The three-year incidence of major hemorrhage among older adults initiating chronic dialysis

Dialysis patients are known to experience a significant risk of hemorrhagic events. Accurate determination of the risk has been difficult due to variations in the definition of hemorrhage and the use of non-validated diagnostic codes in previous studies. Further it remains unclear if or to what magnitude the hemorrhagic risk changes with increasing age and over time.

We found that one in seven patients (14.43%) will experience a hospitalization with a hemorrhage within the first three years of initiating dialysis and this high risk did not increase with age or improve over time.

Patients who initiate chronic dialysis are older, have more co-morbid illness than a decade prior, and subsequently may experience more hospitalizations in follow-up. Hemorrhages are one of most common hospital admission diagnoses in the end-stage renal disease population and contribute significant morbidity and mortality [1]. Compared to the general population, the incidence of hemorrhage is considerably higher in those receiving dialysis with rates ranging from 35.4 to 80 per 1000-patient years (which is contingent on the study design, the definition used, method of outcome ascertainment and concurrent use of anti-thrombotic medications such as clopidogrel or warfarin) [2],[3].

In the non-dialysis population, the incidence of hemorrhagic events has been declining over the last two decades, attributed to the widespread use of gastrointestinal endoscopic procedures, treatments for H.pylori-related ulcers and the use of proton pump inhibitors [4]–[8]. In an American dialysis population, the incidence of non-variceal upper gastrointestinal bleeding had not changed over the last decade despite increasing patient comorbidity [9],[10]. It is unknown whether the same is true in other jurisdictions, or applies to other types of bleeding (lower gastrointestinal, intracranial bleeding) or related procedures (gastrointestinal endoscopy). Furthermore, the risk of bleeding appears to increase with advancing age [7]. In the non-CKD, non-dialysis population, hospitalizations for UGI hemorrhage per 100, 000 population increased dramatically with age (<65 31.7, 66 to 75 197.4, > 75 425.2) with a similar trend observed for lower GI hemorrhage. As the average age of incident dialysis patients is increasing over time, it follows that there may a parallel rise in the risk of major hemorrhage [11]. Lastly, the hemorrhagic risk may differ based on the dialysis modality as patients on peritoneal and hemodialysis are well known to differ based on co-morbidities, exposures (to heparin with the hemodialysis procedure) and functional status.

In this study we examined older adults who initiated chronic dialysis in Ontario, Canada over a ten-year period to determine the 3-year cumulative incidence and incidence rate (per 1000 person years) of hemorrhage and related gastrointestinal endoscopic procedures. We further examined hemorrhage incidence across three age (66 to 74, 75 to 84, ≥85), era (1998 to 2001, 2002 to 2005, 2006 to 2008) and dialysis modality (peritoneal, hemodialysis) and considered whether patients with hemorrhagic events were recently prescribed anti-thrombotic medication.

In this population-level study of older incident dialysis patients in Ontario, Canada, the 3-year cumulative incidence of hospitalization with hemorrhage was 14.43% or one in seven patients. The most common site of bleeding was lower gastrointestinal followed by upper gastrointestinal, intra-cerebral and subarachnoid hemorrhage, and many patients received an endoscopic procedure during their hospital admission. The incidence of hemorrhage was stable across age groups, over time, by dialysis modality and elevated among those recently prescribed anti-thrombotic medications. Among incident dialysis patients, the burden of hemorrhage is high and population-specific interventions are warranted.

Our finding of a high incidence of hemorrhage among incident dialysis patients with an observed rate of 52.64 per 1,000 person years is consistent with previous studies [2],[3],[24]. In prior large dialysis cohort studies event rates have ranged from 35.4 to 80 per 1,000 person-years [2],[3]. Differences in the reported rates are attributable to variations in the definition of hemorrhage and inclusion criteria of individual studies. Olsen et al. examined hemorrhage requiring hospitalization in patients with atrial fibrillation and included urinary tract and airway hemorrhage [3]. Sood et al. reported an event rate of 80/1000 patient-years for hemorrhage with hospitalization with a broader definition including epistaxis, vascular access-related hemorrhage and subdural hematoma, and the study cohort included both incident and prevalent dialysis patients [2]. Of note the Canadian dialysis population may differ from international counterparts with relatively higher use of central venous catheters and warfarin all of which may impact the risk of hemorrhage [2],[11].

As our rate estimates were determined in incident dialysis patients, they could serve to inform patients of their risk prior to dialysis initiation. The ability to quantitate risk of dialysis-related complications prior to dialysis initiation is increasingly important with the shift in demographics of the dialysis population to the elderly and need to balance quantity and quality of life. We further calculated the three-year cumulative incidence of hemorrhage as it may be a simple and meaningful way to communicate risk and understand disease burden from a patient perspective [25]. We found a 3-year cumulative incidence of 14.43% thus roughly 1 in 7 elderly dialysis patients will experience a hemorrhage requiring hospitalization within 3 years of dialysis initiation. This could be used in pre-dialysis discussions with patients to help them understand the risks of complications after dialysis initiation.

Among the specific subtypes of hemorrhagic events, we found upper and lower gastrointestinal hemorrhage (both well validated definitions of hemorrhage) to be the most common and roughly twenty to one hundred-fold higher in the elderly dialysis population compared to the general population [6],[26],[27]. Upper gastrointestinal hemorrhage has been previously reported in large US ESRD cohorts ranging from 23–57 per 1,000 person-years [10],[28]. This is comparable to our rate of 20.94 per 1000 patient-years. Similarly, subarachnoid hemorrhage and intra-cerebral hemorrhage were found to occur at rates 3- to 10- fold higher in the dialysis population compared to those reported in the general population [29]. The rate of subarachnoid hemorrhage among those with ESRD was recently reported as 3-fold higher in the US compared to our study (0.74 vs. 0.33 per 1000-patient years) [29]. Differences among the rates may be attributable to a higher prevalence of associated risk factors (such as women US: 56.3% vs Canada: 42.3% and African Americans US: 25.7% vs Canada <5%) [29].

The risk of hemorrhage was stable over time, by age group and dialysis modality. The steady temporal trends have been reported in the US dialysis populations for upper gastrointestinal and subarachnoid hemorrhage, however, the comparable risk across age groups to our knowledge has not been previously reported [10],[29]. The stable temporal trends in the dialysis population are in contrast to the general population where GI and SAH have been reported to be declining steadily over the last two decades [6],[30]. This has been attributed to advances in prevention (acid suppressing medications) or related risk factors (hypertension treatment for subarachnoid hemorrhage) in the general population. Yet it remains unclear whether therapies are effective or treatment targets are similar or comparable in the dialysis population. The stable risk of hemorrhage with advancing age is reassuring and again provides important prognostic information for the elderly initiating dialysis. Of note, the absence of an increase in hemorrhagic events in the elderly may reflect a selection bias as elderly patients who are less likely to experience a hemorrhagic event may also be more apt to choose dialysis therapy over conservative treatments. Surprisingly despite similar hemorrhage rates, individuals ≥ 85 were much less likely to receive related endoscopic procedures (p = 0.005). Although the specific reasons for not performing endoscopic procedures were unknown, there is an increased risk of complications, often due to the bowel preparation or anesthetics, with advancing age and frailty [31]. The hemorrhagic risk appears to be similar based on dialysis modalities with the only difference detected being in gastrointestinal endoscopic procedures. It should be noted that the two cohorts are known to differ significantly in terms of demographics, co-morbidities, functionality and exposures [32]. These disparities were not accounted for in the present analysis and represent future areas on investigation.

A question of interest is the risk of hemorrhage with the use of anti-thrombotic medications [1],[33]–[35]. We observed a large number of hemorrhagic events occurring following a recent warfarin or clopidogrel prescription. Of concern, the incidence of hemorrhage associated with clopidogrel prescription use was trending upward during the study period. To date, there remains very little evidence for efficacy of these medications in dialysis patients aside from warfarin use for a recent or recurrent thrombus [36],[37]. Despite the increase in prescriptions for clopidogrel associated with hemorrhagic events over time, there was no significant increase in total hemorrhagic events. Whether this represents possible selection bias, increased use of co-interventions or declines in hemorrhagic events among patients not treated with anti-thrombotics remains unknown.

Our study has a number of strengths. Our definition of hemorrhage was stringent and limited to severe events requiring hospitalization. Previous studies included surrogate measures of bleeding (such as blood transfusion), bleeding events managed in the outpatient setting and subtypes with limited clinical impact (epistaxis) [2],[3],[24]. We used well validated definitions for our outcomes and excluded those lacking validation (vascular access related bleeding). We determined a three-year follow-up to ensure estimates of hemorrhage incidence would be useful for sample size calculations in future clinical trials. As CORR captures nearly all incident dialysis patients in Canada, our findings are representative and generalizable.

Our study has some limitations. We did not include traumatic or peri-operative bleeding events. Medication prescription was obtained only in cross section at the initiation of dialysis or 120 days preceding an event, and did not account for medication continuation and discontinuation over time. We lacked information on ASA use or INR control among warfarin users. We did not account for recurrent events or events in prevalent dialysis patients which would have considerably increased our risk estimates. We did not have information regarding heparin-use during the hemodialysis procedure nor did we account for changes in modality over time.

Among incident dialysis patients are at high risk of hemorrhagic events requiring hospitalization. These events are stable over time, across age groups and by dialysis modality. Further studies of risk stratification for these events, and strategies to reduce these events, are needed.