This review showed that many authors define dialysis withdrawal as generalized dialysis discontinuation for reasons including social factors, patient preference, the clinician’s opinion, modality change, transplant, recovery, and death. Few authors have used patient-selected dialysis withdrawal as a reason for dialysis discontinuation. The review identified a scarcity of literature on the relationship between individual factors and dialysis withdrawal. Moreover, there was conflicting evidence in few studies that explored the associations between dialysis withdrawal and demographics, renal diseases, health behaviour, comorbidities, physiological indicators, or dialysis factors.
Influence of demographic factors on dialysis withdrawal
The relationship between demographic factors such as age, gender, race/ethnicity, residence, education, employment and marital status and dialysis withdrawal have been explored by only a few researchers in different geographical locations [
8‐
10,
12,
16,
17,
19,
31,
32]. Older age was associated with higher rate of dialysis withdrawal. Discontinuation of dialysis was more frequent in patients ≥70 years old versus those < 70 years old (29.83% versus 18.14%,
p < 0.001) [
14]. Ellwood et al. (2013) found higher rates of withdrawal in patients aged ≥75 years as compared to patients aged < 75 years, and increasing age was significantly associated with dialysis withdrawal (HR, 1.81; 95% CI, 1.75–1.88) [
10]. Similarly, Findlay et al. (2016) found older age was significantly associated with dialysis withdrawal [
33]. Older age patients have multiple medical problems and comorbidities that worsen with increased duration of dialysis. These results suggest that this drastic change in physical and mental health leads to increased dialysis withdrawal and discontinuation of treatment in older populations [
10,
12,
34]. Yet, Urban et al. (2013) [
32] found insignificant differences with respect to age, gender, and living situation between elective dialysis withdrawal and non-withdrawal groups. This difference among studies was not related to the definition of dialysis withdrawal, as all the above studies used generalized (non-specific) discontinuation of dialysis [
10,
12,
32,
34]. The difference could have been due to a small subsample (
n = 10) in the study by Urban et al. [
32] in which the number of people was insufficient to detect any differences among these factors.
Dialysis withdrawal also appears to vary with race and ethnicity. Patient-elected (based on patient decision) dialysis discontinuation was more frequent in whites than blacks (29.5% versus 14.7%,
p < 0.001) or patients of other races (29.5% versus 19.2%,
p < 0.001) [
14]. Similarly, other authors have reported a higher rate of dialysis withdrawal in white people versus African Americans and Asians [
14,
35‐
39]. This difference in dialysis withdrawal between different ethnicities is unclear although these findings highlight the role of social and cultural values in the decision to withdraw from or continue dialysis [
9,
10]. One of the reasons for increased likelihood of dialysis withdrawal in whites may be related to more liberal values such as religious, societal, and cultural beliefs, which can have an influence in deciding to continue or withdrawal from dialysis [
9,
40,
41]. The differences may also be more pronounced in a geographical setting with historical racial tensions and issues, such as in the US, where non-white races continue dialysis and have a lack of trust in healthcare settings because of inequalities in healthcare in comparison to white populations [
13].
The association between dialysis withdrawal and gender is inconsistent and inconclusive. Few authors have shown that women are more likely to withdraw from dialysis than men and a higher dialysis withdrawal rate in women versus men (26.3% versus. 23.0%,
p < 0.001) [
14]. However, Seshasai et al. (2016) found younger ages, males, and white race had a high withdrawal rate than older ages, females, and non-white races [
8]. Study differences may be related to gender inequality in treatment and decision-making management [
9,
42]. Gender bias in clinical-decision making is still prevalent in many underdeveloped regions and low socioeconomic areas [
9,
42]. In a few religions, cultures, societies, races, and ethnicities, women are less privileged than men and have less access to expensive, quality health care such as renal dialysis and transplantation [
9,
42]. This difference may also be related to the sample population and differences in defining dialysis withdrawal. Gessert et al. (2013) [
14] used all types of dialysis therapies and defined dialysis withdrawal as discontinuation of dialysis due to any reason. However, Seshasai et al. (2016) [
8] included only HHD patients and defined dialysis withdrawal as no HHD during a period of > 60 days.
The area of residence was also associated with dialysis withdrawal since residents of small towns and villages have a higher rate of dialysis withdrawal than residents of large cities and towns (26.9% versus 24.3%,
p < 0.001) [
9,
14]. This difference between dialysis rate may be related to the reduced dialysis facilities in small towns and villages compared to cities. Morton et al. (2012) found that the distance to a dialysis centre and the ability to travel are associated with patient choice of dialysis or conservative treatment (discontinuation of dialysis) [
43]. Similarly, Elwood et al. (2013) found a higher risk of dialysis withdrawal in patients who had to travel longer than 60 min when compared with patients who had to travel 15 min or less to arrive at the dialysis facility [
10]. Authors have also shown that certain marital status such as divorced or widowed and living in nursing homes was also one of the predictors of dialysis withdrawal [
31,
44]. However, Birmele et al. (2004) found that living alone or with family or spouse was not a significant predictor of withdrawal [
12]. This finding may be due to a small subsample size (
n = 40) in the withdrawal group. Similarly, some authors have shown that being married, living alone, or divorced was not associated with dialysis withdrawal although the association was significant in an unadjusted analysis [
17,
18]. Fissell et al. (2005) found living in a nursing home was significantly associated with dialysis withdrawal in both the adjusted and non-adjusted models, and less than 12 years of education was insignificant in either model [
31]. The author also found employment was a significant factor in dialysis withdrawal [
31]. All of the studies above defined dialysis withdrawal as discontinuation for any reason and did not provide a subgroup analysis to make further inferences (such as whether the mentioned demographic factors has more influence on any particular reason for dialysis withdrawal). The reasons for the differences between socioeconomic indicators and dialysis withdrawal may be because patients with less education and lower employment status were underprivileged and lacked access to good quality health care as well as the communication skills, transportation, and community support systems required to continue frequent visitation to a dialysis centre for treatment (3–4 times a week for conventional HD) [
31,
44].
Influence of renal disease aetiology on dialysis withdrawal:
There is a scarcity of literature on the association between the aetiology of renal disease with dialysis withdrawal. Ellwood et al. (2013) found that diabetes-induced ESRD and renovascular disease were associated with increased rates of withdrawal (HR = 1.58 [1.37–1.82] and HR = 1.26 [1.06–1.49], respectively) vs. glomerulonephritis [
10]. However, Birmele et al. (2004) found that causes of renal diseases such as glomerulopathy, diabetic, interstitial, and vascular nephropathies, and polycystic kidney disease were not associated with dialysis withdrawal [
12]. Similarly, another study showed that hypertension, diabetes, and glomerulonephritis were not associated with dialysis withdrawal [
19].
These contrasting findings may be attributed to the age of the sample. For example, Ellwood et al. (2013) found that renovascular disease was significantly associated with dialysis withdrawal in patients in the age group < 75 years old [
10]. The difference may also be attributed to different definitions of dialysis withdrawal used by different authors. Birmele et al. (2004) define dialysis withdrawal as all types of dialysis discontinuation [
12], whereas Ellwood et al. (2013) defined dialysis withdrawal as all types of discontinuation except for recovery patients [
10]. Workeneh et al. (2015) defined dialysis withdrawal as the discontinuation of dialysis for several reasons except for transplant and recovery patients [
19].
Influence of health behaviours on dialysis withdrawal
Behaviour risk factors such as smoking, substance abuse, alcohol dependence, and BMI are associated with dialysis withdrawal. Seshasai et al. (2016) showed that smoking and alcohol use were associated with dialysis withdrawal in the HD group (HR = 1.34 [1.01–1.78]) [
8]. Similarly, Fissell et al. (2005) showed that alcohol dependence for less than 12 months showed higher odds of dialysis withdrawal in the unadjusted analysis although it was insignificant in the adjusted analysis [
31]. Additionally, having a BMI < 18.5 kg/m (HR = 1.37[1.16–1.61]) was associated with increased rates of withdrawal [
10]. Patients having a low BMI may have malnutrition and poor health status, thus increasing the odds of dialysis withdrawal as a result of comorbidity worsening and physically deteriorating conditions associated with dialysis [
10]. However, categorization of BMI into underweight (< 18.5), healthy (18.5–25), overweight (> 25–30), and obese (> 30) were not associated with PD discontinuation [
19]. Similarly, Hazama et al. (2014) found that BMI was not associated with PD withdrawal [
16]. The differences in the relationship between dialysis withdrawal and BMI may be attributed to type of dialysis with withdrawal from PD less dependent on BMI when compared with HD [
8,
10,
16,
19]. The difference may also be attributed to different causes of dialysis withdrawal selected in each study, such as discontinuation or change of modality [
8], technique failure and complications [
16], and all types of discontinuation except recovery and transplantation [
19].
Influence of physiology indicators on dialysis withdrawal
Blood and serum markers such as serum albumin, creatinine and dialysate vascular endothelial growth factor are associated with dialysis withdrawal. Hazama et al. (2014) found lower hemoglobin (≤ 11.2 g/dL) and lower serum albumin (≤ 3.31 g/dL) were associated with PD withdrawal, but creatinine, uric acid, and weekly Kt/v were not [
16]. The reason for the insignificant association of some of the variables may be related to the type of dialysis, as that study included only PD patients. Hazama et al. (2014) found higher (> 27 pg/mL) dialysate vascular endothelial growth factor was associated with dialysis withdrawal in PD patients. Dialysate vascular endothelial growth factor is a biomarker produced in the peritoneal tissue of patients undergoing PD and has been used as an independent predictor of serum albumin levels [
16]. Some authors have shown that excretion of peritoneal albumin was significantly associated with cardiac diseases, resulting in dialysis withdrawal [
45‐
47]. McDade-Montez et al. (2006) found that serum creatinine and phosphate were associated with withdrawal, but not serum potassium [
17]. The relationship between serum phosphate and dialysis withdrawal highlights the importance of the dietary control of phosphorus and the use of phosphate-binding medications during dialysis [
17]. The association between serum creatinine and dialysis withdrawal may be explained by many patients on dialysis having lower BMI, or lack of adequate nutrition, and reduced muscle mass along with low serum creatinine [
17,
48].
Influence of comorbidities on dialysis withdrawal
Authors have found that comorbidities such as dementia, diabetes, cerebrovascular diseases, and malignancies are associated with dialysis withdrawal. Addition of comorbidities and their combinations may also be positively associated with withdrawal from dialysis [
9]. Patients with chronic conditions such as cancer, dementia, diabetes, hypertension, and cachexia are more likely to withdraw than those with acute conditions such as stroke, infection, angina, heart failure, cellulitis and gangrene and infectious diseases such as hepatitis B and C and neurological complications [
8‐
12,
16‐
19,
31]. Patients with poor health status at the start of dialysis also have a higher risk of dialysis withdrawal. Chronic diseases gradually deteriorate patient health status, leading to complications that initiate a cascade of health issues. These health issues increase the burden of disease and lead patients to discontinue dialysis treatment [
38].
In addition to physical health, pain, an important the quality of life measures, is also a significant predictor of dialysis withdrawal. Authors have shown higher withdrawal rates in patients suffering from chronic pain [
49]. Davison (2012) found that almost half of patients (50%) have significant pain at the time of dialysis discontinuation [
50]. However, patients with comorbidities have a higher risk of depression, despair, loss of positive attitude, and hopelessness than patients without comorbidities [
17,
49]. It is difficult to distinguish and understand the biologic plausibility between pain and depression in relation to dialysis withdrawal [
17,
49]. It may be that decisions to discontinue dialysis in patients with comorbid conditions and poor health status is due to depression and not chronic pain or discomfort [
17,
49].
Influence of dialysis indicators on dialysis withdrawal
The relationship between type of dialysis such as HD, HHD, and PD, and dialysis withdrawal is inconsistent. Mizuno et al. (2011) found a higher dialysis withdrawal rate in HD patients than PD patients [
11]. Chan et al. (2012) found significant effects of PD on dialysis withdrawal in both unadjusted and adjusted models [
9]. These differences in findings between studies may be explained by general health status, disease burden, and comorbidities at the start of dialysis [
12].
Peritoneal dialysis is mostly performed at home in patients that have more self-control over the treatment and family support to be able to perform routine dialysis [
12]. This self-control of dialysis management improves patient’s confidence, active participation in daily activities, and mental health and wellbeing, thus ultimately reducing chances of dialysis withdrawal when compared with in-hospital HD [
9,
12]. However, these findings may be attributed to selection bias and confounding factors. Patients having high disease burdens and comorbidities have higher odds of undergoing HD than PD [
12]. Poor mental health status has been associated with dialysis withdrawal; therefore, HD patients have higher rates of dialysis withdrawal than PD [
49]. The differences in findings may also be due to PD-associated complications. Koc et al. (2011) found that 33% of patient-selected dialysis withdrawal was because of peritonitis (50%) and insufficient PD (50%) [
30]. However, few authors found insignificant effects of types of dialysis on dialysis withdrawal [
12]. Ellwood et al. (2013) found patients undergoing HD have a higher rate of withdrawal when compared with non-withdrawal, but this association between HD and PD with dialysis withdrawal was insignificant [
10]. This inconsistent relationship between the type of dialysis and dialysis withdrawal may be related to the definition of dialysis withdrawal. Koc et al. (2011) define dialysis withdrawal as patient-selected discontinuation [
30], while other studies define dialysis withdrawal as the general discontinuation of dialysis for multiple reasons, including technique failure, complications, and preferences of physicians, patients, and relatives [
9,
12,
49].
The relationship between duration of dialysis and dialysis withdrawal is inconclusive; McDade-Montez (2006) found an insignificant association for duration of dialysis (in months) between withdrawal and non-withdrawal groups [
17]. This finding may be attributed to the small subsample of the dialysis withdrawal group (
n = 40). Another study showed that duration of dialysis in years was not significantly different between patients who withdrew or continued dialysis [
12]. Many dialysis patients have short survival, and exploration of duration of dialysis in years was not an appropriate measure.
Implications for the future research
While authors have explored different factors relating to dialysis withdrawal and revealed differences in dialysis attrition rates, the strength of association for similar factors is inconsistent across various studies. These differences may be due to several reasons. The definition of dialysis withdrawal is not consistent, as different authors have used this concept for multiple reasons and causes. Few authors (
n = 3) have used a specific definition of dialysis withdrawal. Most authors (
n = 12) defined dialysis withdrawal as any type of discontinuation, discontinuation, withholding, death, treatment refusal by patients and caregivers, or technique failure [
8‐
12,
16‐
20,
28,
53]. Discontinuation was defined as no dialysis treatment within a 60 -day period [
8]. Withdrawal was defined as either withdrawal from treatment, suicide, accidental death, patient refusal for further treatment, or treatment cessation [
9]. Withholding therapy was defined as stopping, not starting, or increasing a life-sustaining intervention. Technique failure was defined as discontinuation of PD for > 6 weeks [
17]. Furthermore, few studies have provided exclusion criteria when defining dialysis withdrawal, such as excluding patients with a return of kidney function [
10]. Future studies should explore the specific type of dialysis withdrawal or provide a subsequent analysis for each reason or case of dialysis withdrawal to make firm conclusions.
Dialysis withdrawal rate and associated factors are dependent on the type of modality such as HD or PD [
8,
9,
12,
19]. Many of the studies have selected either PD or HD patients but not both, making comparisons and inferences difficult to interpret [
8,
9]. To determine the role of dialysis modality (HD or PD) in dialysis withdrawal, future studies should explore both modalities in one study setting.
Patients with comorbidities such as diabetes, heart and other chronic debilitating diseases have been shown to be associated with dialysis withdrawal [
9]. Poor general health condition due to comorbidities can further reduce dialysis patients’ quality of life, resulting in higher likelihood of dialysis withdrawal than in patients with otherwise good health condition [
12,
26,
52]. However, few authors have shown insignificant effects on dialysis withdrawal from diabetes, vascular disease, stroke, cancer, arrhythmias, and lung disease. This difference may be explained by the number of diseases or comorbidities included in the study, duration, severity and types [
9,
20,
53]. Many studies have used a small sample or subsample and examined the factors associated with dialysis withdrawal without a priori calculation, resulting in type II error (false negative) [
16,
29,
30,
32].
Old ages, females, whites, and those with chronic diseases are associated with dialysis withdrawal [
9,
10,
26,
52]. However, few studies showed that demographic factors are not associated with dialysis withdrawal [
12,
31]. The geographical setting of the study has also accounted for these differences; factors such as race/ethnicity, preferences of dialysis modality, and whether to withdraw dialysis are sociodemographic dependent [
8‐
12,
16,
17,
26,
31,
51]. Authors should conduct multi-centric and population-based comparative studies to evaluate the influence of demographic factors on dialysis withdrawal.