Abstract
Previous work has shown that circulating miR-371a-3p has higher sensitivity and specificity than current biomarkers for malignant germ cell tumors (MGCTs). Herein, the performance of two methods commonly used to measure miR-371a-3p levels was compared: quantitative RT-PCR (RT-qPCR) and droplet digital PCR (ddPCR). Samples from the University of Texas Southwestern Medical Center (Dallas, TX) and the University of Cambridge (Cambridge, UK) were evaluated using both RT-qPCR and ddPCR, as per current protocols (RT-qPCR) or standard manufacturer’s recommendations (ddPCR). Data were available for 70 individuals: 36 patients with MGCT and 34 control participants (nonmalignant GCT, non-GCT cancer, and noncancer). The performance of the two assays was compared using receiver operating characteristic curves and the area under the curve. Raw miR-371a-3p Cq values (RT-qPCR) were generally lower (ie, miR-371a-3p was more abundant) and the number of positive droplets (ddPCR) higher for patients with MGCT compared with control participants. The area under the curve (95% CI) was 0.94 (0.90—0.99) and 0.82 (0.70—0.93) when using RT-qPCR and ddPCR, respectively. Thus, RT-qPCR had better classification performance compared with ddPCR in the present cohort, supporting the continued use of RT-qPCR in standard clinical practice. Further investigation is required to optimize ddPCR before it should be considered for clinical adoption.
Germ cell tumors (GCTs) are the most common solid tumors among adolescent and young adults.1 Although outcomes for patients with GCTs are favorable, late effects of treatment include second malignant neoplasms, early-onset cardiovascular disease, disabling neuropathies, and hearing loss, which significantly impact the lives of patients who are cured.2 Current malignant GCT (MGCT) serum biomarkers include alpha-fetoprotein and human chorionic gonadotrophin; however, because not all MGCT histologic subtypes secrete these serum biomarkers, only approximately 60% of patients are “marker positive” at diagnosis.3 This lack of a biomarker in a large group of patients warrants the need for better diagnostic tools so that patients can receive better counseling regarding management options, for example to reduce instances where there is potential for patients to be overtreated or undertreated.
miRNAs are promising circulating biomarkers for MGCTs. In particular, a single miRNA, miR-371a-3p, has been shown to demonstrate higher sensitivity and specificity in the diagnosis of testicular MGCT than current biomarkers4; however, there is a need to fully understand and define reproducibility, thresholding, optimal assay performance, and interpretation before such miRNA quantification entering the clinical arena. To date, quantitative RT-PCR (RT-qPCR) has been most commonly used to measure circulating miR-371a-3p levels, and protocols have been designed and refined to measure this miRNA in serum, plasma, and cerebrospinal fluid.5 Extensive work in refining the quantification protocols for miR-371a-3p has led to the recent introduction of an indeterminate range (ie, neither a negative nor a positive result).6 In our laboratories, following robust quality control checks, current guidance calls for retesting samples that are found to have raw miR-371a-3p Cq values between 28 and 35 on the first PCR run.6 These samples are classified based on the second PCR run; samples that fall within this range and have a raw Cq value less than 28 on the second run are declared positive, those greater than 35 are negative, and those between 28 and 35 on the second run are classified as indeterminate.6 Recent work investigating performance based on this recommendation resulted in 100% sensitivity and approximately 97% specificity for the circulating RT-qPCR miR-371a-3p assay.7 More recently, the use of droplet digital PCR (ddPCR) has also been proposed as a potential alternative method to quantify circulating miR-371a-3p levels for MGCT diagnosis. Indeed, several groups have explored the performance characteristics of ddPCR for miR-371a-3p in this context.8,9 In this study, the performance of RT-qPCR and ddPCR for the quantification of circulating miR-371a-3p levels across the early-stage MGCT disease spectrum was compared to better understand the performance characteristics of the two methods.
Materials and Methods
Sample Collection and Processing
Following institutional review board approval and patient consent, serum samples were collected from patients treated at the University of Texas Southwestern Medical Center in Dallas, TX (institutional review board: STU-102010-051). For the University of Cambridge (Cambridge, UK) work, the study received approvals from the multicenter generic Children’s Cancer and Leukaemia Group Tissue Bank (East-Midlands/Derby REC reference: 08/h0405/22+5, covering Biological Studies CCLG-2002-BS03 and CCLG-2020-BS02; formerly Trent-REC reference: 02/4/071), Cambridge Local Research Ethics Committee (reference: 01/128), and the London-Bromley Research Ethics Committee (REC reference: 20/PR/0126 and IRAS project ID: 282883). For this analysis, cases were defined as patients with MGCT. Controls included patients with nonmalignant GCT (such as pure teratoma), non-GCT cancer, and non-cancer controls (Table 1). Classification was determined based on surgical pathology and tumor markers. All patients were chemotherapy naive at the time of sample collection, and all samples were collected before surgery.
| RT-qPCR | ddPCR | |
|---|---|---|
| Assay ID | Taqman; 002124 | Taqman; 002124 |
| Total PCR Cycles | 59 (14 pre-amp + 45 qPCR) | 40 |
| Starting Serum Volume | 200 μL | 200 μL |
| RNA elution volume | 100 μL | 25 μL |
RT-qPCR Method
The protocol by Murray et al5 was followed to quantify serum miRNA levels using RT-qPCR. Briefly, blood was collected in serum separator tubes (with clotting activator and serum separator gel), allowed to clot for 30 minutes, and then underwent centrifugation within 4 hours of receipt as per the manufacturer’s recommendations. Samples were then refrigerated at 4° C (for up to 3 days), and residual serum was retrieved, frozen, and stored at −80°C. RNA was isolated using the miRNeasy Serum Kit (Qiagen, Crawley, UK) following the manufacturer’s instructions. Two multiplexed RT reactions were then performed for each sample to quantify the expression of miR-371a-3p using TaqMan miRNA-specific stem-loop RT primers, and the 14-cycle pre-amplification step was then undertaken (Table 2). For the pre-amplification step, the relevant TaqMan assay probes were used to multiplex the reaction, as described by Murray et al.5 Two pre-amplification reactions were run for each sample, and the reaction product was diluted 1:5 with nuclease-free water for each reaction. The final singleplex miRNA PCRs were run in triplicate as per manufacturer’s instructions. These were conducted on a Mastercycler ep_gradient/S realplex (Eppendorf, Stevenage, UK) and Bio-Rad CFX386 qPCR machine (Bio-Rad, Hercules, CA) at 95°C for 10 minutes, then 45 cycles of 95°C for 15 seconds, and 60°C for 1 minute.5,10 For additional details regarding the protocol, refer to Murray et al.5
Table 2:
Patient Characteristics for Patients Included in Analyses.
| MGCT (N = 36) | Controls (N = 34) | |
|---|---|---|
| Age* | 29.9 (SD = 16.1) | 15.8 (SD = 13.6) |
| Male | 34 (94.4%) | 21 (61.8%) |
| Site | ||
| Gonadal | 31 (86.1%) | 17 (50.0%) |
| Extragonadal | 5 (13.9%) | 9 (26.5%) |
| No tumor | 0 (0.0%) | 8 (23.5%) |
| Histology | ||
| Seminoma GCT | 12 (33.3%) | 0 (0.0%) |
| Non-Seminoma GCT | 23 (63.9%) | 0 (0.0%) |
| Germ Cell Neoplasia in Situ | 1 (2.8%) | 0 (0.0%) |
| Pure Teratoma | 0 (0.0%) | 7 (20.6%) |
| Non-GCT Cancer | 0 (0.0%) | 8 (23.5%) |
| Non-Cancer Controls | 0 (0.0%) | 19 (55.9%) |
Mean (SD; standard deviation) presented for continuous variables and n (%) presentedfor categorical variables.
I patient in control group is missing age
ddPCR Method
For ddPCR, blood was collected in serum separator tubes and processed as described in RT-qPCR Method. Separated serum obtained after standard centrifugation was transported to the laboratory on an icepack or stored at −80°C. The miRNA serum/plasma extraction kit AS1680 (Promega, Madison, WI) was used to extract RNA (including miRNA), as per the manufacturer’s recommendations. Specifically, RNA was extracted from 200 μL of serum, with DNase added during the extraction to remove circulating DNA. The RNA was either analyzed immediately or stored at −80°C until use. miRNA-specific multiplex RT was performed using a TaqMan miRNA RT kit (Thermo Fisher Scientific, Waltham, MA) (Table 2). RT primers specific for either miR-371a-3p (test) or miR-30b-5p (an endogenous miRNA for quality control purposes) were mixed with RT buffer, dNTPs, reverse transcriptase, and RNase inhibitor. Five microliters of extracted serum RNA was reverse transcribed in a 15-μL reaction on a Bio-Rad thermal cycler.11
The final ddPCR step was performed using a Bio-Rad QX200 ddPCR system with an automated droplet generator, droplet reader, and inbuilt data analysis program. Nine microliters of diluted cDNA product was used in a 20-μL ddPCR reaction comprising Bio-Rad ddPCR supermix for probes and Thermo Fisher Scientific TaqMan miRNA assay (1×) for either miR-371a-3p or miR-30b-5p. The ddPCR reaction was run as a singleplex assay on a Bio-Rad thermal cycler at 95°C for 10 minutes before 40 cycles at 95°C for 15 seconds and 60°C for 1 minute. After the ddPCR run was completed, the data were analyzed using the default parameters on the provided QuantaSoft Analysis Pro version 1.0 (Bio-Rad).
Statistical Methods
To study the performance characteristics of RT-qPCR and ddPCR, patients who had circulating miR-371a-3p levels quantified using both methods were included. Patient characteristics were described using means ± SD for continuous variables and as frequency and percentage for categorical variables. The distribution of miR-371a-3p for patients with MGCT and control participants was compared for the two assays using histograms and descriptive statistics. The number of positive droplets was used as the measure for ddPCR, as described by Myklebust et al,8 and the raw Cq value was used for RT-qPCR, as described by Lafin et al.6 Samples with undetectable miR-371a-3p levels by RT-qPCR (defined as raw Cq > 40 or completely nonexpressed) were arbitrarily assigned a raw Cq value of 40.6,8 Pearson’s correlation was used to describe the correlation between RT-qPCR and ddPCR measures, and the classification performance of the two methods was compared using receiver operating characteristic curves and the area under the curve. Confidence intervals for area under the curve were obtained using bootstrapping, with 5000 bootstrap samples.12 For the purposes of this analysis, the optimal threshold was estimated for each method by maximizing the Youden Index.13 The corresponding sensitivity and specificity of the two procedures were compared for the respective thresholds. All descriptive statistics were presented overall and stratified by whether the sample was from an MGCT case/control status.
Results
RT-qPCR and ddPCR measures were available for 70 patients, of whom 36 (51.4%) had an MGCT and 34 (48.6%) were control participants (Table 1). Patients with MGCT were older, on average, compared with control participants (MGCT: 29.9 years; control participants: 15.8 years); of note, previous work has shown that there are no baseline differences in circulating miR-371a-3p levels by age or biological sex.7 Approximately 80% of all samples came from male patients (MGCT: 94.4%; control participants: 61.8%), and most patients with MGCT had gonadal tumors (86.1%).
The mean ± SD RT-qPCR value for miR-371a-3p was 25.5 ± 5.1 for patients with MGCT and 36.7 ± 4.4 for control participants (Table 3). Among control participants, 13 had raw Cq values <40, including 7 noncancer controls, 3 patients with pure teratomas, 1 with small cell carcinoma of the ovary, 1 with a benign ovarian cyst, and 1 with a Leydig cell tumor. The mean ± SD number of positive droplets for miR-371a-3p using ddPCR was 298.5 ± 548.1 for patients with MGCT (Table 3), demonstrating substantial variability due to some patients having very high numbers of positive droplets (Figure 1). The mean ± SD number of positive droplets was 7.5 ± 3.9 among control participants (Table 3 and Figure 1).
Table 3:
Mean (Standard Deviation) of RT-qPCR and ddPCR for Patients with Malignant GCT (MGCT) and for Controls.
| MGCT (N = 36) | Controls (N = 34) | |
|---|---|---|
| RT-qPCR (raw Cq) | 25.5 (5.1) | 36.7 (4.4) |
|
| ||
| Cq < 28 | 26 (72.2) | 0 (0.0) |
| 28 < Cq < 35 | 8 (22.2) | 13 (38.2) |
| Cq > 35 | 2 (5.6) | 21 (61.8) |
|
| ||
| ddPCR (positive droplets) | 298.5 (548.1) | 7.5 (3.9) |
Figure 1.
A: Histogram of quantitative RT-PCR (RT-qPCR) stratified by case/control status. B: Histograms of droplet digital PCR (ddPCR) stratified by diagnosis.
The RT-qPCR values among controls followed a bimodal distribution, with the first mode overlapping with the raw Cq values among patients with MGCT (Figure 1). Some overlap in the distribution of the number of positive droplets was also observed in the ddPCR analysis. Note that the histograms of the number of positive droplets are presented separately for patients with MGCT and control participants (Figure 1) due to the large difference in the range of positive droplets for the two groups; the number of positive droplets ranged from 0 to 16 for control participants and from 0 to 2500 for patients with MGCT.
As expected, there was a negative correlation between ddPCR results (presented as the log number of positive droplets) and RT-qPCR (Figure 2); the higher the number of positive droplets, the lower the raw Cq value. The raw Cq value and number of positive droplets present a moderate negative correlation (r = −0.54). Moreover, as seen in Table 3, patients with MGCT typically had larger numbers of positive droplets and lower Cq values compared with control paticipants.
Figure 2.
Scatterplot of quantitative RT-PCR (RT-qPCR) and droplet digital PCR (ddPCR) measurements. Dashed lines represent the threshold of 28.52 for RT-qPCR and log (17.5), where 17.5 is the threshold for ddPCR.
To investigate the classification performance of the two methods, the area under the curve was estimated to be 0.94 (95% CI, 0.90—0.99) when using RT-qPCR and 0.82 (95% CI, 0.70—0.93) when using ddPCR (Table 4). For RT-qPCR, the optimal threshold based on maximization of the Youden Index was determined to be a Cq value of 28.52, with corresponding estimated sensitivity of 81% and specificity of 100%. For ddPCR, the optimal threshold was 17.5 positive droplets, with a corresponding estimated sensitivity of 72% and specificity of 100% based on this threshold.
Table 4:
AUC values Based on RT-qPCR and ddPCR Classifications.
| Method | AUC (95% CI) | Threshold | Sensitivity | Specificity |
|---|---|---|---|---|
| RT-qPCR | 0.94 (0.90, 0.99) | 28.52 | 0.81 | 1.00 |
| ddPCR | 0.82 (0.70, 0.93) | 17.50 | 0.72 | 1.00 |
Threshold values were determined by maximizing the Youden Index.
As previously discussed, current guidance for RT-qPCR states that samples with raw Cq values <28 should be classified as positive, those >35 as negative, and those between 28 and 35 should require a second run. Based on this guidance, approximately 72% of patients with MGCT and none of the control participants would be classified as positive; 2 patients with MGCT (5.6%) and 21 control participants (61.8%) would be classified as negative; and 8 patients with MGCT (22.2%) and 13 control participants (38.2%) would require a second run (Table 3). Although a second run was not available for the samples in this study, investigation of the characteristics of patients who would have required a second run based on current guidance was performed. Patients whose raw Cq value was between 28 and 35 on the first run were observed to be, on average, younger (mean ± SD age, 17.9 ± 12.8 years) than those classified as positive or negative (mean ± SD age, 24.8 ± 18.2 years). No large differences by site or sex were observed. In addition, eight patients (38.1%) who fell within 28 and 35 Cq on the first run were noncancer control participants.
Discussion
Previous work has shown the potential utility of circulating miR-371a-3p as a biomarker for testicular MGCT, and robust standard operating procedures using existing technology as the gold standard are required. As technology continues to advance, however, it is important to seek to further optimize the quantification of this biomarker. Circulating miR-371a-3p levels have typically been measured using RT-qPCR, but more recently, ddPCR technology has been introduced, offering potential benefits such as absolute quantification. It is still not clear, however, whether ddPCR can improve the quantification of miR-371a-3p compared with the existing highly sensitive preamplified RT-qPCR methods. This study conducted a direct comparison of the two methods using current protocols/recommendations and demonstrated that RT-qPCR had a greater area under the curve and assay sensitivity than ddPCR. These results align with the observations of other groups in that both RT-qPCR and ddPCR are found to perform well for the quantification of circulating miR-371a-3p. In one ddPCR study, sensitivity of approximately 89% and specificity of 100% were observed using serum.8 In another study, sensitivity of approximately 94% and specificity of 100% were observed.9 This study observed a lower sensitivity of ddPCR (72%) compared with previously published results. However, this study was performed across the early-stage MGCT disease spectrum, suggesting that RT-qPCR offers better performance characteristics close to or at the lower limit of detection than ddPCR in this clinical setting.
Previous work identified the existence of an indeterminate range when using RT-qPCR to quantify serum miR-371a-3p levels.6 The current protocol recommends that samples with raw Cq miR-371a-3p values between 28 and 35 on the first PCR run be tested again and that these are classified according to the second PCR run.6 If on the second run the sample falls within this range again, the sample is classified as indeterminate.6 For the present work, the thresholds used to classify patients as cases or controls were determined by maximizing the Youden Index for both RT-qPCR and ddPCR. Furthermore, the results of this study are based on a single run due to sample availability. Although the work presented herein classifies patients based on the Youden Index threshold, the number of samples that fell within a raw Cq of 28 and 35 was also investigated. All MGCT samples in this range were misclassified by ddPCR, whereas only five of the eight samples were misclassified using the optimal thresholds in this study, partially explaining the higher sensitivity of RT-qPCR. In addition, potential differences between patients who fell within 28 and 35 Cq on the first run were investigated. In this sample, patients whose raw Cq value was between 28 and 35 on the first run were, on average, younger than those who did not; however, previous work did not identify differences by age. Furthermore, 38.1% of patients who fell within 28 and 35 Cq on the first run were noncancer controls. This is consistent with previous work, which suggests that this phenomenon is a stochastic event. In most cases, true negatives that fall within 28 and 35 Cq on the first run will be classified as negative on the second run.6
There were some limitations to this study. A larger cohort with varying stages of MGCT disease is likely needed to fully understand the relative performance of RT-qPCR and ddPCR. Moreover, the standard protocol for ddPCR quantification was used as per the manufacturer’s recommendations. It is possible that further refinement of the ddPCR protocol, such as increasing the starting input material or diluting RNA in smaller volumes, could improve sensitivity; however, the protocol used reflects current practice.8,9,14 Furthermore, for the purpose of this study, the classification of patients with MGCT and control participants based on RT-qPCR undertaken herein, which used the Youden Index to determine a threshold for positivity, differed from current guidance.6 Reassuringly, however, this gave an optimal raw Cq threshold of 28.52, consistent with the previously identified raw Cq threshold of 28 for positive cases.6 Furthermore, pure teratomas were determined based on pathology and tumor marker information. It is possible that minor nonteratoma components were present and not identified; however, this approach is consistent with standard diagnostic practice. Furthermore, teratomas with malignant components are often seen in postpubertal males, but all patients with teratomas in this study were female or prepubertal males and were classified as control participants based on the Youden Index thresholds for RT-qPCR and ddPCR.
Conclusions
Circulating miRNAs show promise as potential biomarkers for disease and have already shown greater sensitivity compared with the current biomarkers for MGCT: alphafetoprotein and human chorionic gonadotrophin.4 The use of ddPCR has been explored to potentially improve circulating miR-371a-3p quantification. This study conducted a direct comparison of the two methods and found that RT-qPCR outperformed ddPCR in terms of assay sensitivity. Of note, other (non—droplet-based) digital PCR approaches delivering absolute quantification are continuously being developed and should be systematically investigated in due course. Although ddPCR may show promise and offer some advantages, it should also be noted that the use of RT-qPCR has its own advantages. The equipment and infrastructure required to measure miR-371a-3p using RT-qPCR is readily available, less expensive, and more commonly used than that for ddPCR, including in clinical practice. These results, along with these potential benefits, support the continued use of RT-qPCR for the measurement of miR-371a-3p at the present time, whereas new technologies continue to be refined and improved.
Acknowledgments
Supported by NIH National Cancer Institute award UH3CA240688 (A.L.F., A.B., J.F.A., N.C., M.J.M., and M.M.N.), St. Baldrick’s Consortium Award grant 358099 (A.L.F., J.F.A., and M.J.M.), the Malignant Germ Cell International Consortium, the NHS England/Genomics England NHS Cancer Genomic Network of Excellence Circulating Biomarkers Work Package titled “Translating the circulating microRNA (miR-371a-3p) PCR-based assay test for malignant germ cell tumor diagnosis and follow-up” (C.G.S., N.C., and M.J.M.), and National Institute for Health and Care Research Cambridge Biomedical Research Centre grant NIHR203312 (C.G.S., N.C., and M.J.M.).
Footnotes
Disclosure Statement
None declared.
References
- 1.Poynter JN: Epidemiology of germ cell tumors. Edited by Frazier AL, Amatruda JF. In Pediatric Germ Cell Tumors: Biology Treatment Survivorship. Berlin, Heidelberg, Germany: Springer, 2014. pp. 17—36 [Google Scholar]
- 2.Fung C, Dinh PC, Fossa SD, Travis LB: Testicular cancer survivorship. J Natl Compr Canc Netw 2019, 17:1557—1568 [DOI] [PubMed] [Google Scholar]
- 3.Murray MJ, Huddart RA, Coleman N: The present and future of serum diagnostic tests for testicular germ cell tumours. Nat Rev Urol 2016, 13:715—725 [DOI] [PubMed] [Google Scholar]
- 4.Piao J, Lafin JT, Scarpini CG, Nuño MM, Syring I, Dieckmann KP, Belge G, Ellinger J, Amatruda JF, Bagrodia A, Coleman N, Krailo MD, Frazier AL, Murray MJ: A multi-institutional pooled analysis demonstrates that circulating miR-371a-3p alone is sufficient for testicular malignant germ cell tumor diagnosis. Clin Genitourin Cancer 2021, 19:469—479 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 5.Murray MJ, Bell E, Raby KL, Rijlaarsdam MA, Gillis AJ, Looijenga LH, Brown H, Destenaves B, Nicholson JC, Coleman N: A pipeline to quantify serum and cerebrospinal fluid microRNAs for diagnosis and detection of relapse in paediatric malignant germ-cell tumours. Br J Cancer 2016, 114:151—162 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 6.Lafin JT, Scarpini CG, Amini A, Konneh B, Howard JM, Gerald T, Nuno M, Piao J, Savelyeva A, Wang Z, Gagan J, Jia L, Lewis CM, Murray S, Sawa YC, Margulis V, Woldu SL, Strand DW, Coleman N, Amatruda JF, Frazier AL, Murray MJ, Bagrodia A: Refining the serum miR-371a-3p test for viable germ cell tumor detection. Sci Rep 2023, 13:10558. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 7.Scarpini C, Ward D, Lafin J, Nuno M, Amatruda J, Frazier A, Bagrodia A, Coleman N, Murray M: Circulating MiR-371a-3p assay for detecting malignant germ cell tumours: further progress towards clinical adoption. Pediatr Blood Cancer 2023, 70:S8 [Google Scholar]
- 8.Myklebust MP, Thor A, Rosenlund B, Gjengstø P, Karlsdottir Á, Brydøy M, Bercea BS, Olsen C, Johnson I, Berg MI, Langberg CW, Andreassen KE, Kjellman A, Haugnes HS, Dahl O: Serum miR371 in testicular germ cell cancer before and after orchiectomy, assessed by digital-droplet PCR in a prospective study. Sci Rep 2021, 11:15582. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 9.Sequeira JP, Lobo J, Constâncio V, Brito-Rocha T, Carvalho-Maia C, Braga I, Maurício J, Henrique R, Jerónimo C: DigiMir test: establishing a novel pipeline for miR-371a quantification using droplet digital PCR in liquid biopsies from testicular germ cell tumor patients. Front Oncol 2022, 12:876732. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 10.Murray MJ, Bailey S, Raby KL, Saini HK, de Kock L, Burke GA, Foulkes WD, Enright AJ, Coleman N, Tischkowitz M: Serum levels of mature microRNAs in DICER1-mutated pleuropulmonary blastoma. Oncogenesis 2014, 3:e87. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 11.Badia RR, Abe D, Wong D, Singla N, Savelyeva A, Chertack N, Woldu SL, Lotan Y, Mauck R, Ouyang D, Meng X, Lewis CM, Majmudar K, Jia L, Kapur P, Xu L, Frazier AL, Margulis V, Strand DW, Coleman N, Murray MJ, Amatruda JF, Lafin JT, Bagrodia A: Real-world application of pre-orchiectomy miR-371a-3p test in testicular germ cell tumor management. J Urol 2021, 205:137–144 [DOI] [PubMed] [Google Scholar]
- 12.Efron B: Bootstrap methods: another look at the jackknife. Ann Stat 1979, 7:1–26 [Google Scholar]
- 13.Schisterman EF, Perkins NJ, Liu A, Bondell H: Optimal cut-point and its corresponding Youden Index to discriminate individuals using pooled blood samples. Epidemiology 2005, 16:73–81 [DOI] [PubMed] [Google Scholar]
- 14.Campomenosi P, Gini E, Noonan DM, Poli A, D’Antona P, Rotolo N, Dominioni L, Imperatori A: A comparison between quantitative PCR and droplet digital PCR technologies for circulating microRNA quantification in human lung cancer. BMC Biotechnol 2016, 16:60. [DOI] [PMC free article] [PubMed] [Google Scholar]