Is confirmatory testing still necessary to diagnose von Willebrand disease?

TO THE EDITOR:

In 2005, the International Society on Thrombosis and Haemostasis (ISTH) published recommendations regarding the diagnosis of von Willebrand disease (VWD).¹ Criteria for the diagnosis of VWD included clinical features, family history, and laboratory testing. To meet laboratory criteria, patients had to have abnormal results “on >2 determinations.”¹ This recommendation has informed the diagnostic algorithm used at our institution since then. In 2008, the National Heart, Lung, and Blood Institute (NHLBI) published formal guidance for the diagnosis and management of VWD.² This guideline also recommended confirmatory laboratory testing, albeit with qualifiers such as “if indicated” or “in most cases.” The most recent American Society of Hematology/ISTH/National Hemophilia Foundation/World Federation of Hemophilia guideline (published in Blood Advances in 2021) does not mention the need for confirmatory laboratory testing.³ In addition, a recent review of VWD published in the New England Journal of Medicine and publications from 2 prominent children’s hospitals also do not indicate that confirmatory testing is necessary.^4-6 This begs the question, Is confirmatory testing still necessary to diagnose VWD? The most recent UK guideline suggests that it is.⁷ However, the UK guideline, the original ISTH recommendation, and the 2008 NHLBI guideline fail to reference any studies in support of the necessity for confirmatory testing.^1,2,7 

With that in mind, we reviewed the diagnostic record of all patients who underwent von Willebrand (VW) testing at our tertiary children’s hospital located in the United States. This investigation went from the date of first use of an electronic health record (EHR) at our institution in April 2010 until the day the institutional review board granted authorization as exempt research in June 2021.⁸ Our aim was to determine the probability of being diagnosed with VWD if the first panel had abnormal results. For ease of reading, we define this as the posttest probability. Before collecting data, we concluded that if the posttest probability of being diagnosed with VWD was >95%, confirmatory testing was no longer necessary. We also collected demographics, family history of VWD, ISTH-Scientific and Standardization Committee–endorsed Bleeding Assessment Tool Bleeding Score (ISTH-BAT), and other laboratory data related to bleeding. We used information available in the EHR to retrospectively determine an ISTH-BAT as per Doshi et al.⁵ 

The initial EHR database query was for any patient between birth and 22 years of age from whom at least 1 of the following laboratory tests had a result: VW antigen, ristocetin cofactor/activity (RCO), GPIbM, or collagen binding. Patients eligible for further analysis included those who had a disposition regarding VWD by a pediatric hematologist. Diagnostic dispositions were as per the 2008 NHLBI guideline and based on the presence of recurrent abnormal bleeding, and/or a family history, and repeatedly abnormal laboratory results. Patients diagnosed before 2010, those diagnosed at another institution, those with an incomplete evaluation (disposition unable to be determined), those with testing other than for abnormal bleeding (thrombotic risk), and those never seen by a pediatric hematologist were not eligible. Before 2018, the RCO was the predominant functional assay used in our diagnostic algorithm. In 2018, the GPIbM and collagen binding assays became more prominent. Within this region, most patients’ EHR both within and outside our hospital system are linked. This afforded us the opportunity for extended follow-up to determine if the diagnosis changed. VW panels included RCO-based (RCO, and/or factor VIII, and/or VW antigen) or GPIbM-based (GPIbM, and/or factor VIII, and/or VW antigen, and/or collagen binding) assays. VW panels were considered abnormal if at least 1 result other than factor VIII activity was <50 IU/dL. The EHR of patients diagnosed with VWD or whose first VW panel was abnormal was reviewed a second time for data collection errors. This study was approved by institutional review board as exempted research. This study was conducted in accordance with the Declaration of Helsinki.

As shown in Figure 1, 1133 patients had VW testing. Of these, 577 were ineligible, mostly because they were not seen by a hematologist, mostly because they had normal results. Of the 556 eligible patients, 74 were diagnosed with VWD. Of these, 71 (96%) completed confirmatory testing. Seventy-two of the initial VW panels ordered by a pediatric hematologist were abnormal. Of these, 56 patients were eventually diagnosed with VWD. Thus, the posttest probability of being diagnosed with VWD, based on an initial abnormal result, was 78%. Seventeen patients diagnosed with VWD had initial normal results. One patient was given a diagnosis based solely on abnormal results from a referring physician. One-hundred fifty-six patients had VWD testing ordered by the referring physician, none of which included a GPIbM panel. Forty-eight were diagnosed with VWD. The posttest probability of being diagnosed with VWD based on abnormal results obtained by referring physicians was 63%. The GPIbM panel had superior diagnostic accuracy compared with RCO panel, with a posttest probability of diagnosis of 93% vs 65%, P = .003. There were 17 patients with initial panels ordered by a pediatric hematologist and VW assay result(s) <30 IU/dL. Of these, 14 (82%) were subsequently diagnosed with VWD (4/4 GPIbM and 10/13 RCO). Only 1 out of 253 patients with initial VW assay result(s) >99 IU/dL was diagnosed with VWD.

Table 1 shows demographic, clinical, and laboratory data of those diagnosed with and without VWD. Expectedly, those with VWD were more likely to have a family history of VWD, have a higher ISTH-BAT, and report an abnormal ISTH-BAT for their age.⁹ A multivariate logistic regression model was created to predict VWD with predictor variables of ISTH-BAT, family history, and ISTH-BAT for age. Logistic regression showed the odds of having VWD increased by 1.5 times for each unit increase of ISTH-BAT (P = .0009), by 2.4 times if the ISTH-BAT was abnormal for age (P = .03), and by 3.2 times if there was a family history (P = .002). Using our diagnostic algorithm, nearly all patients (96%) were given a disposition within 3 months and 3 panels or less. Only 4 patients initially diagnosed with VWD had a diagnostic change (bleeding symptoms resolved and laboratory results normalized), and none of the patients who were diagnosed as not having VWD, were subsequently diagnosed with VWD. The mean follow-up was nearly 4 years with a range of 0 to 11.

Based on our results, we believe pediatric patients suspected of having a bleeding disorder and initial panel results <100 need to continue to have confirmatory testing to establish or exclude VWD. Our prestudy cutoff of a posttest probability of >95% is subjective, but we believe most physicians would find this reasonable. In our hands, the GPIbM-based panel gave superior results compared with RCO-based panel, and nearly exceeded our diagnostic threshold. If available, we would advise using the GPIbM assay for the initial diagnostic algorithm. The lowest accuracy was seen from RCO-based panels obtained by referring physicians. It is well known that VWD testing outside of tertiary centers can give faulty results.¹⁰ Our data are consistent with this. Previous studies have shown a high negative predictive value for VW assay results >99 IU/dL.^5,11 Our results also support this finding.

This analysis suffers from weaknesses common to single-institution retrospective studies and would be enhanced by a multi-institution study. However, the ability to monitor diagnostic changes for up to 11 years provides an added measure of confidence in our diagnostic algorithm. The results of our analysis may not be applicable to all patients in all tertiary centers but point toward the ongoing need for confirmatory testing when evaluating patients for VWD.

Acknowledgments: The authors thank SSM Health, and especially Howard Williams, for the query of the electronic health record database.

This project did not receive any financial support.

Contribution: J.P. conceived of the project, collected and analyzed data, and wrote and edited the manuscript; and K.N.K. and Z.Z. analyzed data, performed statistical analysis, and edited the manuscript.

Conflict-of-interest disclosure: The authors declare no competing financial interests.

Correspondence: John Puetz, SSM Health Cardinal Glennon Children’s Hospital, 1465 S Grand, St. Louis, MO 63104; email: [email protected].