The relationship between protein energy wasting (PEW) and mortality in patients undergoing maintenance haemodialysis (HD) therapy is well documented.1,2 Paradoxical associations between various nutritional markers such as body mass index (BMI) and weight changes with mortality risk in HD patients have been described, showing better outcomes for those subjects with a higher BMI or that gain weight over time.3–7

While hospitalization is an inherently limited outcome measure in etiological and prognostic research, it is a fundamental outcome for health-economy research and for research on the quality of clinical care. The risk of hospitalization increases in a graded manner by progressive CKD stage8 and, for dialysis patients, it has been estimated to account for 40–60% of total healthcare expenditures.9,10 During the first year on dialysis, 77% of patients require hospitalization and spend 8.6% of patient treatment days in hospital.11 Given the mounting epidemics of obesity and the parallel decline in underweight and malnutrition in the dialysis population,12 identification of potentially modifiable hospitalization risk factors is important to improve the quality of clinical care in this population. Both malnutrition13,14 and obesity15–17 are important causes of hospitalization in the community, with considerable economic and social cost implications, particularly for inpatient care.18 Data concerning the association between nutritional parameters and hospitalization risk in HD patients are, however, scarce and mainly focused on the study of albumin levels.11,19–22 We here address the association between BMI and short-term body weight changes with the hospitalization risk of dialysis patients. Given the parallel risk for hospitalization of extreme nutritional categories, we hypothesize that both obesity and underweight associate with higher hospitalization rates.

The present analysis included 6296 patients from COSMOS cohort, after excluding 485 patients (7.1%) who did not have follow-up data and 16 patients further lacking information on BMI. The majority of patients (94.5%) had undergone haemodialysis for more than 3 months. Baseline patient characteristics at study entry and incidence rates of hospitalization according to four BMI categories are displayed in Table 1 and published earlier.6 Both underweight and obese patients had similar incidence rates of hospitalization and higher than normal and overweight patients.

BMI and the risk of hospitalization

During the 3-year follow-up, 3696 patients were hospitalized at least once, and a total of 9731 hospitalizations were registered in the cohort. The analysis of hazards to a first hospitalization considered a total of 17,797 observations through time-dependent analyses (Table 2). Underweight patients had a higher risk of hospitalization than patients with normal weight, while overweight and obese patients presented a non-statistically significant lower risk (model 2, Table 2). In the analysis of annual incidence rates, the association between baseline BMI and IRRs showed similar results (Fig. 1); correction for potential confounding factors increased the IRRs estimates associated with underweight while it decreased IRRs estimates of overweight and obese patients.

Table 2.

Time-dependent associations (every 6 months) between body mass index (BMI) and time to hospitalization in 6296 haemodialysis from COSMOS stratified by BMI categories.

	Unadjusted	Adjusted
BMI <20kg/m2	1.20 (1.09–1.33)	1.22 (1.10–1.35)
BMI 20–24kg/m2	1	1
BMI 25–29kg/m2	0.93 (0.86–1.01)	0.89 (0.82–0.96)
BMI ≥30kg/m2	0.98 (0.89–1.08)	0.93 (0.84–1.02)

Represented are unadjusted and adjusted hazard ratios and 95% confidence intervals for the risk of one hospitalization using time-dependent Cox regression analyses. Adjusted model includes multivariate adjustment for age, sex, smoking (never, former or current), country, type of centre (public or private), dialysis vintage, haemodialysis hours per week and modality.

Fig. 1.

Associations between baseline body mass index (BMI) and annual hospitalization rates, expressed as incidence rate ratios, in 6296 haemodialysis patients from COSMOS. Normo weight individuals were taken as the reference. Model 1 includes multivariate adjustment for age, sex, smoking (never, former or current), country, type of centre (public or private), dialysis vintage, haemodialysis hours per week and modality; Model 2 includes multivariate adjustment of variables included in model 1 plus primary kidney disease, diabetes mellitus and history of cardiovascular disease.

(0.06MB).

Weight changes and the risk of hospitalization

Patients experiencing weight losses within a 6-month period were at higher risk of hospitalization than patients with stable weight. No association was seen in patients gaining weight (Table 3). When patients were stratified by their initial BMI, the risks estimates associated with weight losses were statistically significant and similar in magnitude across categories, with HR ranging from 1.88 and 2.04 (Fig. 2).

Table 3.

Associations between weight changes, time to first hospitalization and risk of hospitalization per year in 6296 haemodialysis patients from COSMOS.

	Time to hospitalization (hazard ratios)		Annual hospitalization rates (incidence rate ratios)
	Unadjusted	Adjusted	Unadjusted	Adjusted
Weight loss	2.10 (1.94–2.26)	2.05 (1.90–2.21)	1.61 (1.52–1.71)	1.52 (1.44–1.61)
Stable weight	1	1	1	1
Weight gain	1.09 (0.99–1.19)	1.07 (0.97–1.17)	0.99 (0.93–1.05)	0.90 (0.84–0.96)

Represented are unadjusted and adjusted hazard ratios and 95% confidence intervals for the risk of one hospitalization using time-dependent Cox regression analyses (every 6 months), and incidence rate ratios and 95% confidence intervals for calculation of annual hospitalization rates using Poisson regression analyses. Weight changes are considered as weight loss (decreases in weight higher that 1%), weight gain (increases in weight higher than 1%) and stable weight (weight changes of ±1%), taken here as the reference category. Adjusted model includes multivariate adjustment for age, sex, smoking (never, former or current), country, type of centre (public or private), dialysis vintage, haemodialysis hours per week and modality.

Fig. 2.

Fully adjusted hazard ratios for the time to hospitalization (hazard ratios) and annual hospitalization rates (incidence rate ratios) according to 6-month weight changes and stratified by body mass index (BMI) categories. Multivariate adjustment controlled for age, sex, smoking (never, former or current), country, type of centre (public or private), dialysis vintage, haemodialysis hours per week and modality primary kidney disease, diabetes mellitus and history of cardiovascular disease.

(0.15MB).

Patients experiencing weight losses during the first 6 months of follow-up had higher annual hospitalization rates (IRR) than individuals with stable body weight. Likewise, patients gaining weight had significantly lower hospitalization rates (Table 3). When analyses were stratified by initial BMI categories, the magnitude of the association between dry weight losses and the hospitalization rates was similar among obese (IRR 1.41 [95% CI 1.22–1.62]), overweight (1.32 [1.19–1.47]) and normal weight (1.60 [1.46–1.75]) patients. However, a much higher risk was observed for underweight patients losing weight, with an IRR of 2.85 [2.33–3.47]) (Figure 2).

Compared with other chronic illnesses, patients with CKD have more comorbidities, the highest annual number of inpatient stays, and the greatest number of hospital days.26 Heart failure and ischaemic heart disease have been suggested as the top causes of hospitalization in this patient population.8,11 Poor nutritional status, as identified by hypoalbuminemia11,19–22 may also contribute to explain such risk. To the best of our knowledge, this is the first study addressing the hospitalization risks associated with BMI and short-term body weight changes in this patient population.

Contrary to our working hypothesis which, along with mortality data,6 envisaged an U-shaped relationship between the BMI and the risk of hospitalization, in COSMOS we found that BMI is inversely associated with the risk of hospitalization. Overweight and obese individuals were less likely to be hospitalized and presented a lower annual incidence rate of hospitalizations than normal weight patients. On the other hand, underweight individuals presented a higher relative risk of hospitalization. Even though a precious instrument to investigate the effect of nutritional status in epidemiological studies, BMI is an imperfect marker of fat mass excess both at a general population level27 and in the dialysis population.28–30 High BMI in part reflects high muscle mass, a body component which is of fundamental importance for determining disability in chronic diseases.31,32 Thus our analysis likely reflects the importance of frailty and sarcopenia in engendering clinical complications eventually leading to hospitalization in long-term dialysis.33,34 Some observational in our analysis may align with that hypothesis, such as the shorter dialysis vintage across higher BMI categories. In recent years, it has become apparent however that muscle quality and strength may be as important (or maybe more) as muscle mass.35,36

Independently of baseline nutritional status, intercurrent diseases leading to hospitalization have a relevant impact upon protein-energy metabolism in muscle mass. Accordingly, we found that short-term body weight losses have a clear-cut relationship with the risk of hospitalization, increasing it by two fold. In addition, in the long term, weight losses associated with approximately 50% higher annual hospitalization rates than those maintaining a stable body weight over time, in line with the hypothesis that sarcopenia and intercurrent PEW may underlie the excess risk for hospitalization in dialysis patients. Although it is possible that short term BMI changes in dialysis patients may in part reflect subclinical volume expansion,37 BMI increases had just a minor, non-significant effect on the risk of hospitalization.

Our findings need to be interpreted considering several strengths and limitations. Strengths include the prospective and careful cohort design (including their random recruitment), the completeness of data and the record of consecutive hospitalizations to allow the analysis of annual incidence rates. Its large sample size and representativeness of the European HD population should also be remarked. As discussed, BMI does feature several weaknesses as a metric of body fat in CKD patients.38,39 We lack information on length of hospital stay, which would have allowed us to make cost-effectiveness analysis, and we have unsatisfactory collection on hospitalization causes which limits our capacity to define them. A sensitivity analysis grouping countries according to global background hospitalization rates would have been preferred but was not feasible. In addition, we lack information on the intentionality of these body weight gains or losses. Given the short time frame studied (6-month periods) and the absence of consensus guidelines on weight loss programs for dialysis patients, we are assuming them to be unintentional. Other limitations are the ancillary nature of our findings, as well as the lack of information on additional surrogates of muscle and fat stores.40 As in all observational cohort studies, residual confounding by unmeasured or unknown confounders may have played a role. The predominantly prevalent nature of the HD patients included in COSMOS makes it vulnerable to survivor bias despite the careful adjustment for dialysis vintage.

We conclude that both underweight (lower than 20kg/m2 in this analysis) and short-term body weight losses associate with increased hospitalizations rates, likely because it signals an unstable clinical situation including, but not limited to, an unfavourable nutritional status. This, and our previous report on hard endpoints,6 altogether emphasize the importance of nutritional status on HD outcomes and bring special attention to the importance of the short-term weight losses in the underweight patients, a finding that deserves additional surveillance. The high and rising health cost absorbed by the dialysis population is a reason of concern for health authorities and providers. Since hospitalization has a major share in this cost, understanding the implication of nutritional disorders for this outcome may help to set strategies aimed at limiting hospital admission in this population. Cost-effectiveness studies indicate that, along with improved wellbeing, substantial savings in health costs can be obtained by protein and energy supplementation in treatment or prevention of malnutrition in the general population.41 Analyses modelling the effect of such intervention on hospitalization in the dialysis population suggest that the potential savings in hospitalization, treatment costs and mortality may be substantial.42,43 Our observations provide further strength to the quest of performing clinical trials aimed at assessing whether nutritional interventions targeting body weight may improve clinical outcomes and reduce hospitalization in this high risk population.

COSMOS is sponsored by the Bone and Mineral Research Unit (Hospital Universitario Central de Asturias), SAFIM (Sociedad Asturiana Fomento Investigaciones Óseas), the European Renal Association-European Dialysis and Transplant Association, the National Program of I+D+I 2008–2011 and Instituto de Salud Carlos III (ISCIII), the ISCIII Retic REDinREN (RD06/0016/1013 and RD12/0021/1023), the ISCIII (ICI14/00107 and PI17/00384), Fondo Europeo de Desarrollo Regional (FEDER), Plan Estatal de I+D+I 2013–2016, Plan de Ciencia, Tecnología e Innovación 2013–2017 del Principado de Asturias (GRUPIN14-028) and Fundación Renal Íñigo Álvarez de Toledo (FRIAT). Logistics (meetings, secretarial help, printing of materials, development of Web site for data entry, etc.) have been financially supported by AMGEN Europe, Fundación Renal Íñigo Álvarez de Toledo (FRIAT) and the Spanish Society of Nephrology (Estudio Estratégico de la SEN). ICR research stay at Karolinska Institutet was supported by the Río Hortega program, Instituto Carlos III, Spain. The authors are not aware of any additional relationships, funding, or financial holding that might be perceived as affecting the objectivity of this study.

The authors declare no conflict of interest.