FormalPara Key Points

Liquid biopsy in blood is the major application of droplet digital polymerase chain reaction (ddPCR) technology in a wide variety of cancers for diagnostic, predictive, prognostic and monitoring purposes.

The use of ddPCR for liquid biopsy has increased in recent years in other biologic fluids, including cerebrospinal fluid, urine, stool, ocular fluids, sputum, saliva, bronchoalveolar lavage, pleural effusion, mucin, peritoneal fluid, fine needle aspirate, bile or pancreatic juice.

By far the most utilized ddPCR platform to date is the Bio-Rad QX100/200 system. However, new ddPCR platforms have been recently developed.

1 Introduction

Droplet digital polymerase chain reaction (ddPCR) is a molecular biology technique based on sample partitioning into thousands of nanoliter-sized droplets where individual PCR reactions take place, allowing for the detection of very low abundance molecular targets with extremely high sensitivity [1]. Although this technology is nearly 10 years old [1, 2], its use has increased substantially in recent years, particularly in the field of precision oncology, with hundreds of publications demonstrating its clinical utility in many different kinds of malignancies. Liquid biopsy, defined as the analysis of molecular biomarkers in a wide variety of body fluids with diagnostic, predictive, prognostic or monitoring purposes, represents a noninvasive (or minimally invasive) approach with  significant relevance in the management of patients with cancer that requires the implementation of extraordinarily accurate detection methods [3]. In this clinical scenario, ddPCR has gained much attention as a powerful tool for the detection of genetic alterations, including single nucleotide variants, copy number variations, genomic rearrangements and methylation biomarkers, mainly in blood (particularly in plasma and serum) but also in many other biologic fluids, such as cerebrospinal fluid, urine or stool, among others [4]. In this updated review, we describe some recent developments in ddPCR platforms and how they are being applied in oncology. Then, we focus our attention on the major applications of ddPCR in liquid biopsy, particularly in other “alternative” biofluids that are still less frequently used than blood but are gaining increasing interest in different types of cancers.

A literature research for this review was performed in PubMed using the following search strategy: ("Polymerase Chain Reaction"[Mesh:NoExp] OR "Multiplex Polymerase Chain Reaction"[Mesh] OR "PCR"[tw] OR "Polymerase Chain Reaction"[tw]) AND (("droplet based" AND "digital") OR "droplet based digital" OR "droplet digital" OR "bio-rad"[tw] OR "biorad"[tw] OR "raindance"[tw] OR "stilla"[tw] OR "digital droplet") AND (cancer[sb])AND "2016/12/01"[Date - Publication] : "3000"[Date - Publication].

2 Recent Developments in Droplet Digital Polymerase Chain Reaction Platforms

By far the most widely used ddPCR platform in the literature is the Bio-Rad platform (Bio-Rad; Hercules, CA, USA) (Fig. 1). This water-in-oil emulsion system for droplet generation has evolved from a manual workflow—where the user had to pipette the PCR reaction mix and the oil into the cartridges—to a more automated system, with the so-called AutoDG droplet generator, along with a change from the QX100 to the QX200 system. Other new ddPCR platforms have been developed in the last few years, such as the Naica Crystal ddPCR (Stilla Technologies; Villejuif, France) or the SAGAsafe® technology (formerly known as IBSAFE®; SAGA Diagnostics, Lund, Sweden). The main advantage of these newly developed systems is the increased multiplexing capabilities and improved sensitivity, with a lower limit of detection (LoD), respectively.

Fig. 1
figure 1

BioRad droplet digital polymerase chain reaction (ddPCR) platform. (1) Preparation of ddPCR reaction mix containing sample, probes and ddPCR master mix; (2) generation of droplets in the droplet generator by water-in-oil emulsion using vendor-specific oil; (3) droplets containing sample and ddPCR reaction mix; (4) transfer of droplets to a 96-well PCR plate; (5) the plate is run with PCR protocol in a ddPCR thermocycler; (6) droplet fluorescence is checked into the droplet reader; (7) analysis of results using QuantaSoft. Positive and negative droplets are plotted in a two-dimensional graph, setting thresholds for discrimination. Created with BioRender.com

Full size image

The Naica System is based on a principle initially published in 2013 by Dangla et al. [5]. This digital PCR platform relies on a hybrid approach (named crystal digital PCR [cdPCR]) that combines a two-dimensional array of microchambers for partitioning and the use of crystal droplets that are thermocycled and transferred to a fluorescence microscope to detect amplification. All these steps take place in a specifically designed microfluidic chip (the Sapphire chip, containing the preloaded emulsion oil, which represents an advantage because it simplifies the process and prevents contamination [6]), and two different instruments are involved: the Naica Geode for sample partitioning and thermalcycling and a three-color detection system consisting of an automated fluorescence microscope, the Naica Prism3. Results are analyzed using the Crystal Miner software (Stilla Technologies). Madic et al. [7] reported a detailed and comprehensive description of the system and its workflow. This paper also discussed issues related to data analysis and the application of this system to the detection of L858R, L861Q and T790M epidermal growth factor receptor (EGFR) mutations. Thus, the capability of three-color multiplexing of this platform was tested, showing that it did not result in a loss of sensitivity. When compared with massive parallel sequencing (MPS), the cdPCR showed better performance for the detection of known mutations in the plasma of patients with metastatic non-small-cell lung cancer (NSCLC) [8]. In this comparison, the authors found 11 positive plasma samples with cdPCR that could not be detected with MPS, with mutant allele fractions between 0.09 and 7.9%. In longitudinal plasma samples collected for monitoring EGFR along the disease trajectory, MPS also reported six negative samples that digital PCR found to be positive.

In 2018, a customized six-color assay was developed for detecting and quantifying 19 prevalent EGFR sensitizing and resistance mutations using the Naica cdPCR and an inverted Nikon eclipse TI microscope (Nikon Instruments Europe, France) with an appropriate selection of filter sets for reading fluorescence in six different detection channels [9]. The LoD varied, depending on the mutation assayed, between 0.125% (p.C797S c.2389 T > A and p.C797S c.2390 G > C) and 0.0975% (exon 19 ins/del), with 0.25% for EGFR resistance mutation p.T790M and a range of 0.125–0.25% for activating mutations in the Cy5 detection channel. Sensitivity was also tested in tumor and plasma samples from 82 patients, comparing results with those from next-generation sequencing (NGS) and the three-color system. These comparisons showed good correlation, especially for the concentration of mutant copies per microliter and mutant allele frequency (MAF) in plasma samples measured by six- and three-color digital PCR. Longitudinal samples from four patients were also analyzed to monitor the course of disease, and fluctuations in mutant DNA levels were consistent with clinical evolution. This six-color assay for EGFR mutation detection was subsequently optimized on a prototype of a six-color reader instrument and a prototype of six-dimensional Crystal Miner software that are integrated in the new six-color cdPCR prototype platform, with  the  commercial version due for launch in late 2021 [10].

Song et al. [11] also recently developed an integrated digital PCR assay using the three-color version of the Stilla platform. The assay is called dEGFR39 and is the first that allows the screening and monitoring in plasma of all the EGFR mutations known to be clinically relevant in NSCLC for treatment guidance and prognosis. It simultaneously detects 39 mutations of exons 18–21 of this gene, including not only the most frequently identified variants such as L858R, 19Del and T790M but also other less common mutations including L861Q, S768I, G719X, C797S and 20 insertions [11]. This study analyzed the formalin-fixed paraffin-embedded (FFPE) tumor tissue and plasma of patients with NSCLC (N = 30 and N = 33, respectively) and demonstrated that dEGFR39 could detect EGFR mutations with a sensitivity of 0.308 copies/μL and an accuracy of 88.87% (for dEGFR39 in plasma and amplification-refractory mutation system [ARMS] in FFPE), showing a direct association between mutational load and response to treatment. It also anticipated disease progression by detecting T790M mutations earlier than other methods such as SuperARMS PCR and computed tomography (CT) imaging.

It is evident that the main application of cdPCR to date has been detection of EGFR alterations in NSCLC, but research is also ongoing for detection of PIK3CA mutations in plasma in patients with advanced breast cancer, aimed at the selection of alpelisib treatment. As presented in the European Society for Medical Oncology Breast Cancer virtual meeting in May 2020, the three-color detection system of Stilla was used with a multiplex assay for detection of 26 mutations located in exons 4, 7, 9 and 20 of the PIK3CA gene [12]. Quantification of BRAF V600E has also been performed using a novel DNA reference material that is intended to mimic circulating tumor DNA (ctDNA). Low levels of BRAF V600E ctDNA reference material were tested in eight different laboratories, seven of them with the QX200 ddPCR platform from Bio-Rad and one with the Stilla Naica Crystal ddPCR system [13]. Results from the interlaboratory study showed a significant difference in mutant and wild-type copy number concentration between the only laboratory using the Naica platform and the other seven laboratories. This inconsistency between the two platforms was improved considerably by correcting the droplet volume (the droplet volume measured by the authors of this study was 0.476 ± 0.008 nL vs. the 0.44 nL estimated by the manufacturer; this difference of 8.3% was suspected to be the reason for the overestimation observed in both mutant and wild-type copies).

Finally, the use of this platform has also been reported for chimerism monitoring of post-allogeneic hematopoietic stem cell transplantation for the treatment of hematological malignancies, including acute myeloid leukemia (AML) and acute lymphoblastic leukemia [14]. In this study, cdPCR was compared with NGS, with both methods reaching a sensitivity of 0.1%. The results in terms of percentage of chimerism in cdPCR and NGS showed a high concordance with those obtained by the reference techniques (ddPCR, short tandem repeat and quantitative PCR [qPCR]).

The SAGAsafe® platform is a droplet-based proprietary methodology with an improved LoD of ~ 0.001% MAF. It consists of a two-step process that takes place sequentially within the droplets: linear amplification or copying of the target sequence followed by a limited exponential signal generation. This technology minimizes polymerase base-incorporation errors, enhancing true-positive signals while simultaneously reducing the number of false positives, thereby achieving greater sensitivity and specificity [15, 16].

It has already been used to detect TP53 mutations in liquid-based Pap samples from patients with ovarian cancer [17]. In this study, IBSAFE detected TP53 mutations using a custom-designed assay with very high sensitivity (MAF of 0.0068%) in samples with low DNA input (as little as 0.17 ng). The in-sample LoD was reported to be 1 in 50,000. Bio-Rad ddPCR was used as a control, but the small number of samples tested did not allow a direct comparison between the platforms.

Another study reported that IBSAFE detected somatic mutations in plasma ctDNA from patients with breast cancer, showing a small average increase in ctDNA levels in both peripheral and central blood following mammographic breast compression, with no apparent clinical relevance [18]. Additionally, IBSAFE technology has been applied to the detection of EGFR, KRAS and BRAF mutations in the preoperative plasma of patients with lung adenocarcinoma [15]. Finally, the most recent work using this platform demonstrated its applicability for the detection of minimal residual disease in AML [16]. Between five and nine mutation assays were developed for each patient using this technology. The method was more sensitive in identifying residual disease than multicolor flow cytometry, detecting the targeted mutations in all relapsing patients and allowing the tracking of early recurrence in leukemic subclones, revealing the existence of three different mutational patterns of relapse.

SAGA Diagnostics launched its first in vitro diagnostic (IVD) European Conformity (CE)-marked SAGAsafe® kit for EGFR T790M testing in 2020. The company also presented a combined strategy of NGS of tissue to detect chromosomal rearrangements and a digital PCR fingerprint in plasma, called SAGAsign® (formerly known as KROMA) [19]. Two recent references in the literature to another platform should also be mentioned: the MicroDrop-100 ddPCR system (Forevergen, China) is based on water‐emulsion droplet technology and has been used to detect BRAF V600E mutations in thyroid nodules, with better performance than ARMS-PCR [20], and also  to measure the effect of CSNK2A3 expression on hepatitis B virus infection in hepatocarcinoma cells in vitro [21].

Studies reporting clinical applications of these emerging ddPCR platforms in cancer are summarized in Table 1.

Table 1 Clinical applications of recently developed droplet digital polymerase chain reaction platforms in oncology
Full size table

3 Applications of ddPCR in Liquid Biopsy

3.1 Liquid Biopsy in Blood

Hundreds of papers published in the last 4 years have supported the clinical utility of analyzing genetic biomarkers using ddPCR in blood in the field of oncology. It is being tested for application in the monitoring of molecular response and minimal residual disease in hematologic malignancies [22,23,24,25,26,27,28]. In many cases, both blood and bone marrow aspirates are used for liquid biopsy, and BCR-ABL is frequently the biomarker of choice [26, 28,29,30,31,32,33,34], and this  has also been suggested as a useful tool to predict and assess the outcomes after discontinuation of treatment with tyrosine kinase inhibitors (TKIs) [32, 35, 36]. The QXDx BCR-ABL %IS (Bio-Rad) ddPCR assay is the first commercially available ddPCR-based IVD product with US FDA clearance and the CE mark. This assay can detect the e13a2 and e14a2 fusion transcripts (but not e1a2, e19a2 or other rare transcripts) and has an acceptable analytical performance, with results comparable to those of the CE-IVD-marked ipsogen BCR-ABL1 Mbcr IS-MMR (Qiagen, Hilden, Germany) real-time qPCR (RT-qPCR) assay [37]. The main advantages of ddPCR versus the gold standard RT-qPCR method include a superior sensitivity and accuracy for ddPCR (with a LoD of one copy of transcript), as well as the ability to perform an absolute quantification without standard curves. Disadvantages include the lack of standardized methods and its limited availability in laboratories [38]. Longer turnaround times (due to droplet generation and reading) and the possibility of false positives have also been suggested as potential limitations of this technique [37]. We found no consensus in the literature regarding cost and throughput concerns. Alu methylation status has also been quantified by ddPCR in bone marrow samples from patients with chronic lymphocytic leukemia, myelodysplastic syndromes and myelomonocytic leukemia before and after treatment with 5-azacytidine. Decreased levels of this epigenetic marker have been observed after hypomethylating therapy, suggesting a potential utility of this approach for molecular monitoring of response [39]. Other biomarkers include NPM1 [22] and IDH2 [40] mutations, WT1 levels [25] and immunoglobulin heavy chain (IGH) gene [41] or immunoglobulin kappa-deleting-element (IGK‐Kde) [42] rearrangements. Table 2 summarizes some data on these studies.

Table 2 Minimal residual disease detection by droplet digital polymerase chain reaction in liquid biopsy for hematologic malignancies
Full size table

As previously mentioned, this technology has also been utilized to quantify engraftment after hematopoietic stem cell transplantation [14, 43,44,45,46], to monitor the expansion and persistence of chimeric antigen receptor (CAR)-T cells in vivo, reflecting response rates and side effects [47,48,49,50,51,52,53], and to detect vector copy number in clinical CAR/T-cell receptor (TCR)-T-cell products [54, 55].

Genetic biomarkers have been analyzed from different sources, including circulating tumor cells (CTCs) [56,57,58], cell-free DNA (cfDNA) [59,60,61] and cell-free RNA or extracellular RNA [62], nucleic acids derived from exosomes [63,64,65,66,67,68] and extracellular vesicles (EVs) [69,70,71], including long noncoding RNAs [72,73,74,75], microRNAs (miRNA) [76,77,78,79,80] and messenger RNA (mRNA) [57, 81,82,83], isolated from whole blood, plasma or serum (Fig. 2). Many studies have analyzed a combination of several of these genetic materials [84,85,86].

Fig. 2
figure 2

Different sources of genetic biomarkers isolated from whole blood, plasma or serum that can be analyzed by droplet digital polymerase chain reaction (ddPCR): circulating tumor cells (CTCs), extracellular RNA (exRNA), cell-free DNA (cfDNA), cell-free RNA (cfRNA), long non-coding RNA (lncRNA) and messenger RNA (mRNA). Created with BioRender.com

Full size image

ddPCR has been applied in a long list of cancer types, headed by colorectal [60, 87,88,89,90,91,92,93,94,95,96,97,98] and lung cancer, particularly NSCLC, with EGFR and KRAS mutations being the most analyzed markers because of their relevance for therapeutic management of patients, particularly for the selection and monitoring of treatment with TKIs [61, 99,100,101,102,103,104,105,106,107,108,109,110] (Table 3). These are without a doubt the most widespread clinical applications of ddPCR in the field of oncology. Pancreatic cancer [56, 59, 65, 76, 78, 111,112,113,114,115,116,117], breast cancer [82, 118,119,120,121,122,123,124], melanoma [58, 125,126,127,128,129,130,131,132,133,134], prostate cancer [57, 81, 135,136,137,138,139,140,141,142,143] and ovarian cancer [144,145,146,147,148] are also among the most studied neoplasms using ddPCR for clinical purposes. Other less frequent malignancies such as gastrointestinal stromal tumors [149, 150] or peritoneal metastasis from colorectal origin [151, 152] have also benefited from ctDNA profiling in plasma by ddPCR. It is not the aim of this review to cover all the different applications of ddPCR in every cancer type. Notably, in most of these studies, fluctuations in ctDNA levels detected by ddPCR mirrored disease response and predicted recurrence before clinical evidence. Circulating tumor markers detected by ddPCR have also shown a clinical value for early detection of disease and had prognostic implications in many of these studies.

Table 3 Main clinical applications of droplet digital polymerase chain reaction for liquid biopsy in colorectal cancer and non-small-cell lung cancer
Full size table

In many studies, ddPCR was applied in combination with NGS assays, which provide a broader perspective of the tumor mutational landscape, whereas ddPCR focuses on single molecular targets that allow confirmation of the presence of these variants and detection of changes in ctDNA levels over the disease course to track tumor burden [23, 106, 151, 153,154,155,156,157]. This strategy gives the clinician the opportunity to monitor the response to treatment and guide therapeutic decisions, anticipating relapse even months ahead of the emergence of clinical symptoms or evidence in imaging techniques. A remarkable number of studies have demonstrated the utility of ddPCR to unveil intratumoral heterogeneity [158] and clonal evolution in response to treatment [61, 150]. Of note, ddPCR has also proven useful for the detection of nonmalignant mutations present in hematopoietic cells (a phenomenon called clonal hematopoiesis) that can give rise to a confounding false-positive plasma result when non-hematopoietic cancers, such as NSCLC, are monitored using liquid biopsy [159]. In recent years, a trend towards multianalyte or multifactorial models has increased, where several biologic markers are simultaneously analyzed, with ddPCR playing a significant role as an accurate and reliable tool for quantitative nucleic acid-based biomarkers [160,161,162].

ddPCR is usually employed as a validation technique when alternative or newly developed methodologies are tested or to solve discordant cases [163–167].

It should also be noted that there are still some challenges and/or limitations to a more widespread use of ddPCR in routine clinical practice. The number of genetic variants that can be analyzed is limited by the amount of sample available. Multiplexing strategies have been developed to overcome this problem, including additional detection channels in the newer ddPCR platforms, as mentioned earlier. Besides, more than one target can also be detected in a single fluorescent channel by varying the concentration of different probes labeled with the same fluorophore or using amplicons of different sizes marked with DNA binding dyes [168].

Another relevant hurdle is the occurrence of false-positive signals, mainly caused by PCR errors in cfDNA samples, where a rare variant is intended to be detected in a high background of a wild-type allele, limiting the fractional abundance that can be detected [168].

First, standardized protocols for sample collection, storage, processing, nucleic acid extraction and modification are lacking. A range of tubes are used for blood collection, along with a range of anticoagulants and/or conservative compounds (EDTA, Streck, CellSave). The time from collection to processing and the temperature and storage conditions also deserve consideration. A huge diversity of protocols is found in the literature, from one-step to two-step centrifugation, with variable speeds, times and temperatures. The starting volume of plasma or serum for extraction varied from < 1 to ≥ 5 mL. Commercial kits specifically developed for circulating nucleic acid isolation are the most frequent choice, with protocols described by the manufacturers, but elution volume or the method of nucleic acid quantification, for example, usually differ between studies. It is not the aim of this review to delve into isolation methods, but heterogeneity is high and, remarkably, many of these preanalytical steps are of the utmost importance and can lead to measurement errors. ddPCR results are severely affected by factors such as DNA purity or concentration, hence all these procedures still require further optimization.

Other associated challenges refer to data analysis, mainly thresholding setting, and particularly when “rain” occurs (the presence of partitions located between positive and negative populations). Several software tools (both platform specific and independent) have been developed in recent years in response to this problem. The Minimum Information for Publication of Digital PCR Experiments (dMIQE) guidelines were published in 2013 [169] and an update presented in 2020 (dMIQE2020) [168]. These guidelines highlight the key experimental information that should be provided by researchers and helps in understanding every step of the experimental process, from assay design to validation and performance, why they require standardization, and how this could be achieved.

3.2 Other Biological Fluids

The vast majority of studies using ddPCR involve the detection of genetic alterations in plasma. However, as previously reviewed [4], other body fluids can also be used for noninvasive or less invasive molecular examination, including cerebrospinal fluid (CSF), urine or stool. Recent studies have investigated the use of additional body fluids for diagnostic purposes, including saliva and ocular fluids, such as vitreous fluid (VF) and aqueous humor (AH). Of note, the use of these biofluids is not yet widespread in the clinic, and many of these approaches are still under development in research studies. Also, the different targets analyzed in these alternative sources of nucleic acids requires the adjustment of isolation procedures. Limiting factors include the amount of sample collected (e.g., saliva or sputum) or the concentration and/or purity obtained (e.g., in stool).

3.2.1 Cerebrospinal Fluid

IDH1 mutations were among the first genetic alterations detected in CSF with ddPCR, and further research has been undertaken, particularly in lower grade gliomas [170]. However, MYD88 L265P (a myeloid differentiation primary response gene 88 single-base substitution at c.794T > C resulting in a leucine to proline amino acid change) is another hotspot mutation that has recently gained much more attention, with a number of publications accumulating evidence about its clinical value. The major application of MYD88 L265P detection by ddPCR in CSF is for minimally invasive confirmation of a diagnosis of primary central nervous system (CNS) lymphoma (PCNSL) [171,172,173] (a case of secondary CNS lymphoma has also been reported [174]) and of Bing–Neel syndrome [171, 175]. Interestingly, in some cases, this mutation has been detected in cfDNA from CSF at a higher frequency than in cellular DNA [172], and this might provide more sensitivity than cytomorphology and NGS in samples with low cellularity and very low DNA content [171]. In a very recent study, CSF testing by ddPCR for MYD88 L265P proved useful for the detection of early relapse in a patient with Bing–Neel syndrome treated with ibrutinib, showing an increase in mutation levels 2 weeks before the appearance of clinical symptoms and without evidence of recurrence on magnetic resonance imaging (MRI), CSF cytology, flow cytometry analysis and immunofixation electrophoresis [175]. In line with these observations, Bobillo et al. [176] recently combined NGS and ddPCR (covering many tumor mutations apart from the aforementioned MYD88 L265P) and reported better detection of CNS lesions by measuring ctDNA in CSF than flow cytometry, cytology and plasma measurements. This work showed the great potential of CSF-based liquid biopsy using ddPCR to monitor CNS involvement of B-cell lymphomas, predicting CNS relapse and detecting residual disease [176].

However, a previous study including three patients with spinal ependymoma suggested that anatomic sequestration or a low grade of intramedullary spinal cord tumors might hamper the utility of CSF-based liquid biopsy in these malignancies [177].

In recent years, liquid biopsy by ddPCR has gained increasing interest as a powerful tool to detect genetic biomarkers in pediatric brain tumors, particularly in pediatric diffuse midline glioma (DMG), as previously reviewed by Lu et al. [178] and Azad et al. [179]. A recurrent actionable mutation in histone 3, affecting either the H3.1 or the H3.3 protein at lysine position 27 (HIST1H3B K27M and H3F3A K27M, respectively, also known as H3K27M mutation), has been detected using ddPCR assays in the CSF of patients with diffuse intrinsic pontine glioma (DIPG) or high-grade glioma [180–183]. In a study by Panditharatna et al. [181], H3K27M and other obligate partner mutations in ACVR1, PIK3R1 and BRAF genes were detected by ddPCR in the CSF of patients with DMG. Their results also suggested that fluctuations in ctDNA levels in serial plasma samples might have clinical utility to monitor treatment response in patients with DIPG, comparable with MRI evaluation [181]. A recent study also showed the potential of ddPCR determinations in CSF for disease monitoring in pediatric patients with medulloblastoma. Again, a combined strategy of NGS and ddPCR allowed the detection of a wide variety of tumor mutations in CSF, highlighting that it both represents a better source of ctDNA than plasma and has superior sensitivity compared with cytology [184]. Thus, mutations identified by NGS and validated by ddPCR efficiently detected minimal residual disease and shed light on genomic tumor evolution, revealing intratumor and interlesion heterogeneity since this approach was able to unveil the existence of two completely different tumors at diagnosis and relapse.

Apart from tumors directly affecting the CNS as a primary target, the analysis of genetic alterations in CSF has also been applied for detection of metastatic disease in the CNS from other tumor origins, such as leptomeningeal or brain metastasis derived from lung adenocarcinoma [185,186,187], breast cancer [188] or melanoma [189].

3.2.2 Urine

Several publications have reported the application of ddPCR to the assessment of molecular biomarkers in urine for  prostate and bladder cancer. The detection of methylation biomarkers (GSTP1, APC, RASSF1A, PITX2 and C1orf114) in the cell fraction isolated from urine using a filtration device [190] or the development of a two-gene panel (PCA3, PCGEM1) in urinary exosomal mRNA [191] are some of the strategies that have been explored to improve the noninvasive identification of high-grade prostate cancer. The urinary transcriptome was proposed as a valuable source of biomarkers and later validated by ddPCR in a recent study showing that five protein-coding genes (FTH1, BRPF1, OSBP, PHC3 and UACA) distinguished patients with prostate cancer from cases of benign hyperplasia and healthy subjects, both in the centrifuged and non-centrifuged fraction of small-volume urine samples (1 mL) [192]. Similarly, another novel urinary mRNA signature (including three upregulated genes [PDLIM5, GDF-15 and THBS4] and three downregulated genes [UPK1A, SSTR3 and NPFFR2]) was developed using ddPCR to discriminate prostate cancer from benign prostatic hyperplasia within the “prostate-specific antigen [PSA] gray zone” (3–10 ng/mL total PSA) [193]. The expression of the androgen-receptor splice variant 7 (AR-V7) has also been reliably quantified in urine-derived EVs, with higher levels in patients with castration-resistant prostate cancer than in those with hormone-sensitive tumors [194].

On the other hand, the analysis of hotspot mutations in TERT promoter and FGFR3 by ddPCR in urine has been proven useful for the early detection of urothelial cancer, including upper tract urothelial cell carcinoma and bladder cancer [195,196,197,198,199,200]. Tumor-specific mutations in FGFR3 and PIK3CA hotspot mutations, among many others, have been measured in the plasma and urine of patients with bladder cancer for disease and treatment monitoring, showing a remarkable potential to detect early signs of metastasis [201, 202]. The combination of ddPCR and urine cytology yields a higher sensitivity than cytology alone (UroVysion) for detection and prognosis in urothelial bladder cancer [198]. Specific ddPCR assays for TERT promoter mutations have shown results comparable to those with the UroMuTERT NGS-based assay for detection of MAFs > 2%, both in the urine supernatant cfDNA and the urine pellet cellular DNA, although some discrepancies have been found below this allelic fraction [200]. Previous studies suggested that ddPCR may have a limited sensitivity to detect low-grade tumors harboring very low MAFs [195].

PIK3CA p.H1047R mutation has been detected by ddPCR in the urine of a patient with CLOVES (congenital lipomatous overgrowth with vascular epidermal and skeletal anomalies) syndrome, a subgroup of the PIK3CA-related overgrowth spectrum (PROS), who had a Wilms tumor. The mutation was present not only in the affected tissue but also in urine and in the Wilms tumor, which had been resected upon diagnosis 26 months prior to the urine sample collection. These results suggest that urine may be useful for noninvasive mutation screening by ddPCR in patients with CLOVES syndrome, some of whom are candidates to develop Wilms tumor [203]. This suggestion was supported by another study involving patients with PROS that analyzed several kinds of biologic specimens, including plasma, whole blood, saliva, buccal swabs and urine (at its cellular and cfDNA fraction) [204]. Three different PIK3CA variants [c.3140A > G; p.(His1047Arg), c.3140A > T; p.(His1047Leu) or c.1624G > A; p.(Glu542Lys)] were detected in this wide variety of tissues, with the exception of leukocytes (only one case was detected in saliva and the corresponding buccal swab from a clinically affected cheek). Interestingly, patients who had a positive variant of PIK3CA in DNA extracted from the cellular fraction of urine also presented this variant in urine cfDNA, and these levels were much higher in patients with a history of nephroblastomatosis or Wilms tumor than in individuals without known renal involvement. Thus, urine testing by ddPCR could provide information about the renal involvement in PROS, and multiple tissue analysis may help identify patients at high risk for Wilms tumor.

The usefulness of urine as a suitable source of genetic material for molecular analysis in cancers not related to the genitourinary tract was also further explored in pancreatic ductal adenocarcinoma (PDAC) [205], metastatic colorectal cancer (CRC) [206] and lung cancer, particularly NSCLC [207]. KRAS mutations have been examined in the plasma and urine of patients with PDAC, showing a similar detection rate and sensitivity in both fluids, although they are influenced by renal function degeneration [205]. ddPCR has also found KRAS and BRAF mutations in matched plasma and urine samples from patients with metastatic CRC [206]. EGFR mutations have been measured in the urine of patients with NSCLC by ddPCR at different time points after curative intent surgery for disease monitoring, with the aim of detecting relapse and minimal residual disease [207]. Thus, in this study, the presence of detectable mutant DNA in the urine samples post-treatment was associated with disease recurrence, whereas patients with undetectable levels had better disease-free survival. In another study, matched plasma and urine samples from patients with NSCLC were collected and analyzed for EGFR mutation detection by ddPCR after TKI therapy [208], demonstrating that both body fluids may provide complementary information. Baseline plasma values showed better positive predictive value, whereas urine samples seemed to be more useful for serial monitoring since changes in secondary EGFR T790M mutation levels were detectable earlier. It was concluded from both types of samples that patients with higher post-treatment values than at baseline had poorer outcomes (the majority were T790M positive), but urinary cfDNA performed better at identifying patients with potentially worse outcomes.

3.2.3 Stool

In 2017, Herring et al. [209] published the detection of ITGA6 and ITGA6A transcripts (integrin α6 subunit and its α6A variant) by ddPCR in stool samples obtained from patients with CRC. Patients with colorectal lesions showed statistically significantly elevated levels of ITGA6 transcript in stools with respect to the non-pathological controls (approximately eight times higher for adenomas and 6–11 times higher for CRC, being particularly higher in more advanced stages). Meanwhile, a greater than 40-fold elevation of ITGA6A was found in stools of patients with stage II and III CRC with respect to controls. This study directly compared ddPCR and conventional qPCR, with results being similar in terms of sensitivity and specificity.

Stool-derived DNA has also been analyzed by ddPCR for the presence of KRAS G12D mutation in patients with CRC who presented this hotspot mutation in their tumor tissues, being detectable in 80% of stool samples [210]. More recently, hypermethylation in GRIA4- and VIPR2-associated CpG islands was detected by ddPCR in stool samples from patients with CRC, highlighting their potential as early noninvasive biomarkers for diagnosis of this neoplasia [211]. In this study, ddPCR was compared with Methylight qPCR using the same primers and probes in both assays for detection of the two stool-based methylation biomarkers, demonstrating that the sensitivity of ddPCR was superior.

Apart from alterations in cancer-related genes, another possibility explored using ddPCR in stool samples was the detection of DNA from different strains of bacteria that have been associated with malignancy, such as Fusobacterium nucleatum in CRC [212] and Helicobacter pylori in gastric cancer [213, 214]. In a study performed in a Japanese population, F. nucleatum was significantly elevated in the non-advanced adenoma group, the advanced adenoma/carcinoma in situ group and the CRC group compared with the control group of healthy subjects, suggesting that this ddPCR-based assay could be useful for detecting individuals with CRC [212]. Similarly, H. pylori DNA was detected by ddPCR in patients with gastric cancer in a Chinese population by measuring the copy number of the H. pylori 16S ribosomal RNA (rRNA) gene. These authors found levels six times higher in stool from patients with  gastric cancer than in those from healthy controls, in contrast to gastric loads, which were comparable between both groups. Additional cagA detection and cagA EPIYA typing ddPCR assays developed by the same research group [213] were also tested in the stool samples. In this population with a high prevalence of the cagA virulence gene, the East Asian allele was suggested as a risk marker for gastric cancer [214]. Interestingly, stool-based detection of H. pylori clarithromycin resistance-associated genotypes through an assay targeting 23S rRNA mutant alleles (A2143G, A2142G and A2142C) also proved feasible in patients with and without gastric cancer, particularly in cases of heteroresistance, where it seemed to be more sensitive than commonly used methods for testing in routine clinical practice [215].

Along this line, altered microbiota was detected by 16S rRNA analysis using ddPCR in postoperative fecal samples from patients subjected to pancreaticoduodenectomy (with a cancer diagnosis confirmed by surgical pathology in 45 of 50 cases) [216]. A depletion of strict anaerobes and an expansion of some Proteobacteria, with an enrichment in Bacteroides and Klebsiella, was observed in fecal, pancreatic fluid, bile and jejunal samples, deviating from the microbial patterns considered normal in healthy individuals. These results suggest that postoperative fecal microbiota may have a potential predictive value to identify patients at high risk for pancreatic cancer, but this possibility needs to be further explored.

3.2.4 Ocular Fluids

Hiemcke-Jiwa et al. [217] proved that detection of MYD88 L265P mutation in these fluids by ddPCR is feasible and represents a reliable tool for the diagnosis of vitreoretinal lymphoma (VRL) and for treatment monitoring. This hotspot mutation was present in 74% of patients in this study and is a distinguishing mark of VRL. Patients with uveitis were included as a negative control group. The analysis of paired samples from patients with VRL revealed that sensitivity was 75% in VF versus 67% in AH, with positive predictive values and specificities of 100% in both cases. Indeed, ddPCR allowed the detection of MYD88 L265P on cfDNA even in > 100-fold diluted VF samples. Interestingly, the mutation became undetectable in any ocular fluid after intravitreal and systemic treatment [217]. In another recent study, MYD88 L265P mutation (which was present in 75% of patients with VRL, an incidence similar to that in the aforementioned work by Hiemcke-Jiwa et al. [217]) was detected by ddPCR in the VF of patients with diffuse large B-cell lymphomas and in one patient with lymphoplasmacytoid lymphoma [218].

AH can be obtained by paracentesis and is considered a safer and less invasive method of gathering DNA from tumor origin than collecting VF specimens. In particular, taking retinal biopsies by fine needle aspiration (FNA) incurs a high risk of complications, including infection, hemorrhage or retinal detachment [217]. However, the volumes of AH obtained are small and have low DNA content. Thus, a very sensitive technique for analysis is required. VRL diagnosis is extremely complicated and requires the combination of several laboratory tests, including flow cytometry (for detection of clonal B-cell populations), cytomorphology, immunohistochemistry and molecular analysis (determination of cytokine levels, immunoglobulin gene rearrangements and mutational analysis), because no single diagnostic test has sufficient sensitivity and specificity of detection. The use of ddPCR alone is not enough for an accurate diagnosis, but it provides an additional tool that could be integrated into the clinical routine. The analysis of MYD88 L265P mutation in both VF and AH in combination could also contribute to a better diagnosis [173, 217, 218]. Double-side vitrectomy and multisite sampling including CSF have also been demonstrated to improve ddPCR detection efficiency in PCNSL and other primary extranodal lymphomas [173].

3.2.5 Saliva

Recent studies have explored the utility of saliva as a source of DNA for study in different types of cancer. The levels of human papillomavirus (HPV) DNA were previously studied in the plasma of patients with advanced HPV-associated oropharyngeal cancer (OPC) using a ddPCR multiplex assay to detect the most common high-risk HPV subtypes: 16, 18, 31, 33 and 45 [219]. The same authors later hypothesized that viral DNA could also be shed by tumor cells in the oropharynx into the saliva, paving the way for the use of salivary secretions for diagnostic, prognostic and predictive purposes in disease monitoring [220]. To test their hypothesis, they designed an observational study to analyze paired plasma–saliva samples from patients with HPV-OPC. This study confirmed that HPV DNA is detectable in the saliva and correlates with tumor burden and local disease subsite. Interestingly, salivary HPV DNA viral load showed a strong correlation with tumor burden in patients with locoregional disease but not in those with distant disease only, in contrast to plasma, the levels of which were associated with tumor burden among the whole cohort, irrespective of disease site. Furthermore, HPV DNA baseline levels in saliva were almost 20 times higher in patients with clinical and imaging evidence of locoregional disease than in those with distant disease outside the head and neck only, showing a clinically valuable predictive potential. These levels were particularly elevated in those with base-of-tongue tumors compared with tonsil cases. Salivary HPV DNA  levels fluctuated in close relationship with disease progression and response, and changes were observed prior to clinical detection in most cases. High HPV DNA levels in plasma, in turn, were associated with worse outcomes, indicating that saliva and plasma provide different and complementary information. The use of both bodily fluids simultaneously in a combined strategy increased the sensitivity of detection up to 100% in this study.

Salivary exosomes have been suggested as an alternative and enriched source of DNA and RNA in patients with HPV-OPC. An acoustofluidic biocompatible platform previously developed for plasma samples [221] was later optimized to isolate salivary exosomes at a high yield, irrespective of the variable viscosity and collection method of saliva samples [222]. This platform consists of a fusion of acoustics and microfluidics that uses standing surface acoustic waves, and ddPCR was used to evaluate the exosome yield obtained by this platform compared with other isolation methods. Thus, the concentration of the two small RNAs (miR-148-a and piR014923) measured by ddPCR was 15 times higher in the exosome fraction isolated by the optimized platform than in that isolated with the differential ultracentrifugation method. These authors also designed a ddPCR assay that could detect HPV16 DNA in 80% of patients with HPV-OPC in a small cohort. Interestingly, the concentration of the target DNA was 12 times higher in the exosome fraction than in microvesicles.

Saliva has also been interrogated for the presence of EGFR mutations in patients with NSCLC using ddPCR technology. Paired plasma and saliva samples showed a concordance of 83.78% and was correlated with clinical response. However, saliva cfDNA concentrations could not distinguish patients with NSCLC from controls (including healthy subjects and patients with pulmonary benign disease). This suggests it is a qualitative rather than a quantitative indicator and thus could not be applied for diagnostic purposes in this malignancy, but it could complement plasma and tissue biopsies [223]. A recent study also compared ddPCR and a novel technology called electric field-induced release and measurement (EFIRM) for the detection of EGFR mutations in paired plasma and saliva from patients with NSCLC. The sensitivity of EFIRM was 100% in both types of samples, whereas ddPCR showed sensitivities of 84.6% in plasma and 15.4% in saliva. This work by Li et al. [224] revealed that ctDNA was more fragmented in saliva than in plasma, and EGFR L858R was present mainly as ultrashort DNA fragments between 40 and 60 bp in size (known as ultrashort ctDNA) that the ddPCR assay was mostly unable to amplify. Of note, the EFIRM test results indicated that the concentration of mutant ctDNA in saliva was higher than in plasma, in stark contrast with the results of the previous study by Ding et al. [223, 224]. Interestingly, the majority of these EGFR mutant sequences were encapsulated within exosomes [224].

Finally, as previously mentioned, a PIK3CA variant was also detected in the saliva of a patients with PROS [204].

3.2.6 Sputum

In 2018, Su et al. [225] applied droplet digital methylation-specific PCR (ddMSP) to the detection of epigenetic biomarkers in sputum to develop a classifier for the early detection of lung cancer. A comparison of ddMSP and the conventional quantitative MSP (qMSP) showed that the former was more sensitive and provided more precise and reproducible results for methylation quantification. The authors built an epigenetic classifier using ddMSP that included four sputum methylation biomarkers (HOXA9, RASSF1A, SOX17 and TAC1) and demonstrated a sensitivity of 86.6% and a specificity of 90.6% for the detection of lung cancer. This method was superior to the clinical gold standard—sputum cytology—in accuracy and sensitivity, although both had a similar specificity. Additionally, the authors intentionally added inhibitors (sodium dodecyl sulfate and heparin) to the PCR reactions and observed that ddMSP better tolerated the presence of these inhibitory substances than did qMSP [225].

More recently, EGFR mutations were detected in the sputum of patients with NSCLC using ddPCR [226, 227]. Isaka et al. [226] reported that the detectability of EGFR mutations in sputum samples by ddPCR was highly sensitive in cases with positive sputum cytology and very low (3.1%) in patients with negative cytology. Based on these and previous observations, these authors suggested that sputum samples should be collected for EGFR mutation analysis in cases where CT tumor size is ≥ 29 mm, which is considered a potential predictive factor for positive sputum cytology [226].

On the other hand, Hackner et al. [227] analyzed paired plasma and sputum samples for the screening of activating and resistance EGFR mutations in patients with NSCLC, showing that the combination of both fluids for detection of T790M mutation increased diagnostic efficiency in patients with progressive disease compared with the single analysis of plasma. Again, conventional cytology was unable to detect tumor cells in any of the sputum specimens, although ctDNA was detectable [227].

3.2.7 Bronchoalveolar Lavage

Another body fluid that has been examined using ddPCR in the oncologic field is bronchoalveolar lavage (BAL) and, more precisely, the cell pellets obtained from this liquid biopsy. In 2017, ddPCR was used to validate an miRNA-based prediction model (including two miRNAs: miR-205-5p and 944) to distinguish squamous cell carcinoma from adenocarcinoma in BAL samples from patients with NSCLC, showing higher diagnostic accuracy than cytology [228]. Similarly, the determination of BRAF V600E mutation in BAL samples by ddPCR was proposed as a complementary tool for the diagnosis of pulmonary Langerhans cell histiocytosis [229].

More recently, a ddPCR method was developed to quantify the CpG methylation levels of TMPRSS4 and SHOX2 promoters in plasma and BAL samples from patients with NSCLC, showing that TMPRSS4 methylation status allows distinction between patients with early-stage NSCLC and healthy individuals. TMPRSS4 was hypomethylated in BAL and plasma of patients with early-stage disease compared with controls, and an inverse correlation was observed between TMPRSS4 and SHOX2 in patients with early-stage NSCLC versus healthy controls in both body fluids, although this correlation was only observed in BAL in all stages. These results support the potential of TMPRSS4 as noninvasive epigenetic biomarker as an indicator of malignancy in early-stage NSCLC [230]. In line with this, Roncarati et al. [231] recently described a four-gene methylation panel (RASSF1A, CDH1, DLC1 and PRPH) in bronchial washings using ddMSP that showed a remarkable diagnostic value in lung cancer, with a sensitivity of 97% and a specificity of 74%. ddPCR has also been used for the detection of EGFR-TKI-sensitizing mutations in bronchial washing fluid and plasma of patients with NSCLC, showing that the former has greater diagnostic efficacy than the latter [232].

3.2.8 Pleural Effusions/Pleural Fluid

Several publications have reported the detection of EGFR mutations in pleural effusions or pleural fluid samples of patients with NSCLC [233–236]. These mutations were analyzed in the cfDNA obtained from the supernatant of pleural fluid [233], in the cell pellet [236] or both [234, 235]. Hummelink et al. [235] demonstrated that KRAS and EGFR mutations were detectable in paired supernatant and cell pellet samples, from not only NSCLC but also colon carcinoma, appendiceal carcinoma and adenocarcinoma of unknown primary. This study suggested that the cell-free fraction of pleural effusions is an excellent source of genetic material for detection of both driver and resistance mutations and that the combined analysis of both fractions reached optimal sensitivity [235].

3.2.9 Mucin

ddPCR was recently applied to the detection of KRAS mutations in cfDNA from mucin obtained from patients with pseudomyxoma peritonei, a rare malignant disorder characterized by the accumulation of huge amounts of this viscous fluid in the abdominal cavity [237]. To date, the gold standard for routine diagnosis is the screening of mucin in search of tumor cells. However, this pilot study proved that acellular mucin contained cfDNA from tumor origin despite the absence of detectable tumor cells. Paired plasma samples from the same patients were also analyzed but were negative in all cases. These results are concordant with the localized nature of this malignancy and pave the way for the use of ddPCR in further studies aimed at elucidating the complex molecular mechanisms leading to recurrence in this acellular neoplasm [237].

3.2.10 Peritoneal Fluid/Ascites

As mentioned, in many cases, ddPCR represents a valuable tool for validation of NGS results from different kinds of biologic samples. The cellular fraction of second-look peritoneal washings from patients with high-grade serous ovarian, fallopian tube, or primary peritoneal cancer was analyzed using ddPCR to validate the results previously obtained by NGS [238]. Tumor-specific mutations (including the most frequent mutated gene TP53 but also PTEN and HNF1A, among others) were detectable in second-look washings, for both primary and recurrent tumors, providing valuable information about tumor heterogeneity and residual disease [238]. In line with this, a recent study employed ddPCR as a validation technique for quantification of miR-593-3p as a prognostic biomarker in pancreatic cancer undergoing staging laparoscopy [239]. Its expression was upregulated in the supernatant of peritoneal lavage fluids with positive cytology and correlated with worse overall survival and disease-free survival, even in cases with localized disease and negative cytology. These results suggest that elevated levels of miR-593-3p could be an indicator of the presence of subclinical intra-abdominal micrometastasis [239]. In agreement with these observations, another recent publication also reported that high levels of peritoneal lavage tumor DNA were associated with poorer outcomes in terms of disease-free survival and overall survival [240]. In this work by Suenaga et al. [240], ddPCR was employed to analyze peritoneal lavage from patients with PDAC in search of KRAS mutations. Interestingly, tumor-derived DNA was detectable not only in patients with positive cytology but also in 40% of the patients with a negative cytology result, showing a remarkably superior sensitivity for prediction of peritoneal recurrence than cytology, despite having lower specificity [240].

The analysis of KRAS mutations in peritoneal cfDNA by ddPCR was very recently proposed as a biomarker for peritoneal surface malignancies [241]. MAF was found to correlate with the surgical peritoneal carcinomatosis index, showing that MAF < 1% were associated with complete cytoreduction in contrast to MAF > 1%. These results suggest that peritoneal ctDNA testing would be useful as a surrogate for disease burden and as an indicator of resectability in peritoneal carcinomatosis. However, in this study, MAF could not be related to survival. Interestingly, mutant KRAS ctDNA was detected in peritoneal fluid in 20% of patients with KRAS wild-type tumors (as determined by tissue analysis using Sanger sequencing and NGS). Some of these patients had received anti-EGFR therapy prior to sample collection. Thus, this discordance may be related to intratumoral heterogeneity and/or clonal selection in response to treatment [241].

Another recent study evaluated the use of plasma and peritoneal fluid in a combined liquid biopsy approach as a prognostic factor in patients with advanced colorectal and appendicular tumors undergoing complete cytoreductive surgery and hyperthermic intraperitoneal chemotherapy (CC-HIPEC) [152]. In this study, KRAS mutations were analyzed in plasma and peritoneal fluid by ddPCR before and after CC-HIPEC. Patients with detectable ctDNA in plasma or peritoneal fluid after treatment had shorter disease-free and overall survival (including the 3-year survival rate). Patients with positive ctDNA in post-treatment plasma experienced systemic relapses, whereas its presence in post-HIPEC peritoneal fluid was associated with peritoneal recurrence and/or systemic relapses. In turn, all patients with negative liquid biopsy after treatment remained disease free. Interestingly, post-HIPEC cytology was negative in all cases, but ctDNA was not neutralized in peritoneal fluid by this treatment, suggesting that its persistence may predict a worse outcome [152].

3.2.11 Fine Needle Aspirate

ddPCR for detection of RAS and BRAF mutation has been proposed as a complementary tool in association with cytology to increase the diagnostic sensitivity and specificity of FNA biopsy of thyroid nodules [242]. In another study, these mutations were detected by NGS and confirmed by ddPCR [243]. In line with this, BRAF V600E mutation was also detected in FNA fluid from patients with papillary thyroid carcinoma [244]. In fact, the performance of ddPCR was better than that of ARMS-PCR when detecting this mutation from thyroid nodule FNA samples [20].

Two different studies have highlighted the utility of analyzing the supernatant cfDNA from FNA, a fluid that is usually discarded after centrifugation to separate the cell pellet for cytospin or FFPE cell block preparations. Guibert et al. [245] detected EGFR, BRAF and KRAS mutations with ddPCR in supernatant cfDNA from patients with suspected lung cancer and adenocarcinomas with acquired EGFR resistance. Similarly, another study confirmed EGFR, KRAS, BRAF, PIK3CA and NRAS mutations by ddPCR in post-centrifuged supernatant from malignant and benign FNA needle rinses, including cases of melanoma and pancreatic, lung, colorectal, breast, urothelial and hepatocellular carcinoma [246]. The same authors later observed a 100% concordance between NGS and ddPCR results in a small subset of FNA samples from patients with NSCLC [247]. More recently, ddPCR has been used to confirm EGFR mutations in samples from patients with NSCLC when inconsistencies were found between ARMS and SuperARMS-PCR [236].

The analysis of KRAS mutations using washes from endoscopic ultrasound-guided FNA also proved useful for the detection of local recurrence in pancreatic cancer [248].

3.2.12 Pancreatic Juice/Cyst Fluid/Bile

Suenaga et al. [249] determined that KRAS codon 12 and GNAS codon 201 mutations are detectable by ddPCR in pancreatic juice, finding higher concentrations when directly collected from the ampulla using an endoscopic distal cap. In addition, they observed that the optimal time for collection of pancreatic juice to increase the likelihood of detecting KRAS mutations in this biofluid was 10 minutes after secretin infusion [250].

KRAS mutations have also been examined with ddPCR in cfDNA from bile samples and plasma of patients with PDAC and cholangiocarcinoma, with results comparable to those with NGS. These results suggest that bile-based liquid biopsy by ddPCR might be a reliable tool for diagnosis of pancreatobiliary cancers [251].

In a different line of research, ddPCR was used to detect clinically relevant bacteria in pancreatic duct fluid and in bile and jejunal contents from patients undergoing pancreaticoduodenectomy [216].

Studies describing the use of ddPCR in biologic fluids other than blood are summarized in Table 4.

Table 4 Use of droplet digital polymerase chain reaction in biologic fluids other than blood
Full size table

3.3 Concluding Remarks

In this updated review, we emphasize that the number of publications evaluating ddPCR has increased considerably in recent years, particularly in the field of liquid biopsy. The clinical utility of this methodology has been reinforced not only by its combination with other techniques, mainly NGS, but also by the integration of different biomarkers in multianalyte or multiparametric approaches. ddPCR is also usually employed as a validation, control or reference system to confirm the results obtained by other technologies, which highlights its reliability for detection of a wide variety of genetic alterations.

Blood is the main source of nucleic acids for noninvasive biomarker analysis by ddPCR, but the range of biologic fluids exploited for this purpose has extensively broadened in the last 4 years, providing a complementary tool for diagnosis and surveillance in many cancer types. ctDNA is the more widespread nucleic acid-based biomarker studied, but there is an increasing trend towards the investigation of RNA signatures, including long non-coding RNAs, miRNAs or mRNAs, frequently derived from exosomes or EVs, alone or in combination with other markers, such as CTCs.

Despite the evident prevalence of the Bio-Rad ddPCR platform in the literature, the development of new ddPCR platforms with additional detection channels paves the way for improved multiplexing strategies. Even so, limitations regarding issues such as scalability and sensitivity in samples with low DNA content and low-DNA-shedding tumors at early stages or anatomically sequestered locations, and the lack of standardized protocols still need to be overcome. Further research on the application of ddPCR in oncology is warranted.