Rapid ADAMTS13 activity assays for thrombotic thrombocytopenic purpura: a systematic review and meta-analysis Open Access

Thrombotic thrombocytopenic purpura (TTP) is a rare thrombotic microangiopathy (TMA) caused by severe immune-mediated or congenital deficiency (<10% of normal activity) of ADAMTS13 (a disintegrin and metalloproteinase with thrombospondin motif 13), which results in microvascular thromboses and potentially fatal end-organ ischemia.^1-3 There is clinical urgency in initiating treatment of TTP given that delay in therapeutic plasma exchange (TPE) is associated with increased mortality.^4,5 

TTP is diagnosed by testing patients with clinical suspicion based on clinical judgment and risk assessment scores for severe deficiency of ADAMTS13 activity,^6,7 ideally before the receipt of TPE or blood products.³ In patients with intermediate or high clinical suspicion for TTP, empiric treatment with TPE and corticosteroids is recommended and early administration of caplacizumab is suggested while awaiting ADAMTS13 activity results.³ 

There are multiple assays in use for the measurement of ADAMTS13 activity. Two commonly used assays are commercial or standardized in-house enzyme-linked immunosorbent assays (ELISAs) and fluorescence resonance energy transfer (FRET)-based assays using recombinant von Willebrand factor (VWF) substrates. Although these assays have an analytical turnaround time (TAT) of ∼3 hours, they are generally offered only by specialized coagulation laboratories owing to the need for skilled technicians, resulting in batching of assays during core business hours and often delaying results for ≥1 day.^8-11 

Owing to these limitations, commercial rapid assays for ADAMTS13 activity (hereafter, rapid assays) have recently been developed to test single patient samples with relatively less technical expertise and an analytical TAT of <1 hour.^12-14 If a rapid assay rules out TTP, the adverse effects and costs of empiric TPE, steroids, and caplacizumab may be averted and earlier evaluation for alternative causes of TMA may be initiated. If a rapid assay for TTP is positive, a patient may have earlier rituximab administration to lessen the risk of immune TTP (iTTP) exacerbation and relapse.¹⁵ Discrepant results between rapid and reference standard assays have been reported, especially in cases of low clinical suspicion for TTP, iTTP in remission, sepsis, disseminated intravascular coagulation, malignancy, pregnancy, and icteric samples.^9,16-18 To evaluate the diagnostic accuracy of rapid assays, we undertook a systematic review and meta-analysis of studies comparing the performance characteristics of rapid assays with reference standard assays for patients with suspected or known history of TTP.

The protocol for this systematic review was registered in PROSPERO (#CRD42024546185). Our study was conducted according to Preferred Reporting Items for Systematic Reviews and Meta-Analyses guidelines.¹⁹ 

We electronically searched 4 databases (the Cochrane Library, Embase, PubMed, and Scopus) from database inception to 30 May 2024 for studies comparing rapid ADAMTS13 assays with reference standard assays in patients with suspected or confirmed TTP. iTTP and congenital TTP (cTTP) were included. A health science librarian was consulted to develop a search strategy and relevant keywords (see the supplemental Information, available on the Blood website).

Studies were included if they tested patient samples that were being evaluated for diagnosis or monitoring of TTP using a rapid ADAMTS13 activity assay and a reference standard assay against which the performance of the rapid assay could be compared. Rapid assays were defined as ADAMTS13 activity assays designed for single patient sample use with an analytical TAT of <1 hour. Reference standard assays were defined as commercially available or in-house ELISAs or FRET-based assays using recombinant VWF substrates as such assays are recommended for frontline use based on expert consensus, have undergone external quality assessment, and are widely described and frequently used.^8-11 Studies using biobanked or thawed patient samples were permitted given that preliminary evidence review suggested no difference between fresh and thawed plasma samples.^17,20 

Studies were excluded if they were case reports, editorials, or reviews; were non-English-language publications; or did not include sufficient information to compare or calculate ADAMTS13 activity assay performance. Data not included in the relevant study publications were requested from the corresponding author, and studies were included if sufficient unpublished information to calculate assay performance was provided. Rapid assays only described in a single study were excluded from the meta-analysis.

References were imported into Covidence (Veritas Health Innovation, Melbourne, Australia), a data management system.²¹ Three investigators (S.R.D., H.T., and C.D.) independently completed title and abstract screening with discrepancies and disagreements resolved with consensus discussion between 2 investigators (S.R.D. and A.C.). Full-text review of references included after title and abstract screening followed the same protocol. References of included studies were hand searched to identify other potentially eligible studies.

For each included reference, 2 investigators (S.R.D. and either H.T. or C.D.) independently extracted data using a standardized form. Discrepancies were resolved by consensus discussion between 3 investigators (S.R.D., H.T., and C.D.), and disagreements were arbitrated by a fourth investigator (A.C.). Data extraction included study design, setting, and population; type and characteristics of reference standard assay; type and characteristics of rapid assay; diagnoses of patient population; PLASMIC or French score if reported^6,7; number of true positive (TP), false positive (FP), true negative (TN), and false negative (FN) results; and final diagnoses of samples with discrepant results by rapid and reference standard assays. TP, FP, TN, and FN results were determined using 10% as a cutoff value for defining severe ADAMTS13 activity deficiency per International Society on Thrombosis and Haemostasis guidelines.³ For example, an ADAMTS13 activity of <10% by reference standard assay and ≥10% by a rapid ADAMTS13 assay would be coded as a FN.

Risk-of-bias assessment was completed by 2 independent investigators (H.T. and C.D.) with discrepancies resolved by a third investigator (S.R.D.) using QUADAS-2, a tool for quality assessment of diagnostic accuracy studies.²² QUADAS-2 assesses risk of bias across the domains of patient selection, index test, reference standard, and flow and timing, the first 3 of which are also assessed for applicability concerns.

We conducted meta-analyses on several outcomes of interest to provide a comprehensive investigation of rapid assays. The primary outcome was the diagnostic accuracy of each rapid ADAMTS13 activity assay relative to reference standard assays. In particular, we conducted a univariate random effect meta-analysis to obtain the overall estimates for sensitivity, specificity, positive predictive value (PPV), and negative predictive value (NPV) using the reported values, including TP, FP, TN, and FN from the included studies.²³ 

In each univariate meta-analysis, the bivariate generalized linear mixed model by Chu and Cole²⁴ was applied to account for the presence of zero cells, given that this method does not require arbitrary continuity corrections. Continuity corrections were used solely for visualizing study-specific sensitivity, specificity, PPV, and NPV. For analyses without zero cells, the metaprop function from the R package meta (R version 4.1.2, R Project for Statistical Computing) was used.²⁵ The visualization of the study-specific estimates and the estimated overall estimates was implemented using the forestplot R package.²⁶ Heterogeneity between studies was assessed using Higgins’ and Thompson’s $I2$ and tested using Cochran’s Q statistics for statistically significant differences between subgroups, which were interpreted using standard guidelines.^27,28 

In addition, we conducted a secondary analysis to compare each rapid ADAMTS13 activity assay with ELISA and FRETS-VWF73 assays separately. Similar to the primary analyses, we performed univariate random-effects meta-analyses to calculate the overall proportion of discrepant results between each rapid assay and the reference standard assays. The study-specific discrepancy was calculated as the FP and FN values divided by the total sample size within each study.

Recognizing that reference standard assays for ADAMTS13 activity are not perfect,²⁹ we also evaluated concordance between each rapid ADAMTS13 activity assay and reference standard assay. Concordance was measured by Cohen’s kappa (κ) for agreement between the rapid and reference standard assays. The vcd R package was used to calculate study-specific κ statistics and meta R package was used to obtain overall estimates. The interpretation of the Cohen’s κ coefficient for agreement followed previously described guidelines with agreement interpreted as slight for 0 to 0.20, fair for 0.21 to 0.40, moderate for 0.41 to 0.60, substantial for 0.61 to 0.80, and almost perfect for 0.81 to 1.00.³⁰ 

Of the 502 citations identified, 66 were included for full-text review. Of these, 18 met eligibility criteria for inclusion. Upon requesting data from authors, 1 additional study in submission was also included, yielding a total of 19 studies representing 4207 patient samples. The Preferred Reporting Items for Systematic Reviews and Meta-Analyses flow diagram of study selection are presented in Figure 1. Three distinct rapid assays were evaluated in multiple studies and were included in the meta-analysis: (1) HemosIL AcuStar ADAMTS13 Activity chemiluminescence immunoassay (CLIA) (hereafter, HemosIL AcuStar CLIA; Werfen/Instrumentation Laboratory, Bedford, MA), (2) Technofluor ADAMTS13 Activity FRET assay (hereafter, Technofluor FRET assay; Technoclone, Vienna, Austria), and (3) Technoscreen ADAMTS13 activity kit (hereafter, Technoscreen assay; Technoclone, Vienna, Austria). Characteristics of rapid assays are presented in Table 1.

Given that 6 studies reported multiple comparisons between rapid and reference standard assays, the 19 eligible references comprised a total of 27 unique comparisons of rapid assay performance characteristics with reference standard assays. Characteristics of included studies are presented in Table 2. Studies were primarily retrospective (n = 16, 84.2%) with 5 multicenter or multilaboratory evaluations (26.3%). Most of the 27 comparisons (16, 59.3%) used an ELISA as the reference standard assay; the others (11, 40.7%) used a FRETS-VWF73 assay. TTP prevalence (inclusive of cTTP and iTTP and of patients with acute TTP and a history of TTP) was reported in 13 studies with a median of 51.7% (range, 39.6%-100%). Eight studies reported the indication for ADAMTS13 testing in patients with TTP. Of these, 25.7% of samples were tested for TTP diagnosis and the remainder were tested for monitoring in patients with a confirmed history of TTP. No studies stratified results by PLASMIC or French scores.

The performance characteristics of the HemosIL AcuStar CLIA were assessed across 3533 samples in 18 comparisons from 15 studies.^{12,14,17,18,31,32,34,36,37,39-42} Relative to reference standard assays, the HemosIL AcuStar CLIA demonstrated sensitivity of 0.98 (95% confidence interval [CI], 0.94-1.00) and specificity of 0.99 (95% CI, 0.97-1.00) (Figure 2A) and PPV of 0.96 (95% CI, 0.90-0.98) and NPV of 0.99 (95% CI, 0.98-1.00) (Figure 3A). There was significant between-study heterogeneity for specificity (I² = 0.46; P = .02); the remainder of performance characteristics had no significant between-study heterogeneity. Results were similar whether the ELISA or FRETS-VWF73 assay was used as the reference standard (supplemental Information).

The concordance (κ) of the HemosIL AcuStar CLIA was 0.92 (95% CI, 0.89-0.95; Figure 4A), indicating almost perfect agreement of the HemosIL AcuStar CLIA relative to reference standard assays but with significant between-study heterogeneity (I² = 0.62; P$≤$ .01). The pooled proportion of discrepant results between the HemosIL AcuStar CLIA and reference standard assays was 0.03 (95% CI, 0.02-0.04; Figure 5A) with significant between-study heterogeneity (I² = 0.39; P = .04).

The performance characteristics of the Technofluor FRET assay were assessed across 331 samples in 5 comparisons from 4 studies.^14,33,35,36 Relative to reference standard assays, the Technofluor FRET assay demonstrated sensitivity of 0.93 (95% CI, 0.86-0.96) and specificity of 0.98 (95% CI, 0.95-0.99) (Figure 2B) and PPV of 0.97 (95% CI, 0.85-1.00) and NPV of 0.96 (95% CI, 0.93-0.98) (Figure 3B). Results were numerically similar when only the ELISA was used as the reference standard (supplemental Information). Given that only 2 studies compared the Technofluor FRET assay with FRETS-VWF73 assays, a meta-analysis for this comparison was not completed. No performance characteristic demonstrated significant between-study heterogeneity.

The concordance between the Technofluor FRET assay and reference standard assays was 0.90 (95% CI, 0.82-0.98; Figure 4B), indicating almost perfect agreement. The pooled proportion of discrepant results between the Technofluor FRET assay and reference standard assays was 0.04 (95% CI, 0.02-0.06; Figure 5B).

The performance characteristics of the Technoscreen assay were assessed across 343 samples in 4 comparisons from 3 studies.^13,20,38 For the primary outcome of diagnostic accuracy relative to reference standard assays, the Technoscreen assay demonstrated sensitivity of 0.98 (95% CI, 0.42-1.00) and specificity of 0.87 (95% CI, 0.78-0.92) (Figure 2C) and PPV of 0.71 (95% CI, 0.59-0.80) and NPV of 0.99 (95% CI, 0.72-1.00) (Figure 3C). In the subgroup of comparisons using the ELISA as the reference standard assay, the Technoscreen assay demonstrated sensitivity of 0.97 (95% CI, 0.48-1.00), specificity of 0.87 (95% CI, 0.76-0.94), PPV of 0.73 (95% CI, 0.61-0.82), and NPV of 0.98 (95% CI, 0.77-1.00). Given that only 1 study compared the Technoscreen assay with a FRETS-VWF73 assay, a subgroup analysis was not performed. There was significant between-study heterogeneity for specificity between the Technoscreen assay and reference standard assays (I² = 0.71; P$=$ .02).

The concordance of the Technoscreen assay was 0.75 (95% CI, 0.67-0.83; Figure 4C), indicating substantial agreement relative to reference standard assays. The pooled proportion of discrepant results between the Technoscreen assay and reference standard assays was 0.11 (95% CI, 0.07-0.16; Figure 5C).

Ten of the 21 included studies described discrepant results between rapid and reference standard assays, and 51 of the 149 discrepant results were annotated with clinical information. Of the 30 rapid assay–positive and reference standard–negative discrepancies with clinical information, clinical diagnoses included iTTP in remission (n = 11), acute iTTP episode (n = 6) including 2 iTTP relapses, cTTP (n = 1), other TMAs (n = 5), sepsis (n = 3), and unknown diagnosis (n = 4). Among the 21 rapid assay–negative and reference standard–positive discrepancies with clinical information, clinical diagnoses were acute iTTP episodes (n = 14; 2 of whom were pregnant), cTTP (n = 3), other TMA (n = 3), and iTTP in clinical remission (n = 1).

Risk of bias was assessed by QUADAS-2 criteria for all studies with published abstracts or articles (Table 3). Risk of bias for patient selection was unclear in 8 studies, primarily owing to limited reporting of whether a consecutive or random sample was used; however, all studies were considered to have low concern regarding applicability in this domain. One study was assessed to have a high risk of bias and applicability concerns with respect to the index test domain owing to an unexpectedly high number of low Technoscreen assay results. Quality control evaluation revealed a problematic lot of reagent after test devices, kits, and fresh and frozen plasma samples demonstrated consistent results.²⁰ All studies were considered to have a low risk of bias and applicability concerns regarding the reference standard domain. One study reported autofluorescence by FRETS-VWF73 assay, which led to the use of a commercial ELISA for selected samples. This was judged to constitute a high risk of bias in the flow and timing domain given the differential use of reference standard assays for different samples.¹⁸ 

TTP is a life-threatening TMA with a high risk of death in the absence of urgent treatment.^1,4 Current reference standard ADAMTS13 assays are technically demanding and time consuming, leading to limited availability and days-long TAT. Rapid assays for ADAMTS13 activity included in this systematic review provide results within 1 hour by testing single patient samples with simpler and often automated methodology.^8-11 The results of our meta-analysis demonstrate the overall high diagnostic accuracy, high concordance, and low rates of discrepancies of rapid assays, particularly the HemosIL AcuStar CLIA, relative to reference standard assays for patients with suspected or diagnosed TTP.

There are broad clinical implications to the incorporation of highly sensitive rapid assays into diagnostic and treatment algorithms for TTP. Our approach to integrating rapid assays with clinical evaluation and risk assessment scores is outlined in Figure 6. For an acute iTTP episode, International Society on Thrombosis and Haemostasis guidelines recommend the combination of corticosteroids and TPE and suggest caplacizumab in those with a high pretest probability of TTP.^3,43 The primary benefit of rapidly ruling out TTP with a highly sensitive rapid assay is averting empiric therapy. Addition of rapid assays to the diagnostic and empiric treatment algorithms could potentially spare the adverse effects of TPE (plasma exposure, procedural complications, line-associated thrombosis, and central line infection), corticosteroids (neuropsychiatric, endocrine, and cardiac manifestations), and caplacizumab (bleeding events).⁴⁴ It is also possible that earlier rule-out of TTP may hasten evaluation of alternative etiologies of TMA. A positive rapid assay is also beneficial because it can expedite the initiation of caplacizumab in centers where caplacizumab is administered upon confirmation of severe ADAMTS13 deficiency. The greatest benefit of caplacizumab is thought to be accrued early in the acute TTP episode from uncontrolled microvascular thrombosis.^3,43 Given that the PLASMIC score gives a maximal pretest probability of <90% for severe ADAMTS13 deficiency for its high-risk group (score of 6 or 7), this pretest probability in combination with severe ADAMTS13 deficiency on a highly-specific rapid assay (HemosIL AcuStar CLIA or Technofluor FRET assay) can give further clinical certainty for the use of caplacizumab for acute TTP episodes.⁶ Earlier confirmation of ADAMTS13 deficiency can also expedite testing for an ADAMTS13 inhibitor to aid in differentiating iTTP and cTTP: the HemosIL AcuStar CLIA has been compared with the Technozym ELISA in ADAMTS13 inhibitor assays using a Bethesda technique with good comparability by Montaruli et al⁴⁵ (r = 0.88) and Favaloro et al³² (r = 0.99). The use of rapid assays for this indication may hasten the administration of rituximab for iTTP but requires further validation.^15,43 

Although our results demonstrated high sensitivity and NPV across all 3 rapid assays, in our estimation, only the HemosIL AcuStar CLIA has sufficient sensitivity (0.98; 95% CI, 0.94-1.00) and NPV (0.99; 95% CI, 0.98-1.00) to incorporate into diagnostic and treatment algorithms for this “can’t miss” diagnosis. The Technoscreen assay has a similar point estimate for its sensitivity (0.98), but given the relatively few studies comparing it with reference standard assays, its true sensitivity is uncertain and could be much lower (95% CI, 0.42-1.00). The Technofluor FRET assay has a sensitivity of 0.93 (95% CI, 0.86-0.96) and NPV of 0.96 (95% CI, 0.93-0.98), which we consider insufficient to reliably exclude TTP with the current evidence. Given that few studies have compared multiple rapid assays for the same population, further head-to-head studies are needed to determine whether there are significant differences in diagnostic accuracy among rapid assays.

Given that there are significant and potentially fatal consequences to missed or delayed diagnosis, it is notable that there were relatively few FN results. In our discrepancy analysis, some cases deemed to be acute TTP episodes were accurately identified by a rapid assay but not by a reference standard assay. In addition to averting the risks of empiric treatment, rapid assays are likely to be cost saving. Allen et al⁴⁶ conducted a cost-effectiveness evaluation of rapid vs reference standard assays for iTTP and found that the HemosIL AcuStar CLIA was cost saving for intermediate- or high-risk PLASMIC scores by reducing unnecessary TPE and caplacizumab therapy in patients without iTTP. Another study estimated a 16% reduction in costs with local measurement of ADAMTS13 activity using the HemosIL AcuStar CLIA compared with centralized measurement.⁴¹ 

Although our analysis cannot show the superiority of one rapid assay over another, the Technoscreen assay seems to have lower specificity, PPV, and concordance relative to reference standard assays than the other 2 rapid assays. This may be secondary to the semiquantitative nature of the Technoscreen assay or owing to the different patient populations in the included studies. It is possible that samples used in studies on the Technoscreen assay were more prone to interference from preanalytical variables than fresh plasma samples that would be used in real-world practice. Moore et al³⁸ reported that most discrepant samples with the Technoscreen assay were freeze thawed and that the sensitivity and NPV on fresh samples were 100%.

Although there was high concordance of rapid assays to reference standard assays, further research to assess the clinical and preanalytical conditions in which discrepancies occur would be useful. Less than half of the discrepant results between rapid and reference standard assays included annotated comorbidities among patients. A recent retrospective analysis reported that, among patients with low suspicion for TTP with discrepant, “false positive” (rapid assay positive, reference standard negative) results, a concurrent malignancy or sepsis was observed in 40% of cases.¹⁶ Other than iTTP in clinical remission and other TMAs, sepsis was the most common diagnosis noted in “false positive” cases in our descriptive analysis as well. The key yield of rapid assays is in patients with an intermediate or high pretest probability for acute TTP. Given these findings, it is unclear whether rapid assays are useful in the setting of low TTP pretest probability. Although discrepancies between rapid and reference standard assays were a small sample of cases, further real-world evidence on the comorbid conditions that increase the likelihood of discrepancies is needed.

Our study has several strengths. Our systematic review included “gray” literature, such as abstracts with sufficient data to calculate assay performance. For published references with insufficient data to evaluate rapid assays, we contacted the corresponding authors to obtain sufficient data to complete our analyses. Our meta-analysis ultimately included 4207 patient samples, a large sample for a rare disease.

There are also limitations to this analysis. First, our assessment of the performance characteristics of rapid assays presupposes that the reference standard assay result of ADAMTS13 activity <10% is diagnostic of TTP, but as our analysis of discrepant results suggests, there are cases of TTP with severe ADAMTS13 deficiency on rapid assays and not on reference standard assays. It has been suggested that a lower cutoff value for ADAMTS13 activity could be appropriate for the HemosIL AcuStar CLIA. Most comparisons (10 of 17; 58.8%) report a negative mean bias^{12,14,31-33,36,39,42} (ie, underestimated ADAMTS13 activity by HemosIL AcuStar CLIA relative to reference standard assays); however, many comparisons also showed nonsignificant differences^17,40-42 or positive mean bias.^18,34,37 Three studies reported that the negative mean bias with the HemosIL AcuStar CLIA is less significant at clinically relevant levels of ADAMTS13 activity and more apparent at higher values.^39,40,42 To the best of our knowledge, Dimopoulos et al’s study¹⁷ is the only study to assess a lower cutoff of ADAMTS13 activity in a comparison of the HemosIL AcuStar CLIA with a reference standard assay. At a cutoff of <5% for ADAMTS13 activity, 1 case of disseminated intravascular coagulation was appropriately reclassified by HemosIL AcuStar CLIA (ADAMTS13 activity: 8%) and not by a FRETS-VWF73 assay (1%), and 2 cases of TTP were appropriately classified by HemosIL AcuStar CLIA (0% and 4%) but would have been misclassified by FRETS-VWF73 assay (6% for both). Owing to a lack of disaggregated data, we could not confirm these findings. Given that sensitivity decreases at lower cutoff values and the primary utility of rapid assays is to quickly rule out TTP, we would recommend using <10% or a locally calibrated cutoff value for ADAMTS13 deficiency with the current evidence rather than a lower threshold. Our analysis is also limited by the fact that few studies compare rapid assay results with the ultimate diagnosis of TTP (incorporating clinical assessment in addition to ADAMTS13 activity), which led to our use of reference standard assays as the “gold standard” for TTP diagnosis. We conducted concordance and discrepancy analyses in recognition of the imperfect accuracy of reference standard assays for ADAMTS13 activity. Similarly, few studies reported borderline ADAMTS13 activity between 10% and 20%, which could also support a diagnosis of TTP in the appropriate clinical context.³ 

As a second limitation, studies were primarily retrospective and single center or single laboratory in design. Preanalytical variables can also affect results, including excessive hemolysis, icteric samples, and lipemia, although these are unlikely to cause significant interference.³⁸ Likewise, studies of frozen or biobanked specimens were permitted in our inclusion criteria given that most studies suggested no difference between these samples and fresh plasma samples.^17,20 Finally, our inclusion criteria allowed for studies involving patients with suspected TTP and those with confirmed TTP, including patients in clinical remission undergoing ADAMTS13 monitoring. Fewer than half of included studies reported an indication for ADAMTS13 testing, and stratification of results based on indication was not possible owing to a lack of disaggregated data. Further research should confirm diagnostic accuracy and concordance of rapid and reference standard assays in acute or suspected TTP episodes because that clinical scenario likely offers the greatest benefit for rapid diagnosis.

In conclusion, we report, to our knowledge, the first systematic review and meta-analysis on the performance characteristics of rapid assays for ADAMTS13 activity for patients with suspected or confirmed TTP. The HemosIL AcuStar CLIA demonstrated high diagnostic accuracy relative to reference standard assays, supporting its use in clinical practice and highlighting its potential to avert toxic and costly empiric treatment of patients without TTP.

Contribution: S.R.D. was involved in methodology, screening, full-text review, data extraction, statistical analysis, and manuscript writing and editing; H.T. and C.D. were involved in screening, full-text review, and data extraction; J.T., T.Z., and Y.C. were involved in statistical analysis and manuscript editing; A.C. was involved in conceptualization, methodology, screening, full-text review, data extraction, and manuscript editing; and all authors reviewed and approved the final manuscript.

Conflict-of-interest disclosure: A.C. has served as a consultant for MingSight, Pfizer, and Sanofi, and has received authorship royalties from UpToDate. The remaining authors declare no competing financial interests.

Correspondence: Saarang R. Deshpande, Division of Hematology and Oncology, Department of Medicine, Perelman School of Medicine, University of Pennsylvania, 3400 Spruce St, Philadelphia, PA 19104; email: saarang.r.deshpande@gmail.com.