1
Comparison of Droplet Digital PCR to Real-Time PCR for Quantitative Detection of Cytomegalovirus 1
Hayden RT 1, Gu Z 1, Ingersoll J 3, Abdul-Ali D 3, Shi L 2, Pounds S 2, Caliendo AM 3 2
3
1Departments of Pathology and 2Biostatistics, St. Jude Children’s Research Hospital, Memphis, TN; 4
3Department of Pathology and Laboratory Medicine, Emory University School of Medicine, Atlanta, GA 5
and Emory Center for AIDS Research, Emory University, Atlanta, GA 6
7
8
*Corresponding author: 9
Department of Pathology 10
St. Jude Children's Research Hospital 11
262 Danny Thomas Place 12
Rm C5025, Mailstop #250 13
Memphis, TN 38105-2794 14 Randall.Hayden@https://www.doczj.com/doc/823237107.html, 15
16
Running title: Comparison of Digital to Real-Time PCR 17
Key words: Real-time PCR, Digital PCR, Viral load, Cytomegalovirus
18
Copyright ? 2012, American Society for Microbiology. All Rights Reserved.J. Clin. Microbiol. doi:10.1128/JCM.02620-12 JCM Accepts, published online ahead of print on 5 December 2012
2
Abstract 19
Quantitative real-time PCR has been widely implemented for clinical viral load testing, but a lack of 20
standardization and relatively poor precision has hindered its usefulness. Digital PCR offers highly 21
precise, direct quantification without requiring a calibration curve. Performance characteristics of real-22
time PCR were compared to those of droplet digital PCR (ddPCR) for cytomegalovirus (CMV) viral load 23
testing. Ten-fold serial dilutions of the World Health Organization (WHO) and the National Institute of 24
Standards and Technology (NIST) CMV quantitative standards were tested, together with the 25
AcroMetrix? CMV tc Panel (Life Technologies, Carlsbad, CA) and 50 human plasma specimens. Each 26
method was evaluated using all three standards for quantitative linearity, lower limit of detection (LOD), 27
and accuracy. Quantitative correlation, mean viral load, and variability were compared. Real-time PCR 28
showed somewhat higher sensitivity than ddPCR (LOD of 3 log 10versus 4 log 10copies and IU/mL for NIST 29
and WHO standards). Both methods showed a high degree of linearity and quantitative correlation, for 30
standards (R 2≥ 0.98 in each of 6 regression models) and clinical samples (R 2=0.93) across their detectable 31
ranges. For higher concentrations, ddPCR showed less variability than RT-PCR for the WHO standards 32
and Acrometrix standards (p< 0.05). RT-PCR showed less variability and greater sensitivity than did 33
ddPCR in clinical samples. Both digital and real-time PCR provide accurate CMV viral load data over a 34
wide linear dynamic range. Digital PCR may provide an opportunity to reduce quantitative variability 35
currently seen using real-time PCR, but methods need to be further optimized to match the sensitivity of 36
real-time PCR. 37
38
39
40
3
Introduction 41
Over the past several years, viral load testing has evolved from a highly complex and labor-intensive 42
procedure to a routine part of patient care. Such methods are now integral to a diverse range of clinical 43
practice settings and diagnostic and treatment guidelines. These include their implementation in 44
patients with HIV or hepatitis, and post-transplant monitoring for cytomegalovirus, Epstein-Barr virus, 45
adenovirus, and BK virus, among others. Quantitative values have been used to follow efficacy of anti-46
viral therapy and to help determine changes in that therapy. Rising viral burden has been used as a 47
trigger for pre-emptive treatment, to prevent symptomatic infection. CMV testing is the archetype for 48
the latter application; viral load testing by real-time PCR in particular has changed the epidemiology of 49
CMV disease in transplant patients, and is now a standard part of post-transplant care (1, 4, 15, 16, 18, 50
30). 51
Despite these advances, challenges remain in viral load testing which relate primarily to intrinsic 52
limitations of current methodology and center on the issues of accuracy, standardization, and precision 53
(9, 19, 22, 29). Quantitative determinations by real-time PCR are indirect, depending on relationship of 54
cycle threshold of a test sample to a calibration curve. The latter, in turn, is typically generated by 55
testing a series of known standards, across the linear range of the assay. However, marked variation in 56
assay performance characteristics and in materials used as calibration standards may prevent 57
agreement between different laboratories, even when testing identical material. The use of 58
international standards, such as has been made available by the WHO (7, 11, 17), has helped mitigate 59
this issue, but has not resolved the problem. Standards are available only for a few of the most common 60
target analytes. Even when available, they may not be fully commutable, and thus may behave 61
differently, depending on the assay system utilized (5, 25). 62
63
4
Perhaps even more problematic is the poor reproducibility often seen among quantitative molecular 64
tests. Within and between lab precision, even when using identical methodology, can be surprisingly 65
poor (9, 13, 19). Variations in any aspect of these complex methods are magnified by the multistep 66
nature of the process and by the reliance on calibration curves, which themselves may vary over time. 67
The very nature of relying on the measure of a dynamic process, such as rate of target amplification, 68
carries with it intrinsic fluctuations that one could not expect to fully eliminate. This suggests that a 69
more direct method of quantification may be of value. 70
Droplet digital PCR is such a direct method (3, 12, 14, 20, 21, 26). It relies on limiting partition of PCR 71
reaction volume, such that a positive result in any of a large number of micro-reactions (in this case, 72
20,000) indicates the presences of a single target molecule in a given reaction. The number of positive 73
reactions, together with Poisson’s distribution, can then be used to produce a direct, high confidence 74
measurement of original target concentration. This methodology removes both the reliance on rate-75
based measurements (Ct values) and the need for the use of calibration curves. Studies targeting low-76
copy number genes, typically in the field of molecular oncology, have demonstrated a high degree of 77
sensitivity and precision of digital PCR compared to real-time PCR (6, 10, 23, 27). To-date, limited data 78
exists regarding the application of this new technology to viral load testing, but it promises to markedly 79
improve our ability to reproducibly quantify viruses and to obviate the need for developing a costly 80
series of international standards. 81
Herein, we compared ddPCR to real-time PCR for quantitative detection of CMV in both artificially 82
seeded samples and clinical plasma samples, the former including recently developed international 83
standard materials. 84
85
5
Materials and Methods 86
87
Reference materials 88
World Health Organization International (WHO) Standard for human cytomegalovirus (hCMV) 89
WHO-hCMV was purchased from the National Institute for Biological Standards and Control (NIBSC, 90
Potters Bar Hertfordshire, EN6 3QG United Kingdom) and was prepared from whole virus of the HCMV 91
Merlin strain. It was supplied as in a lyophilized format containing a total of 6.7 log 10 International Units 92
(IU) (5x106IU). One vial of the WHO standard was reconstituted with 1.0 mL of nuclease free water to a 93
concentration of 6.7 log 10 IU/mL (5x106 IU/mL). Dilutions were made in CMV negative human plasma to 94
obtain a 10-fold concentration gradient from 1-6 log 10IU/mL. For each concentration, 6 aliquots of 200 95
μL were extracted on the Qiagen EZ1 XL extractor using the Qiagen EZ1 Virus Mini Kit v2.0 (Qiagen, Inc., 96
Valencia, CA) following the manufacturer’s recommended protocol which includes the addition of an 97
internal control (IC) (7.2 μL). The specimens were eluted in 60 μL, pooled, aliquoted, and stored at -80o C. 98
99
National Institute of Standard & Technology (NIST) Standard Reference Material 100
NIST-HCMV was purchased from NIST (Gaithersburg, MD 20899) and prepared from a bacterial artificial 101
chromosome of CMV Towne ?147 BAC containing the genome of the Towne strain of CMV. The standard 102
was supplied in a unit of three component DNA materials, each component containing 150 μL of purified 103
DNA solution at a different concentration of CMV genome/μL. NIST Component C (127 μL, with a given 104
concentration of 19,641 copies/μL (4.29 log 10 copies /μL) was diluted with 373uL of TE for a 105
concentration of 6.7 log 10 copies/mL (5x106 copies/mL). Dilutions were made in TE to obtain a 10-fold
106
6
concentration gradient from 1-6 log 10 copies/mL and stored at -80o C. As the NIST was received as 107
purified nucleic acid; it was not extracted prior to amplification. 108
109
AcroMetrix? CMV tc Panel 110
The AcroMetrix? CMV tc Panel was purchased from AcroMetrix (Life Technologies, Carlsbad, CA), 111
containing intact, encapsulated viral particles of hCMV strain AD169. The panel consisted of a normal 112
human plasma (tested non-reactive for CMV DNA) and four members over a 10-fold concentration 113
gradient from 2.5 to 5.5 log 10 IU/mL (2.9x102 to 2.9x105 (IU/mL) in human plasma. For each 114
concentration, 200 μL was extracted and eluted in 60 μl on the Qiagen EZ1 XL extractor using the Qiagen 115
EZ1 Virus Mini Kit v2.0 (Qiagen, Inc., Valencia, CA) following the manufacturer’s recommended protocol. 116
Internal control (IC) (7.2 μL) was added pre-extraction. Six aliquots of 200 μL from each concentration 117
were processed, extracts were pooled, aliquoted, and stored at -80?C until usage. 118
119
Patient specimens 120
50 de-identified human plasma specimens were used in the study. Samples were collected from 121
residual material, originally collected for clinical testing at Emory University Hospital from June through 122
August of 2011. Samples were chosen to represent a wide spectrum of CMV concentrations, based on 123
clinical test results. For each clinical specimen, three aliquots of 200 μL were extracted on the Qiagen 124
EZ1 XL extractor using the Qiagen EZ1 Virus Mini Kit v2.0 (Qiagen, Inc., Valencia, CA) following the 125
manufacturer’s recommended protocol which includes the addition of an internal control (IC) (7.2 μL). 126
The specimens were eluted in 60 μL, pooled, aliquoted and stored at -80o C. This protocol was approved 127
by the Emory University Institutional Review Board (IRB).
128
7
129
Droplet Digital PCR (ddPCR) 130
A laboratory developed test (LDT) for CMV detection was used, together with the QX100? Droplet 131
Digital? PCR System (Bio-Rad, Pleasanton, CA). The ddPCR reaction mixture consisted of 10 μL of a 2x 132
ddPCR Mastermix (Bio-Rad), 2 μL of CMV primer/probe mix (artus CMV PCR analyte specific reagent, 133
Qiagen, Valencia, CA), and 5 μL of sample nucleic acid solution in a final volume of 20 μL. The entire 134
reaction mixture was loaded into a disposable plastic cartridge (Bio-Rad) together with 70 μL of droplet 135
generation oil (Bio-Rad) and placed in the droplet generator (Bio-Rad). After processing, the droplets 136
generated from each sample were transferred to a 96-well PCR plate (Eppendorf, Germany). PCR 137
amplification was carried out on a T100 Thermal Cycler (Bio-Rad) using thermal profile of beginning at 138
95?C for 10 minutes, followed by 40 cycles of 94?C for 30 seconds and 60?C for 60 seconds and 1 cycle of 139
98?C for 10 minutes, and ending at 12?C. After amplification, the plate was loaded on the Droplet Reader 140
(Bio-Rad) and the droplets from each well of the plate were read automatically at a rate of 32 wells per 141
hour. ddPCR data was analyzed with QuantaSoft analysis software (Bio-Rad) and the quantification of 142
target molecule was presented as copies per μL of PCR. 143
144
Quantitative real-time PCR 145
Quantitative CMV PCR testing was performed using an LDT incorporating artus CMV PCR analyte specific 146
reagents (Qiagen, Inc., Valencia, CA) on the Qiagen Rotor Gene instrument. For each amplification 147
reaction, 20 μL of purified nucleic acid was added to 30 μL of a reaction mixture of master mix and 148
magnesium. The thermo-cycler parameters were as follows: hold 10 minutes at 95?C followed by 15 149
seconds at 95?C, 30 seconds at 65?C and 20 seconds at 72?C for 45 cycles. A four point standard curve as
150
8
well as a positive and a negative control were included on all runs. The laboratory-determined limit of 151
detection (LOD) in this assay using plasma was 2.3 log 10 IU/mL. 152
Statistical analysis 153
For each standard’s dilution series, the concentration was log-transformed as log 10(concentration+1) for 154
purposes of subsequent statistical analyses. Digital PCR measurements were transformed as 155
log 10(copies/ml+1) and real-time PCR measurements were transformed as log 10(IU/1.38 +1) for purposes 156
of subsequent statistical analyses. The factor of 1.38 was used in the transformation of real-time PCR 157
measurements to convert IU/ml to the copies/ml units reported by digital PCR. One was added prior to 158
log-transformation to avoid obtaining undefined values by taking the log of zero for negative samples. 159
For each technology and standard, the limit of detection (LOD) was defined as the smallest nonzero 160
concentration at which all replicates gave a positive qualitative result. For each standard and 161
technology, the log-transformed measurements at or above the LOD were regressed against the log-162
transformed concentrations by simple least squares. Bartlett’s test(2) was used to compare standard 163
deviations of log-transformed results across technologies for each concentration and standard. 164
For clinical samples, the mean log-transformed measurement was computed for each technology and 165
each clinical sample. Simple least squares regression was used to evaluate the quantitative agreement 166
of the mean values of the two technologies. For each sample, the number of repeats giving a positive 167
result was determined for each technology. The number of samples with a difference in the number of 168
positive results was determined. A binomial test was used to determine the significance of the number 169
of samples with fewer positive results by digital PCR than by real-time PCR. Similarly, the standard 170
deviation of results was computed for each sample and technology. A binomial test was used to 171
determine he statistical significance of the number of samples with a smaller standard deviation by real-172
time PCR than by digital PCR.
173
9
Results 174
Evaluation using quantitative standards 175
The three quantitative standards (WHO, NIST, Acrometrix) were tested as unknowns by both real-time 176
PCR and ddPCR to generate analytical operating characteristics for each method. Each concentration of 177
each standard was run in triplicate on three separate runs (total of nine results per sample). Analysis 178
was performed to determine limit of detection, quantitative linearity, quantitative agreement compared 179
both to nominal concentration and to results from real-time PCR, and reproducibility of results using 180
each methodology. 181
182
As shown in Table 1, each method had the same LOD when using Acrometrix standards (2.48 log IU/mL). 183
However, real-time PCR demonstrated a lower LOD by one log unit dilution for both the WHO and NIST 184
standards (10-fold greater sensitivity than ddPCR). Digital PCR (dPCR) detected 100% of samples at log 3 185
IU/mL using WHO material and at log 4 copies/mL using NIST material. 186
187
Both methods showed excellent linearity above the LOD. For each technology and standard, the 188
estimated slope coefficients were close to 1 (0.93-1.04) and R-squared values were very close to 1 189
(≥0.98), indicating both linearity and correlation between nominal and measured values (Figure 1). 190
Quantitative correlation was also shown between methods (Figure 2). Correlation was reduced, 191
particularly for ddPCR, at the low end of each dilution series, corresponding to analyte concentrations 192
with less than 100% detection. Regression model equations for Figures 1 and 2 are shown in 193
Supplemental Tables 1 and 2, respectively.
194
10
The lower sensitivity of ddPCR may be largely attributed to use of a 4-fold smaller input volume. When 195
we plotted the probability of detection against the nominal total number of input viral copies 196
(copies/mL × input volume in mL), the two technologies showed very similar performance 197
(Supplemental Figure S1). This indicated that the sensitivity difference was largely attributable to the 198
input volume differences. 199
Table 1 shows the standard deviation and mean log 10 transformed copies for each technology, standard 200
and concentration. Only matched concentrations of each standard run with both methods are reported. 201
Overall, the methods produced similar results. The maximum differences between mean measurement 202
values of the two technologies for all concentrations tested were 0.12, 0.37, and 1.27 log 10 copies/mL 203
for Acrometrix, NIST, and WHO standards, respectively. Among concentrations above the LOD for both 204
methods, the maximum differences were 0.12, 0.37, and 0.45. Digital PCR showed significantly less 205
variability than real-time PCR for Acrometrix log 10 concentrations ≥ 4.48 and for WHO log 10 206
concentrations ≥ 4. Real-time PCR showed significantly less variability than digital PCR at NIST log 10 207
concentration = 3 and WHO log 10 concentration = 2. The two technologies did not show significantly 208
different variability for the other concentrations of the standards. 209
210
Evaluation using clinical samples 211
Clinical samples were each tested in duplicate on three separate PCR runs (total of six results per 212
sample, Table S3). Mean viral load and result variability was compared for each method. Figure 3 shows 213
linear regression and Bland-Altman plots comparing the mean quantitative values for ddPCR with those 214
of real-time PCR, demonstrating close agreement between the two systems (mean difference=-0.247 215
log 10 copies/ml; SD=0.336 log 10 copies/ml).
216
11
Digital PCR had a significantly lower positivity rate than real-time PCR. Digital PCR obtained fewer 217
positive results among its six repeated replicates than did real-time PCR for 14 samples but the converse 218
was not true for any samples (p = 0.0001). Additionally, the variability of real-time results was less than 219 that of digital PCR results for 34 of 50 (68%) samples (95% CI = 53%- 81%; p = 0.015). 220
221
12
Discussion 222
This study offers one of the earliest published assessments of digital PCR as a viable means for 223
quantitative detection of DNA viremia in clinical samples. Previous work with this relatively new 224
technology has focused on its application in molecular oncology (6, 21). However, the utility of a direct 225
(rather than relative) measure of quantification for the field of clinical molecular virology cannot be 226
overstated, particularly if demonstrated to be accurate, sensitive, and precise. Findings here show 227
ddPCR to have such potential for the measurement of CMV. While the current assay showed somewhat 228
lower sensitivity than real-time methodology, other operating characteristics were comparable. 229
Quantitative accuracy was high and generated viral loads that matched well with nominal values of 230
international and commercial quantitative standards. Linearity and quantitative correlation with real-231
time PCR results were high; and precision matched or exceeded that of real-time PCR, when assessed 232
within the analytical measurement range of the test. Furthermore, our data suggest that the lower 233
sensitivity and greater variability of this ddPCR may be attributed to a lower input volume; however, our 234
experiment was not designed to definitively determine the impact of input volume on sensitivity. 235
Another experiment designed explicitly for that purpose may be necessary to confirm our 236
interpretation. Clinical samples showed greater variability with ddPCR than with real-time PCR, while 237
the converse was true for reference materials. Samples detected near the LOD of either method showed 238
greater variability. Because the LOD of ddPCR was somewhat higher than for real-time PCR, the results 239
of clinical samples disproportionately fell close to the LOD of ddPCR, potentially accounting for the 240
increased variability (since clinical samples tended to have a lower viral load overall, while reference 241
materials had their concentrations evenly distributed along the linear range of the assays). 242
243
13
The disadvantages of real-time PCR for such measurements have been well-documented. Several 244
authors have now shown a high degree of quantitative variability among such assays, not only for those 245
targeting CMV, but also for other blood-borne viruses (8, 9, 19, 22). Laboratory developed assays are 246
particularly prone to such variability, and a wide range of factors have been shown to affect both 247
accuracy and precision of these tests (9). The development of international standards for CMV and other 248
such viruses promises to mitigate such problems (particularly that of accuracy), but numerous issues 249
remain. These include the difficulty of generating widely commutable standards and other aspects of 250
assay design that impinge on amplification efficiency, all integrally related to the generation of 251
reproducible results when measurement depends on relative rates of amplification, as with real-time 252
methods. In addition, gains seen in the development of international standards and automated, FDA 253
cleared assays are unlikely to be duplicated for the wide range of viruses currently of clinical interest 254
and may be limited only to those with a strong commercial market. 255
The clinical import of improved assay characteristics is multifold. Present methods for measuring CMV 256
load may show quantitative variability among laboratories exceeding four orders of magnitude (9, 19). 257
Quantitative accuracy and precision are crucial to assay interpretation and to the ability to set 258
meaningful, universal treatment thresholds, both for clinical disease attribution and for pre-emptive 259
therapeutic strategies. In addition, the portability of results can have direct implications for patient care. 260
Currently, ongoing treatment of patients who move from one institution to another may be challenging 261
and require new baseline viral loads. Comparability of clinical study results that rely on such 262
measurements is a final issue of importance, fundamental to advancing patient care. While the recent 263
availability of an FDA-cleared, commercial assay may improve the situation for CMV testing, its impact 264
remains uncertain. Digital PCR may address all of these issues, utilizing end-point PCR for direct 265
measurement of viral loads, potentially improving accuracy and precision over current, real-time 266
methods.
267
14
Since it was initially described in the 1990’s, other authors have shown digital PCR to have advantages in 268
several early manuscripts, focused primarily on research applications in molecular genetics and 269
oncology. Specifically, it has been used for mutational analysis, assessment of allelic imbalance, cancer 270
detection, and allelic expression analysis (3, 6, 21). Its ability to detect and accurately quantify minority 271
mutations in a predominating background of normal sequence has proven advantageous over other PCR 272
methodologies. This capability has also been exploited in the area of prenatal diagnostics, where its use 273
for the detection of fetal aneuploidies and single gene Mendelian disorders using maternal plasma 274
samples has also been described (31). Few authors, however, have yet investigated the application of 275
this technology to infectious disease diagnostics. Experimental systems have demonstrated proof-of-276
principle for quantification of adenoviral genome copies, GB virus in transfected cell culture lines, serial 277
dilutions of hepatitis C virus and human immunodeficiency virus (HIV) RNA, and purified, serially diluted 278
HIV RNA from two clinical samples (14, 24, 28). The latter studies demonstrated a 4 log 10 linear range 279
and variable agreement with real-time PCR methods. None of these studies examined a commercially 280
available digital PCR system, amenable to routine clinical use; and none utilized large numbers of clinical 281
samples in a clinical laboratory setting. This report supports the clinical applicability of ddPCR for 282
routine use in clinical diagnostic molecular virology. The system used here has a relatively small 283
footprint (requiring less than two linear feet of bench space for both the droplet generator and reader, 284
together). It can be used at roughly the same cost as real-time PCR (as most reagents used here were 285
identical in the two systems) and has scalable, rapid throughput (6.5 hours for 96 reactions and 2.2 286
hours for 8 reactions, compared to 4 hours for 33 real-time PCR reactions). Hands on time for ddPCR 287
was 25 minutes for 8 and 120 minutes for 96 reactions, while for real-time PCR it was 85 minutes for 33 288
reactions. This now makes digital PCR practical for routine use in clinical laboratories, adding relevance 289
to the results shown above.
290
15
The conclusions here are of course limited based on the number and genetic heterogeneity of clinical 291
samples available. In addition, the conclusions here are drawn based on comparison using only one 292
assay design, targeting only one of a number of viruses for which such a method may have clinical utility. 293
Indeed, work ongoing in our laboratory seeks to extend the present study by looking at other 294
comparator assays and other viruses, particularly those with RNA genomes. In addition, the 295
development and use of other digital platforms will be crucial to building upon this work. The 296
somewhat higher LOD, compared to real-time PCR seen can be accounted for entirely by the small total 297
sample volume, compared to the real-time system used (5μL for ddPCR, compared to 20μL for real-time 298
PCR, statistical analysis not shown). Increased reaction volume for ddPCR, currently limited by system 299
design, might also lead to improved quantitative correlation between methods at low concentrations of 300
analyte and might extend the precision advantage of ddPCR, already seen higher concentrations. Finally, 301
future studies comparing ddPCR to the recently released FDA-approved test for CMV viral load testing 302
(and others, as they become available) may prove valuable in assessing the future roles for each 303
technology. 304
305
The data shown here confirm earlier work in experimental oncology and show proof of principle for the 306
application of ddPCR for routine use in clinical, molecular virology. The advantages of direct 307
quantification by end-point, limiting partition PCR could obviate the need for development and purchase 308
of costly quantitative calibrators. This paradigm shift could reduce testing costs, improve accuracy and 309
precision of quantitative viral load assays. Such improvements will have direct clinical utility, particularly 310
in diagnostic microbiology and virology laboratories serving immunocompromised patients, where the 311
number of viruses of importance continues to outstrip our ability to develop methods for quantitative 312
detection, and where our present diagnostic tools remain severely limited in utility due to suboptimal 313
operating characteristics. These findings support both the continued development of this technology
314
16
and further work to explore its value for routine use in clinical diagnostic testing for infectious 315
pathogens. 316
317
Acknowledgements: 318
Artus CMV RG PCR reagents were graciously provided by Qiagen, Inc. (Valencia, CA). This work was 319
supported in part by the Anderson Charitable Foundation, the American Lebanese Syrian Associated 320
Charities (ALSAC), and the Emory Center for AIDS Research (P30 AI050409). Dr. Caliendo is a co-321
investigator on a CMV viral load clinical trial that is funded by Qiagen. 322
323
324
325
326
327
References 328
1. Barker, J. N., R. E. Hough, J. A. van Burik, T. E. DeFor, M. L. MacMillan, M. R. O'Brien, and J. E. 329
Wagner. 2005. Serious infections after unrelated donor transplantation in 136 children: impact 330
of stem cell source. Biol Blood Marrow Transplant 11:362-370. 331
2. Bartlett, M. S. 1937. Properties of sufficiency and statistical tests. Proceedings of the Royal 332
Society of London Series A 160:268–282.
333
17
3. Bhat, S., J. Herrmann, P. Armishaw, P. Corbisier, and K. R. Emslie. 2009. Single molecule 334
detection in nanofluidic digital array enables accurate measurement of DNA copy number. Anal 335
Bioanal Chem 394:457-467. 336
4. Boeckh, M., and P. Ljungman. 2009. How we treat cytomegalovirus in hematopoietic cell 337
transplant recipients. Blood 113:5711-5719. 338
5. Caliendo, A. M., M. D. Shahbazian, C. Schaper, J. Ingersoll, D. Abdul-Ali, J. Boonyaratanakornkit, 339
X. L. Pang, J. Fox, J. Preiksaitis, and E. R. Schonbrunner. 2009. A commutable cytomegalovirus 340
calibrator is required to improve the agreement of viral load values between laboratories. Clin 341
Chem 55:1701-1710. 342
6. Diehl, F., and L. A. Diaz, Jr. 200
7. Digital quantification of mutant DNA in cancer patients. Curr 343
Opin Oncol 19:36-42. 344
7. Fryer, J. F. H., A. B.; Anderson, R.; Minor, P. D.;the Collaborative Study Group. 2010. 345
Collaborative Study to Evaluate the PRoposed 1st WHO International Standard for Human 346
Cytomegalovirus (HCMV) for Nucleic Acid Amplification (NAT)-Based Assays. National Institute 347
for Biological Standards and Control. 348
8. Hayden, R. T., K. M. Hokanson, S. B. Pounds, M. J. Bankowski, S. W. Belzer, J. Carr, D. Diorio, M. 349
S. Forman, Y. Joshi, D. Hillyard, R. L. Hodinka, M. N. Nikiforova, C. A. Romain, J. Stevenson, A. 350
Valsamakis, H. H. Balfour, Jr., and U. S. E. W. Group. 2008. Multicenter comparison of different 351
real-time PCR assays for quantitative detection of Epstein-Barr virus. J Clin Microbiol 46:157-352
163. 353
9. Hayden, R. T., X. Yan, M. T. Wick, A. B. Rodriguez, X. Xiong, C. C. Ginocchio, M. J. Mitchell, and A. 354
M. Caliendo. 2012. Factors contributing to variability of quantitative viral PCR results in 355
proficiency testing samples: a multivariate analysis. J Clin Microbiol 50:337-345.
356
18
10. Hindson, B. J., K. D. Ness, D. A. Masquelier, P. Belgrader, N. J. Heredia, A. J. Makarewicz, I. J. 357
Bright, M. Y. Lucero, A. L. Hiddessen, T. C. Legler, T. K. Kitano, M. R. Hodel, J. F. Petersen, P. W. 358
Wyatt, E. R. Steenblock, P. H. Shah, L. J. Bousse, C. B. Troup, J. C. Mellen, D. K. Wittmann, N. G. 359
Erndt, T. H. Cauley, R. T. Koehler, A. P. So, S. Dube, K. A. Rose, L. Montesclaros, S. Wang, D. P. 360
Stumbo, S. P. Hodges, S. Romine, F. P. Milanovich, H. E. White, J. F. Regan, G. A. Karlin-361
Neumann, C. M. Hindson, S. Saxonov, and B. W. Colston. 2011. High-throughput droplet digital 362
PCR system for absolute quantitation of DNA copy number. Anal Chem 83:8604-8610. 363
11. Holden, M. J., R. M. Madej, P. Minor, and L. V. Kalman. 2011. Molecular diagnostics: 364
harmonization through reference materials, documentary standards and proficiency testing. 365
Expert Rev Mol Diagn 11:741-755. 366
12. Kiss, M. M., L. Ortoleva-Donnelly, N. R. Beer, J. Warner, C. G. Bailey, B. W. Colston, J. M. 367
Rothberg, D. R. Link, and J. H. Leamon. 2008. High-throughput quantitative polymerase chain 368
reaction in picoliter droplets. Anal Chem 80:8975-8981. 369
13. Kraft, C. S., W. S. Armstrong, and A. M. Caliendo. 2012. Interpreting quantitative 370
cytomegalovirus DNA testing: understanding the laboratory perspective. Clin Infect Dis 54:1793-371
1797. 372
14. Kreutz, J. E., T. Munson, T. Huynh, F. Shen, W. Du, and R. F. Ismagilov. 2011. Theoretical design 373
and analysis of multivolume digital assays with wide dynamic range validated experimentally 374
with microfluidic digital PCR. Anal Chem 83:8158-8168. 375
15. Ljungman, P. 2010. Molecular monitoring of viral infections after hematopoietic stem cell 376
transplantation. Int J Hematol 91:596-601. 377
16. Ljungman, P., M. Hakki, and M. Boeckh. 2011. Cytomegalovirus in hematopoietic stem cell 378
transplant recipients. Hematol Oncol Clin North Am 25:151-169.
379
19
17. Madej, R. M., J. Davis, M. J. Holden, S. Kwang, E. Labourier, and G. J. Schneider. 2010. 380
International standards and reference materials for quantitative molecular infectious disease 381
testing. J Mol Diagn 12:133-143. 382
18. Mori, T., and J. Kato. 2010. Cytomegalovirus infection/disease after hematopoietic stem cell 383
transplantation. Int J Hematol 91:588-595. 384
19. Pang, X. L., J. D. Fox, J. M. Fenton, G. G. Miller, A. M. Caliendo, and J. K. Preiksaitis. 2009. 385
Interlaboratory comparison of cytomegalovirus viral load assays. Am J Transplant 9:258-268. 386
20. Pinheiro, L. B., V. A. Coleman, C. M. Hindson, J. Herrmann, B. J. Hindson, S. Bhat, and K. R. 387
Emslie. 2012. Evaluation of a droplet digital polymerase chain reaction format for DNA copy 388
number quantification. Anal Chem 84:1003-1011. 389
21. Pohl, G., and M. Shih Ie. 2004. Principle and applications of digital PCR. Expert Rev Mol Diagn 390
4:41-47. 391
22. Preiksaitis, J. K., X. L. Pang, J. D. Fox, J. M. Fenton, A. M. Caliendo, and G. G. Miller. 2009. 392
Interlaboratory comparison of epstein-barr virus viral load assays. Am J Transplant 9:269-279. 393
23. Sanders, R., J. F. Huggett, C. A. Bushell, S. Cowen, D. J. Scott, and C. A. Foy. 2011. Evaluation of 394
digital PCR for absolute DNA quantification. Anal Chem 83:6474-6484. 395
24. Shen, F., B. Sun, J. E. Kreutz, E. K. Davydova, W. Du, P. L. Reddy, L. J. Joseph, and R. F. Ismagilov. 396
2011. Multiplexed quantification of nucleic acids with large dynamic range using multivolume 397
digital RT-PCR on a rotational SlipChip tested with HIV and hepatitis C viral load. J Am Chem Soc 398
133:17705-17712. 399
25. Vesper, H. W., W. G. Miller, and G. L. Myers. 2007. Reference materials and commutability. The 400
Clinical biochemist Reviews / Australian Association of Clinical Biochemists 28:139-147. 401
26. Vogelstein, B., and K. W. Kinzler. 1999. Digital PCR. Proc Natl Acad Sci U S A 96:9236-9241.
402
20
27. Whale, A. S., J. F. Huggett, S. Cowen, V. Speirs, J. Shaw, S. Ellison, C. A. Foy, and D. J. Scott. 2012. 403
Comparison of microfluidic digital PCR and conventional quantitative PCR for measuring copy 404
number variation. Nucleic acids research 40:e82. 405
28. White, R. A., 3rd, S. R. Quake, and K. Curr. 2012. Digital PCR provides absolute quantitation of 406
viral load for an occult RNA virus. J Virol Methods 179:45-50. 407
29. Wolff, D. J., D. L. Heaney, P. D. Neuwald, K. A. Stellrecht, and R. D. Press. 2009. Multi-Site PCR-408
based CMV viral load assessment-assays demonstrate linearity and precision, but lack numeric 409
standardization: a report of the association for molecular pathology. J Mol Diagn 11:87-92. 410
30. Zhang, L. F., Y. T. Wang, J. H. Tian, K. H. Yang, and J. Q. Wang. 2011. Preemptive versus 411
prophylactic protocol to prevent cytomegalovirus infection after renal transplantation: a meta-412
analysis and systematic review of randomized controlled trials. Transpl Infect Dis 13:622-632. 413
31. Zimmermann, B. G., S. Grill, W. Holzgreve, X. Y. Zhong, L. G. Jackson, and S. Hahn. 2008. Digital 414
PCR: a powerful new tool for noninvasive prenatal diagnosis? Prenat Diagn 28:1087-1093. 415
416
417