ClinicalValidation Study of Quantitative Real-Time PCR Assay for Detection and Monitoring of BK Virus Nephropathy

Transcription

1 Available online at ClinicalValidation Study of Quantitative Real-Time PCR Assay for Detection and Monitoring of BK Virus Nephropathy 455 Satoru Kudose 1 and Jianli Dong 2 Departments of Medicine 1 and Pathology 2, University of Texas Medical Branch, Galveston, Texas, USA Abstract. BK virus (BKV) causes nephropathy (BKVN) in renal transplant patients, but monitoring of BKV loads provides an opportunity to prevent BKVN. However, because viral load measurement is not standardized, each laboratory must validate their methodology. We performed a retrospective analysis of 1371 plasma and 600 urine BKV loads measured by the laboratory developed real-time polymerase chain reaction (RT-PCR) of BKV DNA and 346 biopsies from 284 patients in our renal transplant program. We assessed the ability of plasma and urine viral loads to predict the presence of BKVN in biopsy using the receiver-operator characteristic curve. We determined that the cut-offs 3.7 and 7.2 log copies/ml have the best sensitivity (100% and 100%) and specificity (97.6% and 97.5%) for the detection of concurrent biopsy with BKVN by plasma and urine viral load, respectively. Also, we determined that the presence of at least two viral loads greater than 2.8 log copies/ml for plasma and 6.4 log copies/ml for urine within 30 days of biopsy can detect BKVN with similar operating characteristics. Lastly, among pairs of urine and plasma viral loads from the same day, we found that 375 of 376 urine viral loads <4 log copies/ml were accompanied by plasma viral loads <2.6 log copies/ ml, a finding which can alleviate the need for plasma viral load for most patients. In summary, our RT-PCR of BKV DNA has good operating characteristics, and our findings above can help in development of a better strategy to monitor BKV. Introduction BK virus (BKV) is a non-enveloped double-stranded DNA virus of the polyomavirus family that infects and establishes latency primarily in urothelium [1-3]. Approximately 80-90% of the population has detectable antibody to BKV, which appears in childhood and remains elevated for life [4]. Although asymptomatic reactivation with viruria occurs in immunocompetent individuals, BKV can cause nephropathy (BKVN) in renal transplant patients. The prevalence of of BKVN in renal transplant patients is 1-10%, with graft loss present in up to 80% of cases within 5 years. BKV is thus considered to be one of the leading causes of graft loss, especially in the first two years of transplantation [5-7]. In the absence of specific antiviral therapy and definitive treatment for BKVN, mainstay treatment is reduction of immunosuppression [8]. Address correspondence to Satoru Kudose, 4411 Spicewood Springs Rd., Rm. 905, Austin, TX 78759, USA, phone: ; e mail: sakudose@utmb.edu BKVN, staged from A to C, is often asymptomatic at its early stage. Even when it is symptomatic, early clinical manifestations is nonspecific. Despite the diagnostic challenge, early detection is imperative, because BKVN at stage C is irreversible [3,9]. Nevertheless, because of localized involvement in stage A and difficulty in distinguishing stage B lesions from acute cellular rejection, diagnosis by biopsy alone is also not without limitations, despite its status as the gold standard [10]. In the absence of well-defined risk factors for BKVN, routine monitoring of BK viral loads presents an opportunity for early detection, since BKVN appears to be preceded by viruria and viremia [2,3,5,6,11]. However, because there is currently no standard method to quantitate the BK viral load, measurements from different laboratories are influenced by choice of primers, probes, targets of viral sequences, and reference materials, resulting in significant variability [12,13]. Thus, each laboratory must validate their methodology analytically and clinically /14/ by the Association of Clinical Scientists, Inc.

2 456 Hence, although cut-offs of 4 and 7 log copies/ml for plasma and urine viral load, respectively, are suggested by both the American Society of Transplantation and Kidney Disease Improving Global Outcomes group to detect concurrent BKVN in biopsy, each lab must set their own cutoffs for the presumptive diagnosis of BKVN, which may then be used to guide therapy [14,15]. Materials and Methods This study was conducted through retrospective chart reviews of patients from the transplant nephrology service at UTMB undergoing routine clinical care in accordance with institutional review board approved protocol. BK viral loads were measured by real-time polymerase chain reaction (RT-PCR) using the previously published protocol developed at UTMB [16]. RT- PCR was carried out using Minor Groove Binder Alert Real-Time PCR reagents and a primer to amplify the VP1 region (Elitech Group, Princeton, NJ). Patient Selection. All patients with renal allograft who had at least one BK viral load drawn at UTMB between September of 2011 and August of 2012 (inclusive) were included in this study. Pathologists determined the presence of BKVN with histologic examination. Additional staining for SV40 large T Ag or electron microscopy was performed as needed. Clinical Data. We extracted age, sex, type and reason for transplant, plasma and urine BK viral loads, renal biopsy results, serum and urine creatinine, and urine protein starting the day before the day of transplant from electronic medical records for each patient. Staging for BKVN and other pathologies in the biopsies were also recorded if available. Analysis. Prevalence of BKVN was computed. Statistical difference between viral loads associated with biopsies consistent with BKVN ( positive biopsies ) and biopsies without BKVN ( negative biopsies ) was assessed with Mann-Whitney test, with significance defined as P<0.05. All computations were carried out using statistical software R. In particular, receiver operating characteristic analysis was conducted using ROCR package. For both graphs and computations, <2.6 and >8.6 log copies/ml were replaced by 2.6 and 8.6 log copies/ml, respectively. Those viral loads outside the reportable range log copies/ml were also recorded and replaced by either 2.6 or 8.6 log copies/ml. Pairs of viral loads obtained from the same day with plasma value greater than urine value were assumed to be erroneous due to presence of inhibitors in urine despite few reports of viremia without viruria and were thus excluded from the computations [17,18]. ROC analysis. ROC analysis was carried out by associating the urine and plasma viral loads closest to the day of biopsy, with or without BKVN, within 30, 45 and 60 days. Biopsies without such a viral load were excluded from this analysis. ROC curves were constructed, and optimal cut-offs were chosen using Youden's index. Positive and negative predictive values were computed using the prevalence obtained from above. To examine the ability of a sustained increase in viral load above a certain threshold to predict the presence of BKVN, another ROC curve was constructed by associating the minimum of at least two viral loads to each biopsy within 30 days of it. Biopsies with only one or no viral load were excluded from this analysis. Cut-offs and predictive values were determined as above. Relationship between urine and plasma viral loads from the same day. For each patient, pairs of viral loads drawn on the same day were collected. Pairs with plasma viral load greater than urine viral load were removed. Results Patient characteristics. The prevalence of BKVN among 284 patients included in the study who had at least one BK viral load drawn between September 2011 and August 2012 was 3.87%. Average age of the patient population was 49.2 years old with a range of 15 to 78. There were 104 females and 180 males. Of the 269 patients for which a reason for transplant was explicitly charted, most common causes were hypertension (31.6%), diabetes mellitus type 2(16%), diabetes mellitus type 1(16.7%), both hypertension and diabetes mellitus type 2(10%), and focal segmental glomerulosclerosis (8.2%). Age and sex distribution and reason for transplant did not differ significantly between patients with BKVN and patient without BKVN. From the 284 patients, 346 biopsies were obtained, with 16 positive biopsies distributed among 11 patients and 330 negative biopsies distributed among 149 patients. 130 patients did not have any biopsies. Of the 1371 plasma and 600 urine viral loads obtained, 1155 (84.2%) of plasma and 389 (64.8%) of urine viral loads were reported as < 2.6 log copies/ml. Furthermore, 210 of 284 patients have viral loads <2.6 log copies/ml for both plasma and urine.

3 Clinical Validation of Real-Time PCR for BK Virus 457 Figure 1. Boxplots of plasma (A) and urine (B) viral loads associated with either positive or negative biopsy within 1 month. Three plasma and two urine viral loads from three patients are filled black. Two pairs of plasma and urine viral loads are from the two patients mentioned in the main text with inadequate biopsies. One plasma viral load is from a patient who developed BKVN later despite the initial biopsy without BKVN. Number of viral loads (#VL) for each group is given below each group. Number at the left side denotes the median. Numbers 163 and 63 represent number of viral loads with 2.6 log copies/ml in each group. Likewise, numbers 9 and 1 represent number of viral loads with 8.6 log copies/ml. Figure 2. ROC curve for plasma (A) and urine (B). Each open circle represents the cut-off in log copies/ml. Two biopsies were notable. In one, the pathologist noted that the absence of BKVN was possibly due to sampling error, since plasma and urine BK viral loads from the same day were 4.4 and 8.4 log copies/ml, respectively. It was not known whether this patient developed BKVN at the time of this writing, since a later biopsy was not available. In another, the biopsy was noted to be inadequate for evaluation, but features were suggestive of acute cellular rejection, with accompanying plasma and urine viral loads of 5.6 and >8.6 log copies/ml, respectively. These biopsies were not excluded from the negative biopsy group in the following analysis.

4 458 Table 1. Various cut-offs for plasma and urine viral load (in log copies/ml) from other studies. Reference by the first Plasma Urine Plasma Plasma Urine Urine author and the year Cut-Off Cut-Off Sens. Spec. Sens. Spec. of publication. Hirsch 2002[22] Randhawa 2004[23] Viscount 2007[24] Costa 2008[25] 4.2* Sung 2008[26] Boudreault 2009[27] Pollara 2011[28] Girmanova 2011[29] 3* Miller 2012[30] Chung 2012[21] Our study Sens. and Spec. stand for sensitivity and specificity, respectively. *serum. whole blood. ROC analysis. Of 346 biopsies, 182 and 91 biopsies had plasma and urine BK viral loads available within 30 days of the biopsy, respectively, as shown in Figure 1. There was a statistically significant difference in the median plasma and urine viral loads between the groups (p<0.05). We found that the cut-off of 3.7 log copies/ml for plasma viral load had the best combination of sensitivity and specificity (100% and 97.6%, respectively) for a single plasma viral load to detect BKVN in a concurrent biopsy within 30 days. Similarly, for urine, the cut-off of 7.2 log copies/ml had sensitivity and specificity of 100% and 97.5%, respectively. With the prevalence of 3.87%, corresponding positive and negative predictive values for plasma were 63% and 100%, respectively. Similarly, positive and negative predictive values were 61.4% and 100% for urine. We noted that the cut-offs of 2.9 and 6.3 log copies/ml for plasma and urine had sensitivity of 100% and specificities of 96.4% and 96.2%, respectively. Associating a viral load within 45 days and 60 days did not change the above cut-offs, and sensitivity and specificity only differed by less than 1%. Because multiple viral loads are often available, a sustained increase in viral loads above a certain threshold may be able to predict the presence of BKVN better than a single measurement. For both plasma and urine viral loads, we found 83 biopsies (8 positive and 75 negative for BKVN) with at least two viral loads available within 1 month. We found that the presence of at least two viral loads greater than 2.8 log copies/ml for plasma and 6.4 log copies/ml for urine within 1 month of biopsy can detect BKVN with sensitivity and specificity of 100% and 98.7% for plasma and 100% and 100% for urine viral loads, respectively. Relationship between urine and plasma viral loads from the same day. There were 542 pairs of urine and plasma viral loads obtained on the same day as shown in Figure 3. From 376 pairs with a urine viral load <4 log copies/ml, we found that plasma viral loads were <2.6 log copies/ml in all but 1 pair from a patient negative biopsy (99.7%). Similarly, almost all (399/400) pairs with a urine load <5 had a plasma load <3, with a single exception from a patient with negative biopsy. Discussion Quantitative PCR for BKV is widely used to monitor BKV infection in order to prevent BKVN. However, different laboratories use different primers, probes, viral targets, and reference material, allowing results from different laboratories to differ by over 1 log copies/ml [12,13]. Hence, despite the recommended cut-off of 4 log copies/ml for plasma viral load to presume a diagnosis of BKVN, each laboratory must set their own cut-offs for the assay [7,14]. The aim of this study was to validate the laboratory developed RT-PCR for BKV for clinical detection and management of BKVN. We determined the

5 Figure 3. Plot of 542 paired viral loads from same day. Note the cut-offs 3.7 and 7.2 (dotted lines) for plasma and urine discriminate positive and negative biopsies well. Two negative biopsies above the cut-offs are biopsies deemed to be inadequate as mentioned in the main text. Note that almost all (399/400) pairs with urine load <5 has plasma load <3 and that almost all (375/376) pairs with urine load <4 has plasma load 2.6 or less. Upward and downward triangles represent pairs on the same day as positive and negative biopsy, respectively. Note that there are fewer biopsies shown in this figure than included in ROC analysis, because of inclusion of biopsies with viral loads within 1 month for ROC analysis. Unit for viral load is log copies/ml. optimal cut-offs for detection of BKVN within 1 month to be 3.7 and 7.2 log copies/ml for plasma and urine viral loads with sensitivities of 100% and 100%, and specificities of 97.6%, and 97.5%, respectively. Our cut-offs are consistent with other studies as shown in Table 1 with excellent operating characteristics. Results from our study are therefore likely applicable to other assays to some extent. Furthermore, since a detailed protocol for our RT-PCR is published, and reagents are commercially available, our RT-PCR can be adopted relatively easily by others. Serial monitoring for BKV infection is generally considered to be beneficial. However, the optimal frequency of monitoring and the best usages of serial trends to guide the clinician to decide whether to perform a biopsy are not known, since validation studies usually examine the ability of a single measurement of viral load to predict the presence of BKVN. We found that the presence of at least two viral loads greater than 2.8 log copies/ml for plasma and 6.4 log copies/ml for urine within 1 month of biopsy can detect BKVN with sensitivity and specificity of 100% and 98.7% for plasma and 100% Clinical Validation of Real-Time PCR for BK Virus 459 and 100% for urine viral loads, respectively. This result suggests that almost any sustained viremia is likely associated with BKVN, while a significant amount of sustained viruria may be present without BKVN, which is consistent with the current model of pathogenesis of BKVN [3]. This result can help guide clinicians in cases where sustained viremia or viruria is present, but none of the individual measurements are sufficiently high enough to reach the cut-offs for a single viral load. To control cost, it may be desirable to refrain from obtaining both urine and plasma viral loads in all patients. Although a study by Pang et al. suggested the use of urine viral loads with plasma loads reserved for those with viruria 7 log copies/ml, this threshold is too conservative for our assay and, most likely, for other assays in Table 1 as well [19]. As shown in Figure 2, 99.7% of patients with a urine viral load <4 log copies/ml had a plasma viral load of <2.6 log copies/ml. Hence, viral replication can be monitored using urine viral load alone for most patients, since 73.9% of patients in our data had both plasma and urine viral loads <2.6 log copies/ml. Since a plasma viral load <2.6 log copies/ml is associated with a urine viral load ranging from <2.6 to 8.3 log copies/ml, as shown in Figure 3, a plasma viral load <2.6 log copies/ml can be falsely reassuring, as even low levels of sustained viremia appear to be associated with BKVN. Hence, despite the 100% negative predictive value for BKVN in concurrent biopsy, and a report of near complete prevention of definitive BKVN, it may be prudent to use urine viral load if the goal is to prevent BKVN, since viruria occurs earlier than viremia by 1-3 months and even low level of sustained viremia may be associated with BKVN [20,21]. This study is limited by its design. It is likely that specificities are inflated compared to the results from a study where viral loads were obtained at prescribed intervals because viral loads were ordered by

6 460 clinicians based on their suspicion of BKVN. Similarly, lower cut-offs obtained for sustained viremia may be due to selective exclusion of negative biopsies because of the requirement to have at least two viral loads around the time of biopsy. In summary, this study establishes reliable thresholds of 3.7 and 7.2 log copies/ml for plasma and urine viral loads, respectively. This finding can be used to identify patients with concurrent BKVN for our laboratory developed RT-PCR assay, for which a published step-by-step protocol using commercial reagents is available. We also showed that lower thresholds to detect BKVN are required in the presence of sustained viremia and viruria. Furthermore, our findings argue for measuring BK virus replication using urine viral load alone initially, with a transition to plasma load once viruria reaches 4 log copies/ml. Acknowledgements We would like to thank the members of the Molecular Diagnostics Laboratory for providing the technical details for the BK viral load test. References 1. Funk GA, Steiger J, Hirsch HH. Rapid dynamics of polyomavirus type BK in renal transplant recipients. J Infect Dis 2006;193(1): Hariharan S. BK virus nephritis after renal transplantation. Kidney Int 2006;69(4): Mihatsch MJ. Polyomavirus nephropathy: a brief review with special emphasis on clinico-patholgical aspects. Prilozi 2012;33(2): Knowles WA, Pipkin P, Andrews N, et al. Population-based study of antibody to the human polyomaviruses BKV and JCV and the simian polyomavirus SV40. J Med Virol 2003;71(1): Balba GP, Javaid B, Timpone JG. BK Polyomavirus Infection in the Renal Transplant Recipient. Infect Dis Clin North Am 2013;27(2): Bechert CJ, Schnadig VJ, Payne DA, et al. Monitoring of BK viral load in renal allograft recipients by real-time PCR assays. Am J Clin Pathol 2010;133(2): Hirsch HH, Brennan DC, Drachenberg CB, et al. Polyomavirus-associated nephropathy in renal transplantation: interdisciplinary analyses and recommendations. Transplantation 2005;79(10): Kuypers DR. Management of polyomavirus-associated nephropathy in renal transplant recipients. Nat Rev Nephrol 2012;8(7): Drachenberg CB, Papadimitriou JC. Polyomavirus-associated nephropathy: update in diagnosis. Transpl Infect Dis 2006;8(2): Drachenberg CB, Papadimitriou JC, Ramos E. Histologic versus molecular diagnosis of BK polyomavirus-associated nephropathy: a shifting paradigm? Clin J Am Soc Nephrol 2006;1(3): Hirsch HH, Steiger J. Polyomavirus BK. Lancet Infect Dis 2003;3(10): Hoffman NG, Cook L, Atienza EE, et al. Marked variability of BK virus load measurement using quantitative real-time PCR among commonly used assays. J Clin Microbiol 2008;46(8): Randhawa P, Kant J, Shapiro R, et al. Impact of genomic sequence variability on quantitative PCR assays for diagnosis of polyomavirus BK infection. J Clin Microbiol 2011;49(12): Chadban SJ, Barraclough KA, Campbell SB, et al. KHA- CARI guideline: KHA-CARI adaptation of the KDIGO Clinical Practice Guideline for the Care of Kidney Transplant Recipients. Nephrology (Carlton) 2012;17(3): Hirsch HH, Randhawa P, Practice AIDCo. BK polyomavirus in solid organ transplantation. Am J Transplant 2013;13 Suppl 4: Wu H, Dong J. Quantification of BK Viral Load. In: Hu P, Hegde M, Lennon PA, editors. Modern Clinical Molecular Techniques: Springer; Bressollette-Bodin C, Coste-Burel M, Hourmant M, et al. A prospective longitudinal study of BK virus infection in 104 renal transplant recipients. Am J Transplant 2005;5(8): Vats A, Shapiro R, Singh Randhawa P, et al. Quantitative viral load monitoring and cidofovir therapy for the management of BK virus-associated nephropathy in children and adults. Transplantation 2003;75(1): Pang XL, Doucette K, LeBlanc B, et al. Monitoring of polyomavirus BK virus viruria and viremia in renal allograft recipients by use of a quantitative real-time PCR assay: one-year prospective study. J Clin Microbiol 2007;45(11): Babel N, Fendt J, Karaivanov S, et al. Sustained BK viruria as an early marker for the development of BKV-associated nephropathy: analysis of 4128 urine and serum samples. Transplantation 2009;88(1): Chung BH, Hong YA, Kim HG, et al. Clinical usefulness of BK virus plasma quantitative PCR to prevent BK virus associated nephropathy. Transpl Int 2012;25(6): Hirsch HH, Knowles W, Dickenmann M, et al. Prospective study of polyomavirus type BK replication and nephropathy in renal-transplant recipients. N Engl J Med 2002;347(7): Randhawa P, Ho A, Shapiro R, et al. Correlates of quantitative measurement of BK polyomavirus (BKV) DNA with clinical course of BKV infection in renal transplant patients. J Clin Microbiol 2004;42(3): Viscount HB, Eid AJ, Espy MJ, et al. Polyomavirus polymerase chain reaction as a surrogate marker of polyomavirus-associated nephropathy. Transplantation 2007;84(3): Costa C, Bergallo M, Astegiano S, et al. Monitoring of BK virus replication in the first year following renal transplantation. Nephrol Dial Transplant 2008;23(10): Sung H, Choi BH, Pyo YJ, et al. Quantitation of BK virus DNA for diagnosis of BK virus-associated nephropathy in renal transplant recipients. J Korean Med Sci 2008;23(5): Boudreault AA, Courtemanche C, Latulippe E, et al. Screening for polyomavirus associated nephropathy in renal transplantation with blood viral load measurement. J Clin Virol 2009;45(4): Pollara CP, Corbellini S, Chiappini S, et al. Quantitative viral load measurement for BKV infection in renal transplant recipients as a predictive tool for BKVAN. New Microbiol 2011;34(2): Girmanova E, Brabcova I, Bandur S, et al. A prospective longitudinal study of BK virus infection in 120 Czech renal transplant recipients. J Med Virol 2011;83(8): Miller S, Liverman CS, Post L, et al. Analytical and clinical performance characteristics of the Simplexa BK virus quantitative PCR assay for the diagnosis of polyomavirus-associated nephropathy in renal transplant recipients using plasma and urine specimens. J Clin Virol 2012;55(4):