|Home | About | Journals | Submit | Contact Us | Français|
Objective. To investigate the association between the presence of aPL and/or LA and all-cause mortality among end-stage renal disease (ESRD) patients with and without SLE.
Methods. We included ESRD patients >18 years old followed at an urban tertiary care centre between 1 January 2006 and 31 January 2014 who had aPL measured at least once after initiating haemodialysis. All SLE patients met ACR/SLICC criteria. APL/LA+ was defined as aCL IgG or IgM >40 IU, anti-β2glycoprotein1 IgG or IgM >40 IU or LA+. Deaths as at 31 January 2014 were captured in the linked National Death Index data. Time to death was defined from the first aPL measurement.
Results. We included 34 SLE ESRD and 64 non-SLE ESRD patients; 30 patients died during the study period. SLE ESRD patients were younger [40.4 (12.5) vs 51.9 (18.1) years, P = 0.001] and more were women (88.2% vs 54.7%, P < 0.001) vs non-SLE ESRD patients. The frequency of aPL/LA+ was 24% in SLE and 13% in non-SLE ESRD (P = 0.16). Median (inter-quartile range) follow-up time was 1.6 (0.3–3.5) years in SLE and 1.4 (0.4–3.2) years in non-SLE, P = 0.74. The adjusted hazard ratio (HR) for all-cause mortality for SLE patients who were aPL/LA+ vs aPL/LA− was 9.93 (95% CI 1.33, 74.19); the adjusted HR for non-SLE aPL/LA+ vs aPL/LA− was 0.77 (95% CI 0.14, 4.29).
Conclusion. SLE ESRD patients with aPL/LA+ had higher all-cause mortality risk than SLE ESRD patients without these antibodies, while the effects of aPL/LA on mortality were comparable among non-SLE ESRD patients.
aPLs are a heterogeneous group of antibodies against phospholipids or phospholipid-binding proteins that can develop in individuals with or without autoimmune diseases [1–3]. These antibodies are thought to be directly involved in the pathogenesis of APS, characterized by the presence of persistent aPL and the development of thrombosis and/or pregnancy morbidity . Not all individuals with aPL develop thrombotic complications during their lifetime, and aPLs may persist before thrombotic complications develop . However, the presence of aPLs, even in individuals without past thromboses, is associated with increased cardiovascular mortality [5–8] and end-organ damage, including end-stage renal disease (ESRD) [9–11].
Certain aPLs, in particular LA, are more strongly associated with thrombosis than other aPLs [12–15]. In addition, the risk of thrombosis increases in the presence of multiple (particularly triple) positivity for aCL, anti-β2 glycoprotein I (anti-β2GPI) and LA [16, 17]. The proportion of SLE patients who have APS or are positive for aPLs ranges between 10 and 44% in past studies [18, 19], and the presence of aPLs in SLE is associated with increased morbidity from thrombosis and pregnancy complications, as well as cardiovascular mortality [5–8, 13, 20]. Similarly, a high proportion of aPLs and a possible association between aPLs and an increased risk of thrombotic complications and cardiovascular mortality have been reported among individuals with ESRD undergoing haemodialysis (HD) .
Despite the fact that SLE patients are younger than non-SLE patients when they develop ESRD, mortality in SLE patients on HD is twice as high as mortality in non-SLE on HD [22, 23]. Factors contributing to the high mortality rates among SLE ESRD patients are not well understood. Although the presence of aPLs is associated with adverse outcomes in SLE without ESRD (as well as in the general ESRD population), there are no studies to date comparing the mortality risks associated with the presence of aPLs in HD patients with and without SLE.
Therefore, we compared the proportions of aPLs and/or LA (aPL/LA+) in ESRD patients with and without SLE on HD, and investigated the association between the presence of aPL/LA+ and all-cause mortality. We hypothesized that aPL/LA+ would be associated with increased all-cause mortality in ESRD, with higher risk in SLE than non-SLE ESRD.
We employed the Montefiore electronic medical record (EMR) system using Clinical Looking Glass, a proprietary software application developed at Montefiore Medical Center (MMC), that allows clinicians and researchers to identify populations of interest from the medical centre database and to gather information about laboratory data, medications, demographics and mortality . MMC is a community-based urban tertiary care centre that provides primary and specialty care to over 2 million people in the Bronx, New York (http://www.montefiore.org/community). We identified all patients over 18 years old with ESRD on HD followed at MMC who had aPL measured at least once between 1 January 2006 and 31 January 2014.
ESRD patients were identified using the following International Classification of Disease (ICD)-9 codes: chronic kidney disease stage V (585.5); ESRD (585.6) or admit for renal dialysis (V56.0). Subsequently, all charts were reviewed to exclude patients who had received HD for <90 days, never started HD, or patients who had a functioning kidney transplant at the time when aPLs were measured. All medical records of patients with ESRD and aPLs identified above with at least one ICD-9 code for 710.0 for SLE were reviewed by a rheumatologist (A.B.) to identify SLE patients who fulfilled either the ACR criteria [25, 26] or the SLICC criteria for SLE [26, 27].
From the medical records of the SLE and non-SLE ESRD patients in our final dataset, we collected the following data: age at the time of the first aPL measurement, sex, race (white, black, other/multiracial) and the total number of comorbidities using the Charlson Comorbidity Index (CCI) . The CCI was based upon ICD-9 codes prior to the first aPL measurement. Information on anticoagulation with warfarin or heparin at the time of the first aPL measurement was recorded from reviewing inpatient and outpatient prescription information in EMR.
Mortality data in the Clinical Looking Glass were obtained directly from hospital records and from the Social Security Administration’s Death Master File. This project was approved by the Institutional Review Board at the Albert Einstein College of Medicine/Montefiore Medical Centre.
APL was routinely tested at our centre during this time using BIO-RAD EIA kits (BIO-RAD Laboratories Inc., Hercules, CA, USA). LA was reported as positive or negative in accordance with the guidelines of the International Society on Thrombosis and Haemostasis . Similar to prior studies, low aCL positivity was defined as aCL levels 20 IU and 40 IU (low aCL), and moderate to high levels were defined as aCL levels >40 IU (moderate/high aCL) [2, 30]. Similarly, low anti-β2GPI positivity was defined as anti-β2GPI levels 20 IU and 40 IU (low anti-β2GPI), and moderate to high levels were defined as anti-β2GPI levels >40 IU (moderate/high anti-β2GPI). Based on the 2006 Sydney criteria, aPL positivity (aPL/LA+) was defined as aCL IgG or IgM >40 IU, anti-β2GPI IgG or IgM >40 IU, or positive LA . Possible reasons for aPL measurements were ascertained from medical record reviews and defined as follows: APS history prior to the development of ESRD, confirmed or suspected arterial or venous thrombosis, HD access thrombosis or occlusion, other (including altered mental status, myocardial infarction, accelerated atherosclerosis and pulmonary hypertension), or unknown. We also ascertained the following non-criteria aPLs that are routinely measured at our centre: aCL IgA, anti-β2GPI IgA and [AQ5]aPS IgA, aPS IgG and aPS IgM. Although the latter antibodies are not included in the current criteria, they have been reported to be associated with cardiovascular complications and mortality in a number of studies [31, 32]. For patients with aPLs or LA measured more than once after ESRD onset, the first aPL or LA measurement was used in the analysis of this study.
Baseline characteristics were compared between SLE and non-SLE ESRD groups and between aPL/LA+ subjects and aPL/LA− subjects in SLE and non-SLE ESRD groups. The aPL subtypes, including aCL, anti-β2GPI and aPS, and LA positivity were summarized and compared between SLE and non-SLE ESRD groups. Wilcoxon-Mann–Whitney tests were used to compare continuous variables, while Pearson’s chi-square tests (or Fisher’s exact test when appropriate) were used to compare categorical variables.
Second, we investigated the association between aPL/LA positivity and all-cause mortality during the study period. We chose all-cause mortality as the main outcome rather than thrombosis-related mortality, since the contribution of thrombosis to cause of death was difficult to ascertain , and cause of death listed on death certificates is not always accurate and reliable . Time to death was defined from the first aPL measurement to the date of death; subjects who had not died during the study period were censored on the date of renal transplant, the individual’s last visit date, or the last date of the study (31 January 2014), whichever came first. Kaplan–Meier survival curves were plotted for the four groups stratified by SLE group status and aPL status, and the log-rank tests were used to compare these survival curves. Cox proportional hazards regression models were used to assess whether the subjects with positive aPL/LA had higher risks of mortality in the SLE group compared with non-SLE group using an interaction term between SLE and aPL status. The models used age as the time-scale with the delayed entry approach and additionally adjusted for comorbidities using CCI, which was the only covariate with a P < 0.20 in the univariable analysis. In an effort to develop the most parsimonious models, as race, sex and anticoagulation use were not associated with mortality in the univariable analysis, they were not included in the final models. Adding anticoagulation use, prior APS history, sex and race one at a time into the Cox regression model did not change study conclusions. Statistical analyses were performed using SAS 9.4 (SAS Institute Inc., Cary, NC, USA).
Of the 8683 ESRD patients over 18 years old followed at our centre during the study period, 180 individuals had aPL measured at least once after ESRD onset (Fig. 1). Sixty-two individuals who either recovered renal function at the time of aPL measurement or never started HD and 13 individuals who had a functional kidney transplant at the time of their aPL measurement were excluded. Of the 105 remaining individuals with chronic ESRD on HD, 41 had at least one ICD-9 billing code for SLE. Seven of these individuals were not included in the final analysis because there was not enough information to document sufficient ACR or SLICC criteria for SLE, based on our medical record review. Therefore, 34 SLE ESRD and 64 non-SLE ESRD patients were included in the final analysis.
There were no differences with respect to age, race or ethnicity between patients included in our study and 8503 ESRD patients without aPL measurements. However, 66% of ESRD patients with aPL were women compared with only 47% of ESRD patients without aPL (P < 0.001).
Among 98 study patients, SLE ESRD patients were significantly younger [40.4 (12.5) vs 51.9 (18.1) years old, P = 0.001) and included a higher percentage of women (88.2% vs 54.7%, P = 0.0007) compared with non-SLE ESRD (Table 1). There were no significant differences between SLE and non-SLE groups with respect to Hispanic ethnicity, use of anticoagulation therapy or CCI scores. Potential reasons for aPL measurements varied substantially between the two groups. The two most common reasons identified among SLE patients were prior documented APS history (29.4%), and no reasons identified (41.2%), suggesting that physicians were more likely to measure aPL in SLE patients because of the known association between SLE and the presence of aPL. On the other hand, the two most common reasons among non-SLE patients were arterial/venous thrombosis (either confirmed or suspected) (34.4%) and HD vascular access complications (28.1%).
Baseline characteristics of aPL/LA+ and aPL/LA− patients within SLE and non-SLE groups are summarized in Table 2. The SLE aPL/LA+ subgroup included a higher percentage of men than the SLE aPL/LA− subgroup (37 vs 4%, P = 0.03), and the total number of comorbidities was higher [median (inter-quartile range, IQR): 8.5 (6.5–10.0) vs 5.0 (3.0–8.0), P = 0.04]. There were no differences between non-SLE aPL/LA+ and non-SLE aPL/LA− subgroups with respect to any of the variables in Table 2. The majority of patients were non-Caucasian, and none of the SLE patients were white.
In the aPL/LA+ group, 37.5% of patient were receiving anticoagulation compared with 18.3% of patients in the aPL/LA− group (P = 0.09). The proportion of patients receiving anticoagulation was also higher in the aPL/LA+ SLE subgroup compared with the aPL/LA− SLE subgroup [25 and 11.5%, respectively (P = 0.57)]. Similarly, the proportion of patients receiving anticoagulation was higher in the aPL/LA+ non-SLE subgroup compared with the aPL/LA− non-SLE subgroup [50 and 21%, respectively (P = 0.10)]. However, none of these differences were statistically significant.
The frequencies of aPL and LA among SLE and non-SLE patients are shown in Table 3. Nine percent of SLE ESRD patients had medium/high aCL IgG, whereas none of the non-SLE patients had medium/high aCL IgG (P < 0.001). There was a significant association between aCL IgG and SLE (P < 0.001), and a borderline significant association between SLE and anti-β2GPI IgG (P = 0.07). The frequency of LA was 7/28 (25%) in SLE ESRD and 7/57 (12%) in non-SLE ESRD patients (P = 0.14). Overall, the frequency of aPL/LA+ was 24% in SLE and 13% in non-SLE ESRD patients (P = 0.16). The frequencies of non-criteria antibodies were low in both groups and not significantly different. In the entire sample, only one SLE patient with a prior history of APS had triple positivity, defined as positive aCL, anti-β2GPI and LA, believed to be associated with the highest thrombosis risk compared with individual antibody positivity [16, 17].
In the SLE group, 10 (29%) patients had a history of APS prior to developing ESRD, while in the non-SLE group only 2 (3%) had a prior history of APS. In a sensitivity analysis, after individuals with prior APS history were excluded, the association between aCL IgG and SLE remained significant (P = 0.006), but the proportion of aPL/LA+ decreased to 8% in SLE and 11% in non-SLE (P > 0.99). The frequencies did not change significantly when the seven patients with possible SLE (excluded from the main analysis because of incomplete information) were included in the analysis (data not shown).
Next, we investigated whether aPL and/or LA positivity was associated with all-cause mortality. Median (IQR) follow-up time in the SLE patients was comparable with that of non-SLE patients: 1.6 (0.3–3.5) vs 1.4 (0.4–3.2) years, P = 0.74. In total, 30 patients died during the study period. Among those who died, the median (IQR) duration from the first aPL assessment to the time of death was 0.3 (0.2–0.8) years in the aPL/LA+ group, and 1.2 (0.2–1.9) years in the aPL/LA− group, P = 0.22. In the SLE group (n = 34), 5 of 8 (63%) subjects with aPL/LA+ died, whereas only 4 of the 26 (15%) subjects with aPL/LA− died (P = 0.02). In the non-SLE group (n = 64), 2 of 8 (25%) subjects with aPL/LA+ had died, whereas 19 of the 56 (34%) subjects with aPL/LA− had died (P > 0.99).
In the overall study sample, 25% of patients who survived and 13% patients who died were on anticoagulation at the time of the aPL measurement (P = 0.28). In the SLE subgroup, anticoagulation use was 20% among survivors vs 0% among patients who died (P = 0.29). In the non-SLE subgroup, anticoagulation use was 28% among survivors and 19% among patients who died (P = 0.55). Although the use of anticoagulation therapy was lower among patients who died, none of the differences reached statistical significance.
Kaplan–Meier survival curves were plotted for the four groups stratified by SLE and aPL/LA status (Fig. 2) (log-rank test, P = 0.009). Among SLE ESRD patients, the survival curve for the aPL/LA+ group was significantly different from that of the aPL/LA− group (log-rank test, P = 0.001), whereas among non-SLE ESRD patients, the survival curves were not significantly different between the aPL/LA+ and aPL/LA– groups (log-rank test, P = 0.75).
In the Cox proportional hazards model adjusted for the number of comorbidities using CCI, the adjusted hazard ratio (HR) for death for the aPL/LA+ vs aPL/LA– comparison was 9.93 (95% CI 1.33, 74.19) for the SLE patients; among the non-SLE patients, the adjusted HR was 0.77 (95% CI 0.14, 4.29), and there was a significant multiplicative interaction between aPL and SLE with respect to mortality (P = 0.05). However, the latter HR has to be interpreted with caution because the number of deaths was very small in the non-SLE aPL/LA+ group (n = 2). These results remained largely unchanged after adding prior APS history into the above model and when the earliest recorded ESRD date was used as a starting date instead of the first aPL date (data not shown). These conclusions remained unchanged after anticoagulation use, prior APS history, sex and race were added to the model one at a time.
In this study, we investigated the possible association between all-cause mortality and the presence of aPL/LA+ in HD patients with and without SLE. We found that the frequency of medium/high aCL IgG was significantly higher among SLE ESRD patients, suggesting that aPL profiles differed between SLE and non-SLE groups. We also found that there was a multiplicative interaction between the presence of SLE and aPLs in their association with the risk of death. Furthermore, aPL/LA+ was significantly associated with all-cause mortality only in SLE ESRD, suggesting that aPLs may be more pathogenic in SLE ESRD. This finding is consistent with the notion that aPLs are more pathogenic in SLE compared with in the general population [3, 33]. To the best of our knowledge, this is the first study to report a differential association between aPLs and mortality in SLE and non-SLE ESRD patients. Our findings are strengthened by the fact that we only included SLE patients with confirmed ACR and/or SLICC criteria for SLE, and we reported the prevalence of several antibody subtypes and LA. Interestingly, when stratified by SLE and aPL status, SLE aPL/LA+ patients had the lowest survival, whereas survival was similar among SLE aPL/LA–, non-SLE aPL/LA+ and non-SLE aPL/LA– patients. In the Cox model adjusted for comorbidities, there was a significant interaction between aPLs and SLE. Taken together, these findings suggest that the lower survival observed in SLE ESRD compared with other ESRD patients  may be, at least in part, due to the presence of aPLs.
Some, but not all, previously published studies reported that aPLs were present in up to 50% of ESRD patients, and associated with an increased risk of vascular access thrombosis and mortality in ESRD [34–38], reviewed in . Brunet et al.  studied the prevalence of aPLs and their association with thrombosis and mortality in 97 ESRD patients. Although SLE patients were excluded from that study, no association was found between the presence of aPLs and mortality among ESRD patients without SLE, which is similar to our findings. In the previously published studies, as in the current study, the most frequently reported aPLs were aCL IgG and LA. However, no information was provided about whether or not ESRD patients developed aCL IgG before or after reaching ESRD and initiating HD. In contrast to our study, these studies excluded SLE patients, included only small numbers of SLE patients (<3%), or did not differentiate between SLE and non-SLE ESRD patients. Furthermore, previously published studies have not accounted for prior APS history, and positivity was not defined as low positive vs moderate/high positive.
One of the main limitations of this study is the possibility of selection bias due to its retrospective design and the inclusion of individuals who had aPLs measured in the course of their clinical care. The presence of aPL/LA is a risk factor for progression to ESRD in lupus nephritis, and aPL/LA+ patients may be overrepresented among SLE ESRD patients. Therefore, the true prevalence of aPLs in the ESRD population with and without SLE is likely lower than in this study population. In addition, physicians may have been more likely to measure aPLs in SLE ESRD than in non-SLE ESRD patients because of the known association between SLE and aPLs (and not because of a thrombotic event or complication). On the other hand, physicians may have been more apt to measure aPLs in non-SLE ESRD patients only if they had thrombotic events or suspected thrombotic events. Therefore, the true differences between the prevalence of aPLs in SLE and non-SLE ESRD patients would have been underestimated rather than overestimated in this study population. Finally, since some of these antibodies may have been transiently elevated, the prevalence of persistent aPLs/LA may be lower than what is reported in our study. The 12.5% prevalence of aPL/LA+ in non-SLE was slightly higher than the 5% previously reported in the general population [18, 19]. Although consistent with prior studies reporting the frequency of aPL/LA+ in other SLE populations, these findings should be interpreted with caution, and prospective studies are needed to estimate the true prevalence.
We did not have information on some of the important potential confounders, including SLE activity, and dialysis access type, that may have affected mortality. Although relatively small numbers of aPL/LA+ patients and deaths limits our ability to adjust simultaneously for multiple potential cofounders, this is the largest described group of aPL frequency and association with mortality in SLE patients with ESRD. Even though the numbers were small, we have found significant differences with respect to antibody profiles and the risk of death between SLE and non-SLE ESRD aPL/LA+ patients. Therefore, our study provides compelling evidence for conducting larger prospective studies investigating the role of aPLs in mortality among ESRD patients.
Several factors may limit the generalizability of our study. First, the study was conducted in a large urban care centre in New York, and the majority of our patients were non-white. In the Hopkins cohort, the prevalence of LA was slightly higher in Caucasians than in Blacks [28 and 23%, respectively]. Therefore, further studies are needed to ascertain whether our findings hold in Caucasian ESRD patients. Second, although we did not focus on renal transplantation for this report, aPL/LA+ SLE patients may be less likely to receive renal transplantation due to high complication rates in aPL/LA+ patients . Therefore, the findings of our study only apply to ESRD patients on HD, and cannot be generalized to transplant patients. The contribution of thrombosis to cause of death was difficult to ascertain, and the numbers were too small to differentiate between various causes of death. Further studies are needed to determine the association between aPL/LA+ and specific causes of death.
We did not find an association between anticoagulation and mortality, and our findings were unchanged after adjusting for anticoagulation. Anticoagulation is recommended for preventing recurrent thrombosis in patients with APS (secondary prevention). However, anticoagulation is associated with an increased risk of bleeding, especially in ESRD patients, and it is unclear how effective anticoagulation is in preventing thrombotic events in this population . Therefore, larger studies are needed to assess long-term risks and benefits of anticoagulation in aPL/LA+ ESRD patients.
Despite the above limitations, this study underscores the importance of conducting future prospective studies of a potential causal relationship between aPLs, thrombosis and mortality in ESRD with and without SLE. If the association between the presence of aPLs and mortality in ESRD is confirmed in prospective studies, preventive strategies, such as HCQ use and aggressive control of modifiable risk factors [41–44], could be tested for effects on decreasing thrombosis and mortality in SLE ESRD.
Funding: No specific funding was received from any funding bodies in the public, commercial or not-for-profit sectors to carry out the work described in this manuscript.
Disclosure statement: K.H.C. has received a research grant NIH K24 AR066109. All other authors have declared no conflicts of interest.