Autoimmune Hemolytic Anemia Confers an Independent Risk Factor for Thrombosis: Retrospective Cohort Study Using the “STRIDE” Database

Research Article

Ann Hematol Oncol. 2017; 4(7): 1160.

Autoimmune Hemolytic Anemia Confers an Independent Risk Factor for Thrombosis: Retrospective Cohort Study Using the “STRIDE” Database

Chen EC¹, Loftus PD¹, Weber SC¹, Hoang NM¹, Gilbert J², Rosenthal A² and Kummar S¹*

¹Stanford School of Medicine, Stanford University, USA

²True North Therapeutics, USA

*Corresponding author: Shivaani Kummar, Stanford University School of Medicine, 780 Welch Road, CJ250L, Palo Alto, CA 94305, USA

Received: May 07, 2017; Accepted: June 12, 2017; Published: July 04, 2017


Background: Auto Immune Hemolytic Anemia (AIHA) is a rare autoimmune disorder in which auto antibodies cause haemolysis. While thrombosis is considered a complication of AIHA, whether AIHA confers an independent thrombosis risk has not been thoroughly shown. The relationship between haemolysis and thrombosis is also unclear.

Methods: 312 non-AIHA patients from Stanford University’s clinical database were matched 2:1 with 156 AIHA patients. Thrombosis incidence was measured, and the prevalence of other thrombosis risk factors as defined by the PADUA score, prior splenectomy status, antiphospholipid antibody diagnosis, and concomitant anticoagulation use was also compared. Within AIHA patients, the relationship between thrombosis and haemolysis was analyzed in terms of time and severity.

Results: 29% of AIHA patients developed thrombosis compared to 19% of non-AIHA patients (p<0.05). AIHA conferred an odds ratio of 2.44 (95% CI [1.16-5.10], p<0.05) for thrombosis. The median PADUA score was not different between the two groups (4 for AIHA, IQR [3-7] vs. 4.5 for non-AIHA, IQR [3- 7], n.s.). There was also no difference in prior splenectomy, antiphospholipid antibody status, and concomitant anticoagulation use. AIHA patients with thrombosis had more hemolytic flares than patients without thrombosis (24.5 instances vs. 13.8, respectively; p<0.05) and a higher lifetime drop in hemoglobin (53.4 g/dL vs. 27.1 g/dL, respectively; p<0.05). 81% of patients with thrombosis had the event within one week of a hemolytic flare.

Conclusion: AIHA is an independent risk factor for thrombosis. Thrombosis is associated with periods of haemolysis and a high hemolytic burden.

Keywords: Anemia; Autoimmune; Hemolytic; Thrombosis; Thromboembolism


Auto Immune Hemolytic Anemia (AIHA) is a rare autoimmune disorder in which auto antibodies target red blood cell surface antigens, causing haemolysis. The estimated incidence is 0.8 per 100,000 with a prevalence of 17 per 100,000 [1,2]. Depending on the thermal range at which the auto antibodies are most active, AIHA is classified as warm (wAIHA), cold (cold agglutinin disease, or CAD), or mixed. 50-60% of AIHA cases are believed to be secondary [3-5]. Major risk factors for secondary AIHA include malignancy [6], viral and mycoplasma infections [7,8], and rheumatologic disorders [9]. AIHA treatment remains largely based on expert opinion without a clear consensus due to limited prospective data [10,11]. Treatment is focused on immune suppression, cytotoxic agents, splenectomy, and addressing contributing factors [12].

While thromboembolism is widely considered to be a complication of AIHA, few studies have assessed whether AIHA confers an independent thrombosis risk that cannot be attributed to known thrombosis risk factors such as immobilization, malignancy, and rheumatologic disorders. Doing so requires the comparison of thrombosis rates in matched AIHA and non-AIHA patients, where other thrombosis risk factors can be taken into account. The literature on this topic is summarized by a 2015 meta-analysis by Ungprasert, et al. [13] based on three retrospective cohort studies and one cross-sectional study. This meta-analysis arrived at an overall pooled risk ratio of 2.63 (95% CI [1.37-5.05]) for the development of thromboembolism in AIHA [13]. However, the studies included the meta-analysis were limited in the extent to which they matched their AIHA and non-AIHA cohorts. Yusuf, et al. matched their non-AIHA cohort according to age and gender alone [14], while the other three studies required their non-AIHA cohort to only lack a diagnosis of AIHA [15-17].

Within AIHA patients, the relationship between thrombosis and haemolysis is also not well understood. Haemolysis is generally thought to cause thromboembolism due to the abnormal exposure of phosphatidylserine following red cell destruction, which promotes coagulation [18]. In support, work by Barcellini, et al. [11] and Lecouffe-Desprets, et al. have suggested that thrombosis in AIHA is associated with more severe anemia at the time of AIHA onset and at the time of the thrombotic event [19,20]. However, these studies excluded patients with secondary causes of AIHAi. e., lymphoproliferative disorders, infections, and autoimmune disorders-which comprise the majority of AIHA cases as mentioned above.

Thus, our objectives were (1) to ascertain the risk of thrombosis intrinsic to AIHA that cannot be attributed to other major thrombosis risk factors, and (2) to better characterize the relationship between haemolysis and thrombosis within AIHA. The strength of this study is the use of a longitudinal, retrospective cohort design with propensity score matching between AIHA and non-AIHA patients.


Study population

STRIDE (Stanford Translational Research Integrated Database Environment) is a standards-based informatics platform supporting clinical and translational research at Stanford University [21]. The STRIDE database was initiated in 2005 and houses records for over 2.6 million patients who have received care at Stanford University Medical Center since 1995. Our study cohort was derived from all available records in STRIDE database from 1995 until November 2015. Data was obtained following all applicable institutional and ethics approvals.

The inclusion criteria were as follows: (1) age =18, (2) at least one instance of hemoglobin (hgb) =12 g/dL, (3) at least one instance of haptoglobin <8 mg/dL, (4) diagnosis of AIHA as defined by a positive direct antiglobulin test (DAT), and (5) documentation of the ICD- 9 code for AIHA (283.0). Unlike prior studies, patients were not excluded based on secondary causes of AIHA such as hematologic, neoplastic, or infectious processes. A total of 156 AIHA patients met the above criteria in the STRIDE database and were classified into AIHA subtypes: wAIHA (DAT+ for IgG without evidence of clinically significant cold agglutinins), CAD (DAT+ for C3 only with high thermal amplitude cold agglutinins) [22], and mixed (DAT+ for both IgG and C3 with high thermal amplitude cold agglutinins). DAT positivity for C3 but without follow-up thermal amplitude tests were deemed unclassifiable.

To assess the risk of thromboembolism conferred by AIHA, an age and gender-matched control cohort of non-AIHA patients was derived for comparison. 312 non-AIHA patients were matched with the 156 AIHA patients in a 2:1 ratio. Patients in the non-AIHA group were prioritized to match according to the major risk factors for AIHA: malignancy (where the severity of active disease was estimated by the number of cancer center visits), viral and mycoplasma infections, and rheumatologic diseases. To prioritize matching according to AIHA risk factors, 6 non-AIHA patients were matched by removing the gender constraint, and 8 were matched by removing both the gender and age constraints.

Within the AIHA and non-AIHA cohort, the association between thrombosis and common thrombosis risk factors, based on the PADUA criteria, was determined. The PADUA criteria were originally developed as a tool for assessing the risk factors for venous thromboemboli (VTEs) [23]. These factors include active malignancy, previous thromboembolism, reduced mobility, existing thrombophilic condition (e.g. Factor V Leiden, my eloproliferative disorders), recent trauma or surgery (=1 month), elderly age (=70 years), cardiovascular and/or respiratory failure, myocardial infarctions and/or strokes, obesity (BMI =30), rheumatologic disorder and/or infection, and hormonal treatment (e.g. hormone replacement therapy, oral contraceptives). The PADUA risk factors comprise a weighted sum known as the PADUA score (max of 20), where scores =4 are considered high-risk for VTEs.

The PADUA score omits two additional major thrombotic risk factors, antiphospholipid antibodies and history of splenectomy, and these were considered separately. Patients were considered to be antiphospholipid antibody positive if they were documented as such in clinical notes. Antiphospholipid positivity was not based on lab results since diagnosis requires positive test results 12 weeks apart, and the STRIDE database lacks repeat lab results for work-up initiated during a patient’s presentation for AIHA. For splenectomies, only splenectomies performed >1 month from a thrombotic event was considered, since thrombotic complications soon after splenectomy are more likely due to local surgical factors [24], which are already captured by the PADUA criteria (i.e., “recent trauma or surgery”).

Patients were considered to have been on anticoagulation at the time of their thrombotic event if there was (1) documented use of subcutaneous heparin or enoxaparin within 3 days of their documented date of thrombosis, or (2) documented use of oral agents (e.g. warfarin, rivaroxaban, apixaban) any time prior to thrombosis, since oral agents are typically warranted by long-term indications and were presumed to have been continued through the time of thrombosis unless noted otherwise.

To assess the relationship between thrombosis and haemolysis within the AIHA group, haemolysis laboratory values of patients with and without thrombosis were compared. These labs included hemoglobin (hgb), lactate dehydrogenase (LDH), haptoglobin, and total bilirubin (TBili) levels. Indirect bilirubin levels, though preferred over TBili levels for haemolysis, were not documented in STRIDE for many patients and therefore could not be used.

Statistical analysis

The incidence of arterial and venous thromboembolism in the AIHA and non-AIHA cohorts was compared using the Student’s t-Test (for brevity, “thrombosis” is used in this article interchangeably with “thromboembolism”).

In our study, the median PADUA score between AIHA and non-AIHA patients with thrombosis was compared using the Mann-Whitney rank sum test. Because the PADUA score requires calculation at the time of a sentinel event, for patients with thrombosis the score was calculated at the time of thrombosis, or, for AIHA-patients without thrombosis, calculated at the time of AIHA diagnosis. PADUA scores were unable to be calculated for non-AIHA patients without thrombosis.

In seeking to derive an odds ratio of thrombosis given a diagnosis of AIHA, propensity score-matching was used to ensure stringent matching. The propensity score was constructed using patient characteristics (i.e. age, gender, AIHA risk factors, and PADUA risk factors) for each patient, and indicated their likelihood of having AIHA. Inverse-probability weighting was applied on a final model where thrombosis-status was the outcome and AIHA-status was the primary predictor [25]. Using this full model, an odds ratio was derived for the development of thrombosis given a diagnosis of AIHA. To ensure that the model was not over-fitted and the relationship between AIHA and thrombosis was not over-adjusted, a sensitivity analysis was conducted using a sparse model in which only AIHA and covariates significant at p<0.05 were included.

In the event of haemolysis, hgb and haptoglobina are expected to be low with associated high levels of LDH and TBili. Stanford laboratories cutoffs of LDH >340 IU/L and haptoglobin <8 mg/ dL were used to indicate clinically significant positive and negative results, respectively. Cutoffs of hgb <8 g/dL and TBili >3.0 mg/ dL were chosen according to the Common Terminology Criteria for Adverse Events (CTCAE) for grade 3 (i.e., “severe or medically significant”) anemia and hyperbilirubinemia.

The number of haemolysis flares per patient, defined as any period of continuous hgb decline =2 g/dL week, and patients’ total lifetime drop in hgb (the “hemolytic burden”) were also assessed. Lastly, whether thrombotic events occurred closely in time to hemolytic flares was determined. Differences were compared using the Student’s t-Test or two-proportion t-Test.


AIHA and non-AIHA cohort characteristics

Demographic and clinical characteristics of patients in the AIHA and non-AIHA cohorts are shown in Table 1. Of the AIHA patients that were classifiable into AIHA subtypes, the majority were diagnosed with wAIHA (44%), followed by CAD (19%), mixed (6%), and unclassifiable (31%). A significantly higher proportion of AIHA patients developed thrombosis than non-AIHA patients (29% vs. 19%, respectively; p<0.05).