of cfDI in cases, we divided the suitable studies into two groups to analysis separately. In total, 20 studies from 14 articles with a higher cfDI in case group were analyzed in Group 1. Three studies from three articles with a lower cfDI in case group were pooled in Group 2. Cutoff values here ranged from 0.11 to 0.63. There is no fixed cutoff value of cfDI in different cancers.

Forest plots of data from Group 1 and Group 2 on the sensitivity and specificity of cfDI in diagnosing cancer are shown in Fig. 2. In Group 1, the pooled sensitivity was 0.74 (95% CI: 0.72–0.77) and pooled specificity was 0.84 (95% CI: 0.82–0.86) as shown in Fig. 2. Bivariate random effects for cfDI yielded a pooled negative likelihood ratio (NLR) of 0.31 (95% CI: 0.23–0.42) and a pooled positive likelihood ratio (PLR) of 6.16 (95% CI: 3.06–12.42) of Group 1 in Supplementary Fig. S1. The NLR was 0.31, meaning the probability of being false negative is 31% in the tests, which is not low enough to rule out cancer. The pooled PLR value of 6.16 suggested that cancer patients have a nearly 6.2-fold higher possibility of being tested positive by cfDI compared with noncancer patients. However, in most circumstances, the likelihood ratios above 10 or below 0.1 are regarded as strong evidence to rule in or rule out diagnoses, respectively (46). Pooled diagnostic odds ratio (DOR) is 20.67 (95% CI: 9.10–46.94) as shown in Supplementary Fig. S1. Higher DOR value means a better discriminatory test performance (47). Furthermore, we generated the SROC curve that indicated an overall summary of tests. As shown in Fig. 3A, the bivariate SROC curves showed area under the curve (AUC) of cfDI was 0.814 for cfDI in 20 studies. It suggested that cfDI may perform well to differentiate cancer patients from controls with a relatively high accuracy.

In Group 2 of 3 studies from 3 papers, the pooled sensitivity was 0.58 (0.51–0.65) and pooled specificity was 0.78 (0.71–0.85; Fig. 4). As depicted in Supplementary Fig. S2, the pooled PLR was 3.82 (1.41–10.34) for cfDI when cfDI was low in case group. The pooled negative LR was 0.52 (0.44–0.61). Pooled diagnostic odds ratio was 7.14 (3.34–15.24). The bivariate SROC curves showed a pooled AUC of 0.742 as a marker for cancer diagnosis (Fig. 3B).

As the incidence and mortality are increasing, cancer is a major public health problem all over the world. Early diagnosis of cancer is prominent as it can improve the patients' chances of survival and decrease the mortality rate. Liquid biopsies nowadays hold great promises for its advantages like easily accessible, reliable, and reproducible (12). DNA concentration has been investigated for many studies. Several studies have found that higher DNA concentration in cancer patients compared with control. However, the lower amount of DNA concentration and wide variation of DNA concentration limits its usage. Now, more and more studies focus on the extent of cfDNA integrity that is more reliable and informative compared with DNA concentration. As a novel candidate of cancer diagnostic marker with high stability, cfDI poses the potential of a great biomarker. As the first evidence-based review concerning the overall diagnostic accuracy of cfDI in different cancers, we summarize the results of different studies discussing the cfDI values for diagnosis of cancers. In general, there are two methods of calculating cfDI (direct cfDNA concentration ratio and e ^{(-ΔΔCp x ln (2))} methods). Overall, these two methods are quite similar, because gene expression fold change can be calculated as 2^−ΔΔC_t. However, because of the different treatment of calculation from the standard DNA concentration, the results of two methods were not always the same (39). Most studies have employed the method of direct ratio of gene expression.

In our review, we observed two different directions of altered cfDI between cancer patients and controls. It is controversial; however, it can achieve significance from patients to controls. Opposite direction of cfDI in case–control studies is unacceptable for a qualified clinical marker. So the actual underlying mechanisms are still not clear. Originally, it is generally assumed that the DNA fragments from malignant cells that underwent nonapoptotic cell death would have larger size (long fragments), whereas the size of DNA fragments from apoptotic cells is mainly about 180–200 base pairs (short fragments). It was hypothesized that cfDNA fragments were mainly released by apoptotic cells in healthy controls, whereas in cancer patients, cfDI was regarded as predominantly released by malignant cells undergoing pathophysiologic processes like necrosis and autophagy (48). However, necrotic DNA has been proven to account for a very small fraction of cell-free DNA in cancer patients (25).

On the other hand, Giacona and colleagues in 1988 found that cfDNA from healthy individuals had 3- to 5-fold multiples of nucleosome-associated DNA length compared with cfDNA in pancreatic cancer patients by gel electrophoresis (49). Recently, by massive parallel sequencing, Jiang found elevated amounts of shorter mitochondrial DNA molecules in plasma of carcinoma patients compared with healthy subjects (50). Many studies observed an enrichment of 166-bp fragments corresponding to nucleosomal DNA in the cancer patients (51–55). Thus, more studies needed to explain this biological significance of cfDI. Hence, it showed the utmost disadvantage of cfDI as a diagnostic marker. If this problem cannot be solved, cfDI can hardly be a biomarker for clinical usage.

As for cfDNA, many studies have debated the usage of serum or plasma. cfDNA concentration was found to be higher in serum than in plasma (56), and serum has a low level of contaminating extraneous DNA released from leukocytes. However, studies observed the coagulation process affecting the spectrum of circulating nucleic acids in serum (57). Sample preparation could also affect the cfDI results significantly (58). Different centrifugation conditions were applied. Increased studies conducted a two-round centrifugation with a low speed first and a fast speed later. The second step of high-speed centrifugation can remove cell debris significantly (11). As extraction methods were quite important, most of the studies used QIAamp DNA Blood Mini Kit that has been proven to be a reliable extraction kit (59). To verify the cfDI results, the confirmation of qPCR results were also summarized. Gel electrophoresis is the direct evidence to confirm the qPCR products. Other studies adopted melting curve analysis, amplification efficiency, and calibration curve as different verification methods to make cfDI accurate and authentic. Other factors such as different conditions of laboratory, different cut-off criteria, and individual difference can also produce influences. However, because of the great heterogeneity of sample preparation in our review, it is impossible to analyze the specific conditions.

Our system review included 25 articles which is the first review to identify the overall accuracy of cfDI in cancer diagnosis. Through our strict screening process, we made sure only high quality studies were adapted to further analysis. Limitations of our study should also be noted. Because of great heterogeneity between studies, it was difficult for us to do subgroup analysis. Because of limited subjects, there was not enough data for a specific type of cancer, which may elucidate our results as cancer is a heterogeneous disease. In our studies only 32 studies were included; therefore, further studies are needed for the final conclusion. Because few studies analyzed the difference of cfDI at different stages as the small case number, we also hope to summarize the cfDI difference at different stages in further studies. Last but not the least, the cut-off values inconsistently varied in different studies and were not available in some studies.

In summary, the current systematic review assessed cfDI as a promising noninvasive biomarker in diagnosing different cancers. Although the opposite direction was observed in the case–control study, the sensitivity and AUC were relatively high in each group. However, because of the great heterogeneity between different studies, more high-quality studies are needed to further explore its diagnostic accuracy in malignant diseases.

No potential conflicts of interest were disclosed.

This work was supported by China Scholarship Council (grant number: 201306090129; to J. Cheng).

The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.