In this study probable, confirmed or severe sarcopenia was present in 67.6% of elderly, stable satellite dialysis patients, with a median age of 71 years, using EWGSOP2 criteria for diagnosis. In a 2014 study of adults receiving haemodialysis, the prevalence of sarcopenia depended on the criteria and assessment method used [20]. Confirmed sarcopenia estimated from DXA appendicular lean mass index (ALMI) and grip strength ranged from 31 to 63% depending on ALMI cut points [20]. Using BIA and grip strength, confirmed sarcopenia ranged from 13 to 45%, depending on lean body mass index BIA cutpoints [20].

Patients on dialysis have a high risk for developing sarcopenia due to anorexia, poor nutrient intake, dialysis related factors such as nutrient loss into dialysate fluid, acidosis, chronic inflammation, comorbid illnesses and hormonal disorders [21]. While in the general population, sarcopenia is associated with frailty, functional impairment, disability, reduced quality of life and mortality, in dialysis populations its ability to predict mortality is not well described [9, 22]. In this study, mortality was not predicted in a number of models by the absence or presence of sarcopenia, although severe sarcopenia was likely to influence survival risk in univariate analysis [HR 2.46 (95% CI 1.03–5.90)] (Table 6), and the influence of severe sarcopenia may have been limited by patient numbers. On the other hand, lower baseline MAP and a higher total comorbidity score (which including a number of cardiovascular risk factors) predicted all-cause and cardiovascular mortality in adjusted analyses. This is not surprising, because cardiovascular disease is the leading cause of death in dialysis patients, and lower blood pressure may signify cardiovascular pathology. No other assessed baseline variables were independently associated with mortality in Cox analyses. The phenomenon of traditional risk factors performing poorly in patients with CKD is well recognised, and likely to be caused by the high number of competing risks these vulnerable patients have for mortality [23]. Information on educational level, smoking, alcohol intake and physical activity was not collected at baseline. However, end organ damage resulting from lifestyle factors is reflected in the comorbidity index, which included hypertension, ischaemic heart disease, other vascular disease, respiratory disease, malignancy, and diabetes. Only 16% of males and no females were able to complete 400 m in the 6-minute walk test, indicating that this group had limited exercise capacity prior to study entry.

Several prognostic markers have been recommended for use in patients with CKD. In an earlier study of patients on haemodialysis, those with high BMI and normal or high muscle mass (based on 24-hour urinary creatinine excretion) had a lower hazard ratio for death than patients with a normal BMI [24], however patients with high BMI and low muscle mass did not have improved survival. Similarly, improved survival of maintenance haemodialysis patients has been associated with greater mid-arm muscle circumference, which is a surrogate for lean body mass [25]. On the other hand, another study of haemodialysis patients reported that high fat mass provided a survival advantage in both sexes, whereas a higher lean body mass was only protective in women, [26] and a study of patients commencing dialysis reported a survival advantage for patients with a BMI > 25 kg/m2 and those with higher fat body mass index [27].

Irrespective of BMI, survival is reported to be lower for patients defined as having PEW based on assessment by SGA, and the SGA has been associated with increased mortality in other studies, including over 7 years for patients with a mean age of 59 years on haemodialysis [28] and over 4 years for patients on peritoneal dialysis [29]. A number of modifications of the SGA have been used in patients on dialysis, including a ‘dialysis malnutrition score,’ which was reported to correlate with biochemical parameters associated with malnutrition more closely than the SGA [30]. Nevertheless, the lack of uniformity between versions of the SGA does make it difficult to compare results for nutritional status between studies, and to provide consistent methodology guidance to clinicians [31]. The subjective nutritional assessment score used in this study was adapted from the SGA, but excluded functional capacity, comorbidities, and body composition, because these were quantitated by methods less prone to subjective error.

Other indices found to predict early mortality following commencement of dialysis include age, comorbidities, and recent hospitalisation, but most of these perform only moderately well, and more accurate tools are required [32]. Three-year mortality was predicted in incident dialysis patients using patient demographics, comorbid conditions and laboratory variables with a C statistic of 0.73 (95% CI, 0.71–0.76) [33] and a ‘new comorbidity index’ assigning weights to each of 11 comorbidities has shown good predictive value in older dialysis patients followed for nearly 10 years [34]. However, no model has been accepted into general use [7].

An important question is why our results differ from those of some reported studies showing positive associations between sarcopenia and mortality in patients on dialysis. In a recent meta-analysis, [7] sarcopenia (defined as low muscle mass plus low muscle strength or performance) was reported to predict mortality in 8 studies that included dialysis patients, although four of these were non-significant after adjustment for covariates. The mean age of participants in 3 of the 4 positive studies was \(\le\)61 years, which is younger than the mean age of our cohort, and known cardiovascular disease was not included as a covariate in all positive studies. Two positive studies had fewer patients than the current study. Selection bias may have contributed to our outcome, because not all patients were suitable to undergo BIA or baseline functional testing. However, particularly for older patients, choices to commence dialysis or to proceed down a non-dialysis, supportive care pathway, are generally based on discussion between the patient, family, medical and allied health team members, focussing on the benefits of dialysis to quality of life and survival. Older dialysis patients have therefore undergone extensive filtering, and are likely to be healthier, with a better prognosis for survival, than patients offered management through a non-dialysis, renal supportive care program. For patients in this study, the relatively long period from commencing dialysis until death (median 5.6 years) or end of follow up (median 9.1 years) may reflect such selection bias. If all prospective patients were commenced on dialysis without allocation bias, sarcopenia might have impacted mortality differently. Selection may have even greater impacts as more elderly patients are considered for dialysis. Selection bias may also have influenced our finding that the nutritional assessment score, which excluded independently assessed physical performance, body composition and comorbidities, did not predict mortality. In addition, the nutritional assessment score was not associated with the absence of sarcopenia, or category of sarcopenia at baseline, although lower serum phosphate and lower transferrin, which reflect nutritional status, were associated with baseline sarcopenia. However, neither improved prediction of mortality in Cox models. On the other hand, a lower MAP, and the total comorbidity score, which can be determined from patient records or a simple questionnaire, were predictive.

A variety of diagnostic criteria have been used to define sarcopenia, resulting in inconsistent estimates of its prevalence and impact. However, use of the EWGSOP2 criteria has improved the ability of clinicians to establish a diagnosis. The current focus of these criteria on muscle strength, then muscle quantity or quality and finally performance, represents a shift from earlier definitions based primarily on measurement of appendicular or skeletal muscle mass, or LTI (kg/m2). Recommended cut-points are generally 2 to 2.5 SD below mean reference values derived from meta-analysis of studies recruiting healthy young adults.

Muscle strength can be assessed by several validated techniques, and this study utilised grip strength, while also assessing knee extension and recurrent chair stands. Grip strength is simple and inexpensive, and in the general population is independently associated with poor patient outcomes, including prolonged hospitalization, functional limitations, reduced quality of life and mortality [6]. However, in this study mortality was not predicted when substituting the sarcopenia category with low grip strength.

Whilst MRI and CT are considered the gold standard for measuring muscle mass [6], and DXA is a recommended reference method, BIA is widely accessible, portable, affordable, easy to use, has no radiation and has been validated against other techniques [35]. In addition, BIA is available in many dialysis units, where it is used to assist the evaluation of fluid status. Because estimates of muscle mass differ between BIA machines and reference populations, the EWGSOP2 suggests that raw measures are preferable [6]. This study classified sarcopenic-muscle mass as \(\ge\)2 SD from a gender-matched young reference range. Using that definition, 79.5% of male and 57.6% of female patients could be classified to have a LTI in the sarcopenic range. However, in this study, the LTI was not significantly associated with mortality. This is consistent with a study of 330 incident dialysis patients, with 23% aged >65 years, of whom 20% fulfilled criteria for confirmed sarcopenia [9]. Over a median follow-up of 29 months, low muscle mass alone was not associated with increased mortality, whereas individuals with low muscle strength had increased mortality, irrespective of their muscle mass.

To assess muscle function (physical performance), we used the validated TUG and 6MWT. By EWGSOP2 criteria, and with TUG as the physical performance measure, probable, confirmed, or severe sarcopenia was present in 67.6% of patients in this study, with 70.5% of men and 63.6% of women fulfilling criteria for sarcopenia. TUG did not predict mortality when normal / increased TUG was substituted for sarcopenia category in the Cox analysis.

Strengths of this study include use of the ANZDATA registry, which contains information on all patients receiving renal replacement therapy in Australia, resulting in complete follow-up. We also used robust, validated methods to test muscle strength and function and applied current sarcopenia definitions. Limitations include the observational nature of the study design, potential for residual confounders and a relatively small sample size. Nevertheless, we are not aware of other studies that have assessed the survival of patients on haemodialysis using the 2019 EWGSOP2 sarcopenia criteria, and despite the relatively small numbers, mortality was 65% at 6 years. There may also be demographic differences between our participants and other patients on dialysis that limit the generalizability of our results, because Caucasians made up 75% of participants, and Caucasian reference values were used. Fluid retention can influence body composition calculations using BIA; however, we minimized that risk by assessing body composition following a mid-week dialysis session with patients at their ‘dry weight’. There was also potential for exclusion bias, by excluding patients unable to complete baseline functional testing.