Colin H Jones MB ChB, MD, FRCP

Renal Unit, York Hospital, York, UK.

Serum albumin is much discussed by doctors treating patients with renal disease. A Medline search combining serum albumin and peritoneal dialysis yields nearly 400 citations. If you include haemodialysis, renal transplantation and other kidney diseases there are many, many more. Why is there such an interest in serum albumin in PD patients?

Firstly, albumin is both easy and cheap to measure, available with routine biochemical analysis. Secondly, most renal databases contain albumin results over a long time period, facilitating its use in database analysis. Most importantly serum albumin concentration is a powerful predictor of outcome in peritoneal dialysis patients. Early studies in PD were small. Teehan found that low serum albumin predicted an increased number of hospital days and was the most powerful predictor of mortality, as compared to age, time on dialysis and small solute clearance[1]. Avram reported that the serum albumin at the start of his study (not the start of PD treatment) was significantly lower in non-survivors than survivors (31 vs. 36 g/l) and was an independent risk factor for mortality[2]. When patients were grouped according to serum albumin (> 37, 32-37, <32 g/l), Kaplan-Meier survival analysis demonstrated progressive worsening of outcome with lower serum albumin (400 day survival of 93, 80 and 64% respectively). In a Cox proportional-hazard model, serum albumin just prior to commencing CAPD was the most powerful predictor of probability of death, as compared to age, low serum creatinine or high serum cholesterol[3]. In a retrospective study of 225 patients Jones found that both patient and technique survival were lower in patients with a mean serum albumin over the first 6 months of treatment below the normal range[4]. Of note the increased mortality and decreased technique survival were predominantly seen in subjects with a decreasing serum albumin over the first year of treatment rather than in those with a reduced but stable serum albumin.

The Canada-USA Peritoneal Dialysis Study Group was a 2 year prospective study of 680 new CAPD patients[5]. Each 1 g/l decrease in serum albumin was associated with both a 6% increase in the relative risk of death and a 5% increase in the relative risk of technique failure.

These findings raise the question of why serum albumin predicts outcome? Most of the early literature focused on the relationship between serum albumin and malnutrition. In renal failure serum albumin has frequently been used as a nutritional marker and hypoalbuminaemia as a synonym for malnutrition. What is the evidence that serum albumin is related to nutritional state?

Serum albumin is correlated with dietary protein intake in patients with renal disease but normal renal function, although this correlation becomes less strong as renal function deteriorates and renal replacement therapy is required [6]. Serum albumin correlates with dietary protein intake in some studies [7]. The protein catabolic rate normalised to actual body weight (nPCR) is correlated with dietary protein intake [8],[9],[10] and nPCR is correlated with serum albumin in some studies[11]. Pollock found a correlation coefficient of 0.28 in 134 patients, Lindsay and Spanner [12] a coefficient of 0.5 from 70 estimates of non-normalised PCR in 20 patients, and Nolph [13] a correlation coefficient of 0.4 in 68 patients. Others studies have not confirmed this relationship [1, 14-17].

Many other techniques for determining and monitoring nutritional state have been described in PD. Actual body weight is the most commonly used nutritional parameter in clinical practice. Subjective global assessment (based on a clinical history of appetite, food intake, weight change, and metabolic demand, and a clinical inspection of muscle bulk and subcutaneous fat) has been validated in PD[18]. Anthropometric measures (skinfold thickness and derived lean body mass), bioelectric impedance techniques, dual energy X-ray adsorptometry (DEXA) scanning, total body potassium, total body nitrogen, and lean body mass derived from creatinine kinetics have all been investigated in PD.

Actual body weight as a percent of ideal was not related to serum albumin in a cross-sectional study of 546 general hospital patients [19]. In PD patients there was no relationship between the difference in ideal and actual body weight and serum albumin [20]. Serum albumin has been reported to be lower in female, but not male, PD patients identified as malnourished by SGA [18] and lower in males and females on both haemo- and peritoneal dialysis identified as malnourished [21]. SGA score identified PD patients with significant abnormalities of body mass index, lean body mass (determined by anthropometry, bioimpedance and creatinine kinetics), midarm muscle circumference, and forearm muscle strength in males and females, in addition to back strength and fat mass in females only[22]. None of these parameters differed when comparison was made between patient groups defined by a serum albumin above or below the median value. In the same study, 48% of hypoalbuminaemic males (61% of females) were scored as well nourished and 41% of normoalbuminaemic males (62% of females) as malnourished. Pollock [23] reported significant correlations between serum albumin and a number of nutritional parameters, including body mass index, midarm circumference, midarm muscle circumference, and triceps skinfold thickness in 134 patients. Lean body mass, estimated from creatinine appearance, expressed as a percentage of total body weight has been correlated with serum albumin in 68 patients [13]. Serum albumin was not correlated with lean tissue measured by DEXA scanning, in either PD or HD patients [24]. There are no consistent data on the relationship between serum albumin and either total body nitrogen or total body potassium in dialysis patients. Using a composite nutritional score that included SGA, percentile ranking of ideal body weight, dry weight, body mass index, midarm circumference and triceps skin fold thickness, and serum albumin, albumin was the only variable that was not correlated with severity of malnutrition [25].

These studies show that the relationship between serum albumin and physical measures of nutrition is inconsistent. It can be concluded that serum albumin is not a reliable measure of nutritional state.

Albumin is a 60 kilodalton protein synthesized by the liver. Albumin metabolism is sensitive to nutritional intake with both synthesis and catabolism decreasing during protein malnutrition [26]. In both acute and chronic infective and inflammatory disorders, hepatic metabolism is switched towards the production of acute phase proteins, including alpha-2 macroglobulin, ferritin and C-reactive protein, and the synthesis of 'negative' acute phase proteins (including albumin) decreases. Redistribution of both albumin and water between the intravascular and extravascular space may alter serum albumin concentration in a number of pathological disorders, including cardiac failure, the systemic inflammatory response syndrome and nephrotic syndrome. Extracorporeal albumin loss occurs in the urine in nephrotic syndrome. Thus the serum albumin concentration is a function of synthesis, catabolism, distribution and extracorporeal loss.

Kaysen and Schoenfeld [27] investigated albumin kinetics in 9 CAPD patients and 5 control subjects with normal renal function. The mean serum albumin was lower in PD patients (37 vs. 46 g/dl in controls). Mean albumin and total protein losses in urine and dialysate were 4.23 and 8.79 g/1.73m2/day in patients. The total body albumin mass was not significantly different between groups, but albumin catabolism was significantly reduced in the patients and albumin synthesis was increased in two patients. The increase in albumin synthesis correlated with the extent of the extracorporeal protein loss. In these patients therefore serum albumin was maintained, all be it at a lower serum concentration, by a compensatory increase in synthesis and decrease in catabolism.

Chronic glomerulonephritis and diabetic nephropathy are the commonest causes of end-stage renal failure requiring renal replacement therapy in Western countries, and both are characterised by proteinuria, frequently in the nephrotic range. This may result in hypoalbuminaemia at the start of dialysis, and a persistent significant urinary protein loss may continue into dialysis. In clinical practice a sudden increase in serum albumin is often seen in diabetic PD patients when they lose their residual renal function. An increased urinary protein loss has been reported in PD patients with a low serum albumin [28].

Proteins and free amino acids are also lost across the peritoneum, and total protein losses may be considerably greater than in haemodialysis. A strong correlation between serum albumin and peritoneal protein loss has been reported in 18 patients (Schoenfield 1992) and a weak correlation in 134 patients [23]. This has not been confirmed in a number of other studies [2, 17, 29, 30]. The utilisation of amino acid-based dialysate can reverse this net loss [31].

The dialysate to plasma (D/P) creatinine ratio is inversely correlated with serum albumin [14, 32]. High or high-average transport status is an independent predictor of serum albumin [33]. This has been interpreted as indicating an increased transmembrane permeability to protein and amino acids and hence an increased loss. However there are other possible explanations of this association. Increased glucose absorption in 'high transporters' may suppress appetite. Non-specific peritoneal inflammation could both increase peritoneal permeability and suppress serum albumin via an acute phase response. A higher D/P ratio is also characteristic of decreased net ultrafiltration volume. Patients grouped as high, high average, low average and low transporters had serum albumin concentrations of 32.8, 33.8, 36.2 and 37.8 g/l respectively and mean 24-hour drained dialysate volumes of 8.22, 8.59, 8.93 and 9.38 litres respectively [34]. High transporters have a higher blood pressure [35] and blood pressure in PD patients is at least partly dependent on salt and water retention. This raises the possibility that this association may be due to fluid excess and resultant dilutional hypoalbuminaemia.

End-stage renal failure results in salt and water retention and a predisposition to fluid overload. Retained salt and water expand the extracellular space, including both the intravascular volume and (predominantly) the interstitial space. When fluid is removed during haemodialysis haemoconcentration occurs and albumin increases in the same way that haematocrit increases. In a random sample of 49 chronic haemodialysis patients undergoing a net fluid removal of 2.0±1.2 kg the serum albumin increased from 36.9 to 41.4 g/l in a laboratory with a reference range for serum albumin of 35–50 g/l [36]. The increase in serum albumin over a dialysis session correlated both with the actual amount of fluid removed and the decrease in extracellular water estimated by bioelectric impedance. Serum albumin, both pre and post dialysis, correlated with extracellular fluid volume normalised to total body water and this relationship was independent of the significant correlation between serum albumin and C-reactive protein. Serum albumin in my units PD patients is in the same range as our HD patients’ predialysis levels.

An increased plasma volume (50.6 ± 20.9 vs. 36.8 ± 3.6 ml/kg) with no change in total albumin mass has been reported in PD patients compared to controls, although this finding did not reach statistical significance [27]. The extracellular fluid volume as a percentage of total body water (determined by multifrequency bioelectric impedance) was increased in hypoalbuminaemic as compared to normoalbuminaemic PD patients [28]. PD patients have a greater requirement for antihypertensive medication than disease, sex, and age matched controls on HD [37]. They have peripheral oedema more frequently than HD patients (30% vs. 12%), as well as a lower serum albumin, despite greater weight and equivalent midarm muscle circumference [38]. There is therefore considerable evidence that PD patients have greater fluid overload than HD patients and that the degree of fluid overload is associated with the serum albumin concentration.

Serum albumin decreases during any infective or inflammatory episode. In stable haemodialysis patients, hypoalbuminaemia was associated with decreased albumin synthesis, decreased serum insulin-like growth factor-1, and increased serum C-reactive protein, 2-macroglobulin and ferritin concentrations [39]. There was no relationship between serum albumin and dietary protein intake or indices of body composition (skinfold thickness and midarm muscle circumference). Acute inflammatory response was considered the major determinant of serum albumin. A similar conclusion has also been reported in PD [40]; in this study an acute phase response, measured by C-reactive protein or serum amyloid A, was significantly correlated with serum albumin.

Serum albumin is also significantly lower in diabetics [14] and in PD patients with systemic disease (defined as diabetic nephropathy, hypertensive nephrosclerosis, lupus nephritis, amyloidosis, and scleroderma) [20]. Age independently predicts serum albumin, demonstrating an inverse relationship [14], and serum albumin is lower in older dialysis patients [41].

Does a low serum albumin cause an increased mortality? This is unlikely to be the case. But why is serum albumin such a strong predictor of poor outcome? Protein calorie malnutrition is unlikely to be the explanation. Fewer than 5% of deaths reported to the EDTA registry are caused by cachexia.

A low serum albumin is associated with both identifiable and more occult disease processes. In a prospective study of 61 new PD patients, survival was lower in patients with a serum albumin below the median value compared to those with a serum albumin higher than the median (49 vs. 79% at 2 years) [20]. However in a Cox proportional-hazards model, serum albumin ceased to be an independent risk factor for increased mortality once the presence of a systemic disease, defined as diabetes mellitus, hypertensive nephrosclerosis, lupus nephritis, amyloidosis and scleroderma was taken into account. Similar conclusions have been reported in two further studies, one of 97 UK and the other of 201 French PD patients. In both studies albumin only predicted survival in univariate, but not multivariate, analysis. In the UK study [42] co-morbid disease and age were significant determinants of survival in multivariate analysis. Left ventricular dysfunction and vascular disease were particularly important predictors of outcome. In the French study solute clearance, age, and the presence of diabetes, cardiovascular disease (cardiac, cerebrovascular or peripheral vascular disease), or other co-morbid conditions, including malignancy, multisystem disease, infection, respiratory failure, and liver cirrhosis, were predictors of outcome in multivariate analysis [43]. In some patients the source of the inflammatory response is unclear. Recently a strong relationship between inflammation, malnutrition and atherosclerosis has been reported [44].

A low serum albumin is associated with an increased risk of developing left ventricular dilatation and de novo or recurrent heart failure in ESRF [45]. In a database analysis of both HD and PD patients a low serum albumin was associated with an increased incidence of congestive heart failure with each 1 g/dl reduction in serum albumin carrying an odds ratio of 1.35 [46]. In PD, chronic hypervolaemia is associated with an increased left ventricular mass [47]. A high D/P ratio is associated with a decreased probability of patient and technique survival at 2 years from start of dialysis (48%, 52%, 61% and 86% for high, high average, low average and low transporters respectively) [34]. Patients with a higher D/P creatinine ratio have a lower serum albumin, poorer ultrafiltration and worse survival despite better small solute clearances. Salt and water overload may therefore cause both dilutional hypoalbuminaemia and precipitate heart failure and its associated excess mortality.

The first step in managing a low serum albumin in a PD patient is to identify the cause. A patient work up might include a full nutritional assessment (independent of biochemical markers), a 24 hour urine and dialysate collection for protein losses, a PET test for D/P creatinine ratio, C reactive protein and full physical examination. Even if evidence of malnutrition is present other causes of a low albumin should be excluded, as the relationship between serum albumin and nutrition is poor.

If the patient is actually still nephrotic, which is often the case in type 1 diabetics starting dialysis, then serum albumin will improve as renal function declines (although this is in itself undesirable). The finding of a high CRP should lead to an exhaustive search for its cause. CRPs at the high end of the normal range may be associated with either occult or overt atherosclerotic disease. Salt and water overload should be suspected in all PD patients even in the absence of oedema. This is particularly the case in patients with hypertension even if controlled with antihypertensive medication. Ultrasound measurement of vena caval diameter or bio-electric impedance may help to identify unsuspected fluid excess.

The next step is to embark on measures that will improve serum albumin. Unfortunately there are no good controlled studies demonstrating that any therapeutic manoeuvre increases serum albumin. Nutritional interventions have included oral supplementation, enteral feeding, administration of intraperitoneal amino acids and the administration of anabolic agents including growth hormone, insulin-like growth factor-1, anabolic steroids and clenbuterol. None have demonstrated a consistent effect on serum albumin, although overt underfeeding clearly needs to be addressed. Increasing dialysis dose may increase nutritional intake. Davies [48] identified 48 malnourished PD patients from a cohort of 153. These patients had a decreasing body weight and midarm muscle circumference, a reduced appetite for calories and protein and a low serum albumin. Dialysis prescription was altered to increase peritoneal KT/V urea by 22.5% and total KT/V urea by 18%. Both oral and peritoneal dialysis derived calorie intake increased. This was accompanied by stabilisation of body weight and midarm circumference and an increase in albumin. Whether this was due to improved nutrition or due to enhanced salt and water excretion needs clarification.

If a definite inflammatory process can be identified then this needs appropriate treatment. There are no data to indicate whether anti-inflammatory agents such as thalidomide have any role in managing chronic non-specific inflammation. Whether statins would have an effect on the high normal range CRPs in patients with atherosclerotic disease is not yet clear. There is certainly no data to show that these approaches would increase serum albumin. Optimising salt and water balance may be important. In an uncontrolled study of 21 patients a mean increase of 290 mls/day in the achieved ultrafiltrate volume resulted in significant decreases in body weight, extracellular fluid volume, systolic and diastolic blood pressure and the number of anti-hypertensive drugs required. Serum albumin increased (34.6 ± 4.1 to 35.9 ± 3.6 g/l) in parallel with these changes [49].

In conclusion hypoalbuminaemia is associated with a poor outcome in PD patients. The causes of a low serum albumin are multifactorial and the underlying cause accounts for the increased mortality. While malnutrition is frequent in PD, it is rarely the sole cause of a low serum albumin and other factors should be actively sought and corrected where possible.