In Hospitalized Patients Undergoing Therapeutic Plasma Exchange, Major Bleeding Prevalence Depends on the Bleeding Definition: An Analysis of the Recipient Epidemiology and Donor Evaluation Study-III

Abstract

Background:

Major bleeding in patients undergoing therapeutic plasma exchange (TPE) has been studied in large databases; but without standardizing bleeding definitions. Therefore, we used standardized definitions to evaluate major bleeding in hospitalized patients undergoing TPE using public use data files from the Recipient Epidemiology and Donor Evaluation Study-III (REDS-III).

Study Design and Methods:

In a retrospective cross-sectional analysis, we identified TPE-treated adults in a first inpatient encounter. We evaluated major bleeding prevalence using 1) international Classification of Diseases (ICD) or Current Procedural Terminology (CPT) codes, 2) packed red blood cell (PRBC) transfusion, or 3) hemoglobin (Hgb) decline. Patients with major bleeding prior to their first TPE were excluded from the analysis.

Results:

Among 779 patients undergoing TPE, major bleeding by at least one of the three bleeding definitions occurred in 135 patients (17.3%). For each of the ICD/CPT, PRBC, and Hgb definitions, the prevalence of major bleeding was 2.8% (n=31), 7.4% (n=81), and 5.4% (n=59), respectively. Only 3.7% of bleeds (5/135) were captured by all three definitions and 19.3% (26/135) exclusively by any two pairwise definitions. The addition of PRBC transfusion and Hgb decline to ICD/CPT code definitions increased bleeding prevalence 3-fold.

Conclusion:

Among hospitalized adults undergoing TPE in the REDS-III study, the prevalence of major bleeding was 17.3%. The addition of PRBC and Hgb decline to ICD codes increased bleeding prevalence 3-fold. Future studies are needed to develop validated models that identify patients at risk for major bleeding during TPE.

1. INTRODUCTION

Although therapeutic plasma exchange (TPE) is a safe and effective procedure, it can be associated with adverse events, including major bleeding.^1–3 Major bleeding is an important TPE safety endpoint because it is associated with worse outcomes, such as transfusion-related complications, surgical intervention, procedure discontinuation, and increased mortality.^1,4–6 In patients undergoing TPE, effective bleeding mitigation or prophylactic strategies depend on a clear understanding of the epidemiology, risk factors and extent to which major bleeding events occur. Previous studies have reported a wide variability in bleeding prevalence.^1,3,7–13 However, most of these studies were not developed specifically to assess bleeding prevalence.¹ Bleeding prevalence may also be poorly assessed due to the absence across TPE studies of standardized major bleeding definitions.^3,4,9,13

Standardized bleeding definitions have been developed by international consortiums and organizations. These organizations specify definitions for minor bleeding, clinically relevant non-major bleeding, and major bleeding.^14–17 Based on guidelines from the International Society on Thrombosis and Haemostasis (ISTH), major bleeding is defined by the presence of one or more of the following: 1) fatal bleeding, 2) symptomatic bleeding in a critical organ, or 3) bleeding causing at least a 2 g/dL fall in hemoglobin (Hgb) or leading to transfusion of two or more units of whole blood or red cells.^15,16 These definitions have been widely applied in studies investigating antiplatelet and anticoagulant therapies but have not been applied in studies evaluating major bleeding in patients undergoing TPE.¹⁷

In patients undergoing TPE, prior large database studies have estimated major bleeding prevalence using International Classification of Diseases (ICD) codes alone. To date, no large database studies have estimated major bleeding prevalence during TPE using data from packed red blood cell (PRBC) transfusion and Hgb decline.^2–4 Therefore, to better characterize major bleeding events in patients undergoing TPE, we sought to incorporate data from ICD codes, PRBC transfusion, and Hgb decline using the Recipient Epidemiology and Donor Evaluation Study-III (REDS-III) database, an observational multicenter study developed to evaluate transfusion outcomes.^15,16,18 As shown in subsequent sections of this manuscript, we determined that the addition of PRBC transfusion and Hgb decline definitions increased major bleeding prevalence over ICD codes alone.

2. MATERIALS AND METHODS

2.1. Data Source

This study is an analysis of public use data files from REDS-III, a National Heart, Lung and Blood Institute-sponsored four-year multicenter study of electronic health record data obtained from four United States (US) blood centers and 12 associated hospitals across four regions.^19,20 From July 2012 to December 2016, participating blood centers contributed data on blood donors/donations (6.5 million blood components). Participating hospitals contributed data on blood donation recipients (120,290 patients in over 234,277 encounters).^18,19,21,22 Transfusion recipients contributed data from both inpatient and outpatient encounters. For transfusion recipients, an encounter was defined as either an episode of outpatient transfusion or hospitalization during which transfusion was received.^18,19,21,22 All non-transfused patients that were hospitalized during the study period in an inpatient encounter (an unique hospitalization event; 1,285,359 encounters) were included as a comparison group.^18,19,21,22 Information provided in REDS-III public use files includes the following: demographics; medications (with time and route of administration); imaging; laboratory values; ICD-9/10 codes; and Current Procedural Terminology (CPT) codes.^18,21 Additionally, REDS-III data files contain information related to blood product processing, such as the issue time, issue location (general ward, procedure suite, emergency room, intensive care unit [ICU], operating room), ABO type, irradiation status, storage age, Codabar codes and ISBT (Information Standard for Blood and Transplant) codes.^18,20,22 All identifiers within the database were replaced by random numbers, which protects patient confidentiality.²³ Therefore, this study was reviewed by the Duke University Health System Institutional Review Board and determined to be exempt.

2.2. Study Design and Participants

The study design is a retrospective cross-sectional analysis of the REDS-III database for the study period 2012–2016. TPE was identified using validated codes: ICD 9 code 99.71 (therapeutic plasmapheresis); ICD 10 codes 6A550Z3 (pheresis of plasma, single) and 6A551Z3 (pheresis of plasma, multiple); and CPT code 36514 (therapeutic apheresis; for plasmapheresis).^2,3

The inclusion criteria were adult (≥18 years old) patients in an encounter in which a TPE procedure was done. Since pediatric patients may have a different TPE-associated bleeding risk they were excluded from this analysis. Due to the limited inclusion of outpatients in REDS-III, also excluded were outpatient encounters.^{1,18,19,21,22} Therefore, only TPE procedures performed in an inpatient encounter (unique hospitalization event) were included in this analysis. Although the REDS-III data exist at the encounter level, a patient level analysis was done by merging the recipient data files by Encounter and Subject ID. This strategy was chosen to allow for better control and adjustment of confounding variables so that patients who may be more prone to bleeding would not skew the data toward major bleeding. Therefore, in patients with two or more encounters, subsequent encounters were excluded. TPE-treated patients were further classified based on the following: 1) the presence of major bleeding (Yes-Major Bleeding vs No-Major Bleeding) and 2) time of major bleeding event relative to the TPE procedure (Prior to TPE Initiation vs. After TPE Initiation). Patients with major bleeding prior to TPE initiation were further excluded from the analyses. Finally, any outliers that indicated a data entry error were also excluded.

2.3. Study Definitions

Using modified ISTH criteria and considering the transient impact of TPE on hemostasis, major bleeding was defined as follows: 1) the presence of an ICD 9/10 or CPT code for major bleeding coded after the first TPE procedure; or 2) ≥ 2 PRBC transfusion within 24 hours of one another and within 2 days of any TPE procedure; or 3) a Hgb decline of ≥ 2g/dl in any consecutive measurements after the first TPE and within 2 days of the last TPE procedure.^1,16 Only major bleeding events occurring exclusively after TPE initiation were considered possibly TPE-associated events. The same criteria were used to identify patients with major bleeding events prior to TPE initiation.

Major bleeding ICD/CPT codes were defined using validated codes in the literature and as used in our prior studies (see Supplementary Appendix).^3,15,24–27 PRBC transfusion was identified using ISBT codes for issued PRBC products.^18,22 In REDS-III, the blood product was considered to be transfused only if it was not returned to the blood bank.²⁰ Additionally, REDS-III also allows the longitudinal analysis of the database by using the variable, “DaysSinceStartEncounter,” which was used to determine the Hgb decline in consecutive measurements.¹⁸ Although the TPE procedure itself can cause a decline in the hematocrit and or Hgb, these changes are usually small (0% to 6%) and not likely to be responsible for declines ≥ 2g/dl in consecutive measurements.^1,6,28–30

We also evaluated the impact of variables known to be associated with bleeding events that were present prior to TPE initiation (baseline risk factors). These risk factors include thrombocytopenia, anemia, coagulopathy, and hypofibrinogenemia (see Table 1 for definitions).¹

Table 1.

Study Definitions

Number of TPE procedures	Based on the number of ICD and CPT codes; If both ICD and CPT codes were reported in the same day, only one was counted.
Time of Major Bleeding Relative to the TPE Procedure	Determined based on the variable “DaysSinceStartEncounter” which allows the longitudinal analysis of the database
Invasive Procedures	Patients were identified if they were transferred FROM or TO the Procedure Suite; We also included patients identified based on ICD9/10 codes related to renal, cardiac and muscle biopsy.
Intensive Care Unit Stay	Patients were identified if they were transferred FROM or TO the Intensive Care Unit
Operating Room/Surgery	Patients were identified if they were transferred FROM or TO the Operating Room
100% Plasma-TPE	Defined as a patient who received plasma for every day there was a TPE procedure AND did not have albumin for any of their TPE
Antiplatelet therapy	Defined as utilization of aspirin, clopidogrel, prasugrel or ticagrelor on at least one day of a TPE procedure
Therapeutic anticoagulation	Defined as 1) unfractionated heparin administered intravenous; 2) Enoxaparin >40mg or 0.4ml; 3) Dalteparin > 5000 units/day; 4) argatroban or bivalirudin utilization regardless of dosage; 5) Apixaban ≥5mg/day; 6) Dabigatran=150mg/day; 7) Rivaroxaban ≥15mg/day; 8) Betrixaban ≥80mg/day; 7) warfarin or edoxaban utilization regardless of dosage
Baseline Risk Factors Prior to TPE
Anemia	Defined as the Hgb measurement <13g/dL for men or <12g/dL for women closest o the first TPE procedure (but not more than 30 days)
Hypofibrinogenemia	Defined as fibrinogen level <200mg/dl or 0.2g/L within 2 days before the first TPE procedure
Coagulopathy	Defined as activated partial thromboplastin time or International Normalized Ratio above the upper limit of normal within 2 days before the first TPE procedure
Thrombocytopenia	Defined as platelet counts <150×109/L within 2 days before the first TPE
Hemolytic conditions	Patients were identified based on ICD9/10 codes of both acquired and hereditary hemolytic conditions

CPT = Current Procedural Terminology; Hgb = Hemoglobin; ICD = International Classification of Diseases; TPE = Therapeutic Plasma Exchange.

Additional TPE parameters, such as TPE schedule (frequency of TPE procedures) and volume exchanged, were not available or could not be accurately determined. Therefore, they were not included. Detailed information related to ICD/CPT codes used in this analysis are provided in the Supplementary Table 1.

2.4. Statistical Analysis

The statistical analysis was a patient-level analysis using mean, median, interquartile, range, and standard deviation (SD) or counts and percentages as appropriate. Tests of associations between bleeding status and baseline characteristics were examined by Chi-Square for categorical variables and Kruskal-Wallis tests for continuous variables. SAS version 9.4 (SAS Institute Inc., Cary, NC) was used for the analysis. A p-value of <0.05 was considered statistically significant. Univariate results are presented without, and with adjustment for multiple comparison using Bonferroni correction. Even though only 56 are shown, the Bonferroni adjusted p-values are adjusted for 97 comparisons.

To better characterize TPE-treated patients with major bleeding identified by the PRBC transfusion and Hgb decline definition, we reported the mean, SD, and range for Hgb decline (for patients with multiple instances of a Hgb decline ≥ 2 g/dL, we used the maximum decline) and the maximum number of PRBC transfusions received within 24h. This was also used as a data cleaning step in our analysis to identify potential outliers and data entry errors.

2.5. Sensitivity Analysis

To exclude the impact of hemolysis on bleeding defined by PRBC transfusion or Hgb decline definitions, we performed a sensitivity analysis excluding patients with ICD codes for hemolysis coded prior to or on the same day as the first TPE procedure, including thrombotic microangiopathy (TMA), hemolytic uremic syndrome, autoimmune hemolytic anemia, and sickle cell disease (a full list of excluded hemolysis codes is in Supplementary Table 1). Because prior studies have demonstrated that the overall specificity of major bleeding events identified by ICD codes is 99%,³¹ the sensitivity analysis did not exclude patients with major bleeding events identified by ICD codes.

3. RESULTS

Between 2012 and 2016, a TPE procedure was coded for 1,154 patients (1,600 encounters). After exclusion of pediatric patients (n=43) and patients receiving TPE in an outpatient encounter (n=22), the final number of patients receiving TPE in a first inpatient encounter was 1,089 (250 subsequent encounters excluded). Among these patients, 310 (28.5%) had a major bleeding event prior to TPE initiation and were excluded. Therefore, the final study population was 779 patients, of which 135 (135/779; 17.3%) had a major bleeding event that occurred exclusively following TPE initiation (see Figure 1).

Figure 1. Study flow diagram.

There were a total of 1,154 patients (in 1,600 encounters) treated with therapeutic plasma exchange (TPE). After excluding pediatric patients (43 patients and 52 encounters) and outpatient encounters (22 patients in 209 encounters), the number of adult patients in an inpatient encounter in which a TPE procedure was done was 1,089. Among those, 310 (28.5%) experienced major bleeding events prior to the first TPE procedure and were excluded. Therefore, the final study population for analysis included 779 patients. Of these, major bleeding events exclusively following TPE initiation occurred in 135 (17.3%).

3.1. Patient Characteristics By Bleeding Status

Patient demographic information separated by bleeding status are shown in Table 2. Using a threshold of α = 0.05 and considering the Bonferroni adjusted p-values, no statistically significant differences were found for demographic characteristics (gender, age, race, and ethnicity).

Table 2.

Baseline Characteristics of Therapeutic Plasma Exchange-treated patients by Bleeding Status

	Major Bleeding Exclusively Following TPE Initiation (N=135)	No-Major Bleeding (N=644)	Total (N=779)	p value	Bonferroni adjusted p value
Age				0.42681	>0.9999
Mean (SD)	49.4 (17.3)	50.7 (17.2)	50.5 (17.2)
Median	48.0	51.0	50.0
Range	(18.0–86.0)	(18.0–90.0)	(18.0–90.0)
Interquartile Range (Q1-Q3)	35.0–63.0	37.0–65.0	36.0–64.0
Gender				0.01222	>0.9999
Female	87 (64.4%)	339 (52.6%)	426 (54.7%)
Male	48 (35.6%)	305 (47.4%)	353 (45.3%)
Race				0.82532	>0.9999
White	84 (64.1%)	424 (68.2%)	508 (67.5%)
Black or African American	31 (23.7%)	133 (21.4%)	164 (21.8%)
Asian	3 (2.3%)	14 (2.3%)	17 (2.3%)
Other	13 (9.9%)	51 (8.2%)	64 (8.5%)
Not Reported	4	22	26
Ethnicity				0.86922	>0.9999
Hispanic	12 (8.9%)	64 (9.9%)	76 (9.8%)
Non-Hispanic	119 (88.1%)	557 (86.5%)	676 (86.8%)
Not specified	4 (3.0%)	23 (3.6%)	27 (3.5%)
Body Mass Index *				0.37381	>0.9999
Mean (SD)	27.8 (7.5)	28.6 (7.0)	28.4 (7.1)
Median	26.1	27.3	27.0
Range	(3.0–50.0)	(13.9–59.1)	(3.0–59.1)
Interquartile Range (Q1-Q3)	23.4–32.3	23.5–33.5	23.5–33.0

^*= total of 410 patients due to missing data (83 with bleeding complications and 327 without bleeding complications)

¹= Kruskal Wallis

²= Chi-Square

Additionally, patients with major bleeding exclusively after TPE were more likely associated with 100% Plasma-TPE (26.7% vs. 10.4%; p= 0.0097) and an intensive care unit (ICU) stay (59.3% vs. 25.2%; p = 0.0097) (see Supplementary Table 2 for detailed information).

Regarding American Society for Apheresis (ASFA) TPE-indications,^32,33 patients with major bleeding were more likely associated with multiple myeloma (5.9% vs. 1.1%; p=0.0194), nephrogenic systemic fibrosis (5.9% vs. 1.1%; p=0.0194), and sepsis (5.9% vs. 1.2%; p=0.0485). Other common ASFA TPE-indications are summarized in Supplementary Table 3.

3.2. Prevalence of possibly TPE-associated major bleeding events

The results by major bleeding definition are summarized in Figure 2. Overall, the addition of PRBC transfusion and Hgb decline to the ICD/CPT definition increased bleeding prevalence 3-fold (identified 104 more patients).

Figure 2. Numbers of TPE-treated patients with major bleeding exclusively following TPE initiation.

Venn diagram of TPE-treated patients with major bleeding based on each major bleeding definition.

Among the 81 patients with major bleeding identified by the PRBC definition, 2 units of PRBC transfusion were given within 24h for 53 patients (65.4%). For the remaining patients, 8 (9.9%) received 3 units each, 6 (7.4%) received 4 units each, 3 (3.7%) received 5 units each, 2 (2.5%) received 6 units each, 5 (6.2%) received 8 units each, 1 (1.2%) received 9 units and 3 (3.7%) received 10 units each (See Supplementary Figure 1). Additionally, in our Hgb decline definition, we initially identified 60 patients with major bleeding events, including one outlier that was further excluded from the Hgb cohort (a Hgb drop of 28g/dL likely represented a data entry error. Although excluded from the Hgb cohort, this patient was also identified by the PRBC definition and was included in the PRBC cohort). Therefore, the final Hgb cohort had 59 patients. Among these, the mean Hgb decline was 2.749 g/dL (SD ± 0.809g/dL; range 2–5.6g/dL).

3.3. Sensitivity Analysis – Prevalence of possibly TPE-associated major bleeding events

The sensitivity analysis excluded 23 patients with major bleeding due to an ICD code for hemolysis. Therefore, in this analysis, the prevalence of major bleeding was 112 (112/756, 14.8%) (see Figure 3 for details). In the sensitivity analysis, adding the PRBC transfusion and Hgb decline to the ICD/CPT definition increased bleeding prevalence 3-fold (identified 81 more patients).

Figure 3. Numbers of TPE-treated patients with major bleeding exclusively following TPE initiation in the sensitivity analysis.

Venn diagram of TPE-treated patients with major bleeding based on each major bleeding definition.

4. DISCUSSION

In a cross-sectional analysis of 779 TPE-treated patients in the REDS-III database, we found that: 1) The prevalence of major bleeding occurring after initiating the TPE procedure was 17.3%; and 2) The addition of Hgb decline and PRBC transfusion to ICD increased the identification of major bleeding in TPE-treated patients by 3-fold.

Our finding of a 17.3% prevalence of major bleeding events exclusively following TPE initiation was within the broad range of other studies published in the literature showing bleeding prevalence that range from 0.6% to 55.5%.^6,7,11,34,35 On the higher end of the spectrum of reported bleeding events is a 55.5% bleeding prevalence reported in study of 18 (mostly kidney transplant) patients with TPE-associated acquired factor XIII deficiency.⁶ In this study, bleeding occurred in 10/18 patients (55.5%). In applying ISTH criteria to this study, major bleeding occurred in 9/18 (50%) patients and minor bleeding in 1/18 (5.5%). On the lower end of the spectrum is a 0.6% bleeding prevalence reported in a study of 509 patients receiving TPE for mostly neurologic conditions (e.g., myasthenia gravis and Guillain-Barre Syndrome). In this study, bleeding events occurred in three patients, including two patients with prolonged bleeding after needle removal and one patient with transient nasal bleeding.¹³ Thus, in patients undergoing TPE, there is heterogeneity in reported bleeding outcomes. Furthermore, the lack of standardization of reported bleeding events hinders comparison between studies. Future studies are needed to standardize definitions of TPE associated bleeding events, including major, minor, and clinically-relevant non-major bleeding, so that bleeding events can be appropriately compared across studies.

Heterogeneity in reported TPE-associated bleeding events may be impacted when studies report bleeding prevalence differently. Bleeding prevalence may be reported based on the number of TPE procedures or on a per patient basis.^10,35–38 Because the number of TPE procedures is typically higher than number of patients treated, reporting bleeding prevalence per procedure may cause TPE-associated bleeding events to be underestimated. In fact, when we consider studies reporting bleeding as a percentage of the number of TPE procedures, reported bleeding events range from 0.1% to 8.1%.^10,35–38 In a study of 174 patients receiving 1,727 TPE procedures, bleeding events occurred in 23 TPE procedures (1.3%).³⁶ Similarly, in another study of 1,034 patients undergoing 9,501 TPE procedures, bleeding events were reported in 13 procedures (0.1%).³⁸ In these studies, the number of patients experiencing bleeding events was not reported. For this reason, studies that report bleeding events on a per-procedure basis may artificially lower TPE-associated bleeding estimates when compared to studies reporting events on a per patient basis. On the other hand, reporting bleeding events on a per-procedure basis may provide greater accuracy in assessing dynamic changes that occur during hospitalization, such as the temporal relationship of the TPE procedure with invasive procedures, transient thrombocytopenia and/or coagulopathy.¹ Therefore, standardizing the reporting of TPE-associated bleeding events using one or both strategies would decrease heterogeneity and improve prevalence estimates in future studies.

Our understanding of the true prevalence of TPE-associated bleeding event may also be influenced by strategies used to identify bleeding in administrative databases.^{3,4,7–13,39} In this study, the use of ICD codes identified only 2.8% of patients, thus underestimating bleeding prevalence compared to other bleeding definitions. In our prior large database analysis of 3,991 TPE-treated cases that used ICD codes to identify major bleeding, the prevalence was 5.4%.³ Similarly, another large database study that used ICD codes alone to identify bleeding events in patients with ITP, some of which received TPE, found lower bleeding prevalence than expected for patients with ITP.⁴ Lower bleeding prevalence has also been found in other large database studies using only ICD codes.^2,31,40 In light of these data, the present study suggests that the prevalence of possibly TPE-associated major bleeding may be higher than previously reported in the literature. Nevertheless, a key challenge in studies of major bleeding is the need to achieve an appropriate balance between specificity and sensitivity.¹⁷ Our sensitivity analysis excluding patients with hemolysis did not demonstrate major changes in the prevalence of possibly TPE-associated major bleeding. Nevertheless, future studies are needed to determine the specificity and sensitivity of each major bleeding definition in TPE-treated patients.

To our knowledge, this study is the first large analysis designed to evaluate the prevalence of major bleeding events in TPE-treated adult inpatients using a definition that incorporates ICD codes, PRBC transfusion, and Hgb decline.^15,16 Strengths of our study include the innovative use of a large electronic health record database with transfusion outcomes, which allowed us to identify major bleeding using different strategies. However, as with any secondary analysis of electronic health record data, we note important limitations. First, the key to our study lies in the definition of major bleeding events occurring before and after TPE, which depends on the dates when these procedures were coded. TPE procedures that were coded on a different date than the date of the actual procedure would skew the direction of the identified major bleeding events. Additionally, our analysis may have picked up bleeding episodes as possibly TPE-associated in patients with conditions that are known to cause bleeding prior to TPE. Finally, since the transition to ICD 10, the development of a single code to capture multiple TPE procedures (6A551Z3) limits accurate determination of the numbers of TPE procedures among patients in whom this code was used. Notwithstanding these limitations, our study is an important milestone in the understanding of major bleeding prevalence in TPE-treated patients and provides data to guide future studies to assess main risk factors and strategies to improve major bleeding outcomes.

5. CONCLUSION

Among patients in the REDS-III database undergoing TPE, the prevalence of major bleeding events exclusively following TPE initiation as identified by ICD codes, PRBC transfusion, and Hgb decline was 17.3%. The addition of PRBC transfusion and Hgb decline to ICD codes increased bleeding prevalence 3-fold. It is important to note that these findings only show association of bleeding events following TPE and does not, in any way, establish causality. Future studies are needed to develop validated and accurate models to identify patients at risk for major bleeding following TPE.

DATA AVAILABIITY STATEMENT

REDS-III public use datasets are available through the National Institute of Health (NIH) Biologic Specimen and Data Repository Information Coordinating Center (BIOLINCC).²²