SEARCH

SEARCH BY CITATION

Keywords:

  • apoB/apoA-I ratio;
  • chronic kidney disease;
  • dialysis;
  • mortality;
  • protein-energy wasting;
  • reverse epidemiology

Abstract.

  1. Top of page
  2. Abstract.
  3. Introduction
  4. Patients and methods
  5. Results
  6. Discussion
  7. Conflicts of interest
  8. Acknowledgements
  9. References

Background.  In the general population, a high apoB/apoA-I ratio is a strong risk factor for cardiovascular disease and mortality. However, whether this is the case in chronic kidney disease (CKD) patients is currently unknown.

Study design.  The apoB/apoA-I ratio was evaluated in 391 incident CKD stage 5 patients examined close to dialysis initiation, and again after 1 year of dialysis in a subgroup of 182 patients, subsequently followed for up to 3 years.

Results.  Baseline values of the apoB/apoA-I ratio as well as changes in the ratio during the first year of dialysis correlated with body mass index (BMI) and fat mass. The baseline apoB/apoA-I ratio showed no association with 4-year mortality. However, after adjustment for confounders, a high apoB/apoA-I ratio (>0.9) predicted short-term (first year) survival [hazard ratio (HR): 0.35; 95% confidence interval (CI): 0.13–0.85)] and long-term (next 3 years) mortality (HR: 1.72; 95% CI: 1.01–2.96). An increase in the apoB/apoA-I ratio during the first year of dialysis was linked to a survival advantage thereafter (HR: 0.48; 95% CI: 0.22–0.98). However, this association lost its significance (HR: 0.62; 95% CI: 0.26–1.36) after adjustment for indices of protein-energy wasting.

Conclusions.  A high apoB/apoA-I ratio and an increase in this ratio during the first year on dialysis were associated with short-term survival advantage in CKD patients. This paradoxical relationship represents an example of the so-called reverse epidemiology phenomenon in CKD patients and suggests that the apoB/apoA-I ratio should always be interpreted with caution in this patient population.


Introduction

  1. Top of page
  2. Abstract.
  3. Introduction
  4. Patients and methods
  5. Results
  6. Discussion
  7. Conflicts of interest
  8. Acknowledgements
  9. References

Patients with advanced chronic kidney disease (CKD) have a very poor prognosis, mainly accounted for by cardiovascular disease (CVD) and infectious complications [1]. In contrast to the general population, cholesterol appears to show no association with mortality in CKD or even a paradoxical relationship, referred to as a ‘reverse epidemiology’ phenomenon [2–5]. Potential explanations for this apparent reversed epidemiology include time differential of competing risks [6], ability of lipoproteins to bind bacterial endotoxins and modulate inflammatory immune response [7] and/or confounding by protein-energy wasting (PEW) conditions [4].

Apolipoproteins (apo) B and A-I are critical in lipoprotein formation, stability and clearance [8]. ApoB is mainly present on very low-density lipoproteins (VLDL), intermediate density lipoproteins (IDL) and low-density lipoproteins (LDL). Therefore, its measurement reflects the number of potentially atherogenic lipoprotein particles [9]. ApoA-I is the major apolipoprotein of HDL particles. It acts as a structural protein and mediates reverse cholesterol transport. It is also involved in anti-inflammatory and anti-oxidant properties of HDL [10]. Thus, measurement of apoA-I reflects the level of potentially anti-atherogenic lipoproteins.

The ratio of these two apolipoproteins (apoB/apoA-I) is a reflection of the balance between atherogenic and anti-atherogenic lipoprotein particles, and is regarded as a strong lipid risk factor for predicting CVD and mortality in the general population. In fact, several studies have shown that the apoB/apoA-I ratio is superior to the predictive power of total cholesterol, LDL cholesterol, total/HDL cholesterol ratio or triglycerides [9, 11–14].

In the CKD population, measurement of apolipoproteins could present some advantages over traditional lipid risk factors. First, it is a direct measurement, in contrast to LDL cholesterol (derived from the Friedewald formula which cannot be used in case of hypertriglyceridaemia, a common feature in advanced CKD) [15]; second, apoB gives a better insight into the number of small dense LDL particles which are believed to be highly atherogenic [16] and often increased in CKD [15]; finally, fasting conditions are not needed for apolipoprotein measurement, being an important advantage especially for diabetic CKD patients [9].

Whether the apoB/apoA-I ratio is a valid risk factor for mortality in CKD patients or whether it is affected by the so-called reverse epidemiology phenomenon in CKD is currently unknown. Therefore, the aim of the present study is to assess the potential prognostic value of the apoB/apoA-I ratio as a predictor of mortality in a large cohort of carefully phenotyped incident dialysis patients.

Patients and methods

  1. Top of page
  2. Abstract.
  3. Introduction
  4. Patients and methods
  5. Results
  6. Discussion
  7. Conflicts of interest
  8. Acknowledgements
  9. References

Patients

Three hundred and ninety-one CKD stage 5 patients were studied close to the start of dialysis therapy. The present evaluation was based on an ongoing prospective study and part of the data had already been reported [17]. The study protocol was approved by the Ethics Committee of Karolinska University Hospital Huddinge, Stockholm, Sweden, and informed consent was obtained from each patient.

The patients comprised 243 males (62%) and 148 females (38%) with a median age of 55 years (range 19–70 years), who were enroled at initiation of dialysis therapy in the renal program of the Karolinska University Hospital Huddinge (Sweden) between 1994 and 2007. Exclusion criteria were subjects aged younger than 18 years or older than 70 years, acute infectious complications, or unwillingness to participate in this study. The causes of CKD were diabetic nephropathy in 121 patients (31%), chronic glomerulonephritis in 102 patients (26%), polycystic kidney disease in 43 patients (11%) and other or unknown causes in 125 patients (32%). Diabetes mellitus was present in 32% of the patients, and CVD defined as a clinical history of cerebrovascular, cardiovascular and/or peripheral vascular disease, was present in 37% of the subjects. The majority of patients was on antihypertensive medications: angiotensin-converting enzyme inhibitors and/or angiotensin II receptor antagonists, n = 213 (54%); betablockers, n = 226 (58%); calcium-channel blockers, n = 170 (43%) and other commonly used drugs in end-stage renal disease, such as phosphate and potassium binders, diuretics, erythropoiesis-stimulating agents, iron substitution and vitamin B, C and D supplementation. Only 83 (21%) patients were on lipid-lowering medication (HMG-CoA-reductase inhibitors) at the start of dialysis therapy.

One hundred and eighty-two patients completed a second evaluation after about 12 months (± 3 months) of dialysis therapy and were included in the longitudinal analyses. The median time between basal and follow-up visit was 12.2 months (interquartile range: 11.7–13.4). Reasons for dropout before the second assessment included death, transplantation, transfer to another hospital, or unwillingness to participate in the follow-up. Out of the 182 patients who completed the 12-month evaluation, 96 (53%) were treated by peritoneal dialysis (PD), whereas 86 (47%) patients were on haemodialysis (HD) therapy. HD patients were treated by conventional bicarbonate-buffered dialysate using polyamide, cellulose acetate, or haemophane filters, whereas PD patients were on a four to five exchanges per day schedule with standard dialysis bags. In accordance with current therapy recommendations, diabetic patients initially taking oral antiglycaemic agents or on restricted diets had been switched to insulin therapy at the time of inclusion in the study.

Laboratory analyses and anthropometric evaluations

Venous blood samples were drawn after an overnight fast and stored at −70 °C for biochemical analyses. Levels of apoA and apoB were determined using an immunonephelometric procedure (Behring AG, Marburg, Germany) and high-sensitive C-reactive protein (hsCRP) was measured by nephelometry. All other biochemical analyses were performed using routine methods at the Department of Clinical Chemistry at Huddinge Hospital. Body mass index (BMI) was calculated as weight in kg per (height in m)2. Fat body mass and lean body mass were evaluated by using dual-energy X-ray absorptiometry (DEXA) (Lunar Corp, Madison, WI, USA). Subjective global assessment (SGA) was used to evaluate signs of PEW, and patients presenting with an SGA score of 2–4 were defined as wasted [18]. This resulted in a diagnosis of wasting in 31% of the patients. Glomerular filtration rate (GFR) was estimated as the mean of creatinine and urea clearance from a 24-h collection of urine and corrected for body surface area.

Statistical analysis

All values are expressed as mean and SD, median and interquartile range, unless otherwise indicated. Comparisons of continuous variables between two groups were assessed by using Student’s unpaired t-test or Mann–Whitney test, as appropriate. As many values were not normally distributed, Spearman’s rank correlation was used to determine correlations between values of baseline apoB/apoA-I ratio or the ratio changes during the first year of dialysis therapy and other variables. Survival analyses were made with the Kaplan–Meier survival curve or Cox proportional hazards model. Survival was determined from the day of baseline examination, and for the analysis of apoB/apoA-I ratio changes from the day of the second visit. Patients were censored at transplantation or when completing the follow-up period with no loss of follow-up of any patient. Hazard ratios (HRs) for mortality were determined by using univariate and multivariate Cox regression analysis and presented as HR and 95% confidence interval (CI). In order to discern high and low apoB/apoA-I ratio groups, a cut-off point of 0.9 was used, as suggested previously [19–21]. In the general population, values of apoB/apoA-I ratio were typically higher in men, and therefore a different cut-off point for men and women has also been proposed [19]. However, in our cohort, apoB/apoA-I ratio did not differ between genders. Therefore, applying different cut-off points for men and women would result in significantly more women belonging to a ‘high ratio’ group. For evaluating the influence of duration of follow-up on associations between the apoB/apoA-I ratio and mortality, time-stratified effects of baseline apoB/apoA-I ratio on mortality were assessed [22]. In this analysis, HR for mortality during the first year of dialysis therapy was compared with the HR for mortality during the next 3 years conditional upon having survived the previous period. Therefore, for the long-term survival analysis we considered only these patients who survived the first year of dialysis treatment. P-value less than 0.05 was considered statistically significant. Statistical analyses were performed using statistical software sas, version 9.1 (SAS Institute, Cary, NC, USA).

Results

  1. Top of page
  2. Abstract.
  3. Introduction
  4. Patients and methods
  5. Results
  6. Discussion
  7. Conflicts of interest
  8. Acknowledgements
  9. References

Baseline characteristics and univariate correlates

The apoB/apoA-I ratio values did not follow a normal distribution, showing a median of 0.78 (0.58–1.00), similar for men (0.78 [0.60–1.00]) and women (0.77 [0.56–0.99]). Moreover, the apoB/apoA-I ratio did not differ with regard to the presence of diabetes mellitus or clinical history of CVD (data not shown). Baseline characteristics of the investigated patients divided according to an apoB/A-I ratio cut-off point of 0.9 are listed in Table 1. As seen in the table, BMI was lower and statin use higher amongst subjects with a lower ratio.

Table 1.   Characteristics of the incident dialysis patients included in the study subdivided according to a low and high apoB/apoA-I ratio
 ApoB/apoA-I ratioP-value
Low ≤ 0.9 (n = 251)High > 0.9 (n = 140)
  1. NS, not significant; BMI, body mass index; SGA, subjective global assessment; hsCRP, high sensitivity C-reactive protein; GFR, glomerular filtration rate.

Age (years)55 (43–63)55 (46–64)NS
Male gender (%)62%63%NS
Diabetes mellitus (%)33%29%NS
Cardiovascular disease (%)37%36%NS
BMI (kg m−2)24.3 ± 4.325.6 ± 4.3<0.01
Total fat mass (kg)19.9 (13.7–27.0)22.2 (16–28.3)NS
Lean body mass (kg)48.8 (41.6–56.4)49.8 (40.0–58.4)NS
Wasting (SGA > 1) (%)30%33%NS
S-albumin (g L−1)33.3 ± 6.233.2 ± 6.0NS
hsCRP (mg L−1)4.4 (1.8–13.0)6.7 (2.3–16.0)NS
GFR (ml min−1)6.5 ± 2.36.5 ± 2.4NS
Statin use (%)28%10%<0.0001
Total cholesterol (mmol L−1)4.5 (3.8–5.5)6.1 (5.1–7.1)<0.0001
HDL cholesterol (mmol L−1)1.4 (1.1–1.7)1.0 (0.8–1.2)<0.0001
Triglycerides (mmol L−1)1.5 (1.1–2.0)2.4 (1.8–3.2)<0.0001
Lipoprotein (a) (mg L−1)254 (102–562)232 (87–523)NS
ApoB (g L−1)0.82 (0.68–1.03)1.29 (1.09–1.53)<0.0001
ApoA (g L−1)1.35 (1.17–1.56)1.13 (0.98–1.30)<0.0001
ApoB/apoA-I ratio0.63 ± 0.171.21 ± 0.35<0.0001

Univariate correlations using Spearman Rank test between the apoB/apoA-I ratio and relevant parameters are listed in Table 2. A positive correlation was found for BMI and fat body mass as well as for hsCRP. The association with hsCRP was due to a correlation between apoA-I and hsCRP (Spearman rho = −0.28; P < 0.001), rather than between apoB and hsCRP (Spearman rho = 0.10; P = NS). No associations were observed between apoB/apoA-I ratio and age, S-albumin levels, and total body mass or GFR.

Table 2.   Spearman’s rank correlation between apoB/apoA-I ratio values and selected variables with CKD stage 5 patients starting with dialysis (n = 391)
VariableSpearman rhoP-value
  1. GFR, glomerular filtration rate; BMI, body mass index; hsCRP, high sensitivity C-reactive protein; CKD, chronic kidney disease.

Age0.026NS
GFR0.020NS
BMI0.193<0.001
Total fat mass0.135<0.05
Lean body mass0.052NS
S-albumin−0.062NS
hsCRP0.186<0.001

Baseline survival analysis

The patients were followed up for up to 4 years after commencement of dialysis therapy, with a median observation time of 23 months. During the observation period, 107 patients died, 68 of them due to CVD. Survival Kaplan–Meier analysis for all-cause mortality was performed according to an apoB/apoA-I ratio cut-off point of 0.9, and no differences between the two groups were observed (HR: 1.12; 95% CI: 0.75–1.64) (Table 3).

Table 3.   Crude and adjusted overall and time-stratified hazard ratios (and 95% confidence intervals) of a high baseline apoB/apoA-I ratio on all-cause mortality (n = 391)
ModelFour yearsPFirst year of dialysisPNext three years of dialysisP
  1. DM, diabetes mellitus; BMI, body mass index; SGA, subjective global assessment; hsCRP, high sensitivity C-reactive protein; NS, not significant.

Crude1.12 (0.75–1.64)NS0.35 (0.13–0.78)<0.011.74 (1.09–2.80)<0.05
Adjusted for age, gender, DM and statin use1.01 (0.69–1.51)NS0.33 (0.12–0.75)<0.011.61 (1.01–2.59)<0.05
Adjusted for above + BMI, SGA and hsCRP1.07 (0.71–1.64)NS0.35 (0.13–0.85)<0.051.72 (1.01–2.96)<0.05

We used conditional time-stratified relative risk analysis in order to investigate the short-term and the long-term prognostic use of the apoB/apoA-I ratio. We first tested survival during the first year of dialysis (short-term prognosis) and compared it with the following 3 years of dialysis (long-term prognosis) in subjects who survived the first year. In the first year of dialysis therapy, 35 patients died, 26 of them due to CVD. During this period, a higher apoB/apoA-I ratio was linked to a survival advantage with a HR of 0.35 and 95% CI: 0.13–0.78 (Table 3). However, for subjects surviving the first year, the survival during the next 3 years of dialysis, when 72 patients died, 42 due to CVD, showed an opposite relationship, i.e. a higher apoB/apoA-I ratio was associated with an increased mortality, (HR: 1.74; 95% CI: 1.09–2.80) (Table 3). As seen in the table, these results remained essentially unchanged after adjustments for potential confounders. Because a higher apoB/apoA-I ratio had a larger proportion of patients on statins (Table 1), all analyses were repeated excluding statin users, showing essentially the same results (not shown).

Longitudinal changes of ApoB/A-1 ratio

As described above, 182 patients (47%) were re-examined after approximately 12 months of dialysis therapy. A comparison between this group of patients and those who were not re-examined showed no differences with regards to baseline apoB/apoA-I ratio, age, gender, presence of diabetes mellitus, CVD and wasting or inflammation (data not shown).

The median apoB/apoA-I ratio after 12 months of dialysis therapy was 0.83 (0.60–1.02). In univariate analysis, significant associations between changes in the apoB/apoA-I ratio and changes in BMI, total fat mass and hsCRP were observed between the baseline and the second visit (Table 4). Again, association with changes in hsCRP were due to the correlation between changes in apoA-I and changes in hsCRP (Spearman rho = −0.25; P < 0.005). There was no association between changes in apoB and changes in hsCRP.

Table 4.   Spearman’s rank correlation between changes in apoB/apoA-I ratio values during the first year of dialysis therapy and changes (Δ) in measurements of body composition and inflammation (n = 182)
VariableSpearman’s rhoP-value
  1. BMI, body mass index; hsCRP, high sensitivity C-reactive protein; NS, not significant.

Δ BMI0.177<0.05
Δ Total fat mass0.410<0.001
Δ Lean body mass−0.065NS
Δ hsCRP0.155<0.05

Finally, although the number of events was rather low, we conducted a 3-year exploratory survival analysis, starting from the second visit onwards. The apoB/apoA-I ratio increased between the two visits in 46% and decreased in 54% of the patients. An increase in the apoB/apoA-I ratio was associated with an improved survival [Log-Rank (χ2): 6.81; P < 0.01] with an unadjusted HR for mortality of 0.45 (95% CI: 0.24–0.82) (Table 5). This remained unchanged after adjustment for age, gender, dialysis modality, diabetes mellitus and statin use. However, after additional adjustment for markers of PEW (BMI, SGA and hsCRP), the HR, although numerically low (HR = 0.62), was no longer statistically significant.

Table 5.   Unadjusted and adjusted risk for 3-year mortality, associated with an increase in the apoB/apoA-I ratio during the first year (median: 12.2 months) of dialysis therapy (n = 182)
MortalityHR95% CIP
  1. DM, diabetes mellitus; BMI, body mass index; SGA, subjective global assessment; hsCRP, high sensitivity C-reactive protein; HR, hazard ratio.

  2. Adjustment for BMI, SGA and hsCRP was carried out with values at the time of the second visit.

Crude0.450.24–0.82<0.01
Adjusted for age, gender, dialysis modality, DM and statin use0.480.22–0.98<0.05
Adjusted for the above + BMI, SGA and hsCRP0.620.26–1.36NS

Discussion

  1. Top of page
  2. Abstract.
  3. Introduction
  4. Patients and methods
  5. Results
  6. Discussion
  7. Conflicts of interest
  8. Acknowledgements
  9. References

The present study demonstrated that, whereas a single measurement of the apoB/apoA-I ratio did not predict overall mortality, time-stratified relative risk analysis, adjusting for confounders, indicated that in the short term (1-year mortality), a high apoB/apoA-I ratio is associated with a survival advantage. This finding supports the notion of a reverse epidemiology phenomenon in this patient group, independent of confounders. Our results were confirmed by a longitudinal analysis of changes in apoB/apoA-I ratio, where an increase in the ratio during the first year of dialysis therapy was associated with an improved survival, although adjustment for indices of PEW weakened this association. Finally, the paradoxical relationship of a high apoB/apoA-I ratio to survival was not seen during long-term follow-up, probably illustrating a temporal impact of competing risks.

In our study, CKD stage 5 patients had a lower apoB/apoA-I ratio compared with that of the general Swedish population [23]. This might seem to be a surprising finding, since CKD patients are believed to have both high apoB and low apoA-I concentrations [15]. The association between the apoB/apoA-I ratio and BMI, reported in this study and previously by Gardner et al. [24], may at least in part, explain this observation. Jungner et al. [23] reported a higher apoB/apoA-I ratio in men than in women. Intriguingly, our data showed no difference in ratio values between the two sexes, suggesting a different gender pattern in CKD. Generally, female sex hormones are associated with a more beneficial lipid profile [25]. Whether this may be linked to the lack of survival advantage in the female CKD population when compared with men [26] remains to be elucidated. However, it should be stated that whereas Jungner et al. [23] included women with all ages, most women in our study were postmenopausal.

Overall, our results demonstrated a complex relationship between the apoB/apoA-I ratio and survival amongst incident dialysis patients. Wasting is a serious problem in this and other chronic disease patient groups, and low fat mass, low BMI as well as weight loss over time have been associated with increased mortality [27–29]. The role of BMI is reversed in CKD, and Kalantar-Zadeh et al. [6] recently suggested that a time difference of competing risks could explain these paradoxical findings, i.e. an adverse effect of an increased fat mass and/or hypercholesterolaemia may have a more limited impact in CKD patients due to the susceptibility to premature death due to PEW. In contrast, wasting conditions resulting in a low BMI may be contributory to an increased risk for mortality, suggesting a ‘time discrepancy’ between the risk associated with short-term risk factors (such as PEW) and long-term risk factors (such as obesity or hyperlipidaemia). To address this issue, we used the concept of time-stratified relative risk analysis of the apoB/apoA-I ratio [22]. Thus, patients who died during the first year of dialysis therapy had a lower apoB/apoA-I ratio, perhaps as a reflection of PEW. However, those who lived longer had a risk profile similar to the general population, with a high apoB/apoA-I ratio predicting mortality risk. Our longitudinal analysis, where a short-term (1-year) increase in the ratio was associated with a survival advantage, was in agreement with this concept. Thus, our results imply a dynamic role of the apolipoprotein profile, as reflected in the apoB/apoA-I ratio over time.

In our study, the apoB/apoA-I ratio was associated (both cross-sectionally and longitudinally) with inflammation and weight gain, consistent with the hypothesis of PEW as contributor to the ‘reverse lipid epidemiology’ phenomenon. However, adjustment for PEW and inflammation did not affect the prognostic value of the apoB/apoA-I ratio. Notably, Kilpatrick et al. [5] reported that after multivariate adjustments, including PEW, both low total cholesterol and low LDL cholesterol levels predicted poor outcome in prevalent HD patients. A growing body of evidence supports the notion that in certain chronic debilitating conditions low cholesterol and low lipoprotein concentration might be harmful. A hypothesis offered by Rauchhaus et al. [7] states that higher concentrations of lipoproteins could be beneficial due to their ability to bind endotoxins and modulate inflammatory immune response. Further studies are needed to shed light on these complex associations in the context of uraemia.

We acknowledge several limitations of our study. First, although our analysis demonstrated associations between the apoB/apoA-I ratio and mortality, we cannot predict causality. Second, we recruited incident dialysis patients and did not include a control group of healthy individuals. However, it would probably be not meaningful to follow healthy controls for only 4 years due to an expected low number of events. Third, we cannot exclude a possibility that the time stratification analysis and longitudinal changes might be affected by small number of events. Fourth, both BMI and DEXA measurements may be influenced by gross imbalances in hydration status of dialysis patients. Moreover, although we corrected for statin usage we cannot exclude that the putative pleiotrophic effects of statins could have affected the observed results. Finally, given the complexity of lipid metabolism on the one hand and renal disease on the other, this study of a rather small heterogeneous group of incident dialysis patients need to be confirmed in larger cohorts.

In conclusion, the present study showed that both a high baseline apoB/apoA-I ratio and an increase in the ratio over time were associated with a survival advantage amongst incident dialysis patients. Over a longer term, however, a high apoB/apoA-I ratio was a significant mortality risk factor, similar to the general population. Thus, the possible role of a time difference of competing risks needs further attention when evaluating mortality risk factors in dialysis patients. Although the ratio appeared to be influenced by nutritional or inflammatory status, adjustment for these confounders did not dramatically affect the size of the effect. This observation suggests that other factors might contribute to the paradoxical association between a high apoB/apoA-I ratio and mortality in CKD.

Conflicts of interest

  1. Top of page
  2. Abstract.
  3. Introduction
  4. Patients and methods
  5. Results
  6. Discussion
  7. Conflicts of interest
  8. Acknowledgements
  9. References

BL is employed by Baxter Healthcare Inc. PS participates as a scientific advisor for Gambro.

Acknowledgements

  1. Top of page
  2. Abstract.
  3. Introduction
  4. Patients and methods
  5. Results
  6. Discussion
  7. Conflicts of interest
  8. Acknowledgements
  9. References

We would like to thank the patients and personnel involved in the creation of this cohort. Our gratitude to Marion Verduijn and Friedo W. Dekker from the Department of Clinical Epidemiology at the Leiden University Medical Centre for their teaching of the time-stratified survival analysis. Also, we are indebted to our research staff (Ann Dreiman-Lif, Annika Nilsson, Anki Emmoth, Björn Anderstam, Monica Ericsson and Anki Bragfors-Helin). MC and JJC acknowledge a generous support of the ERA-EDTA research fellowships, MEC (EX2006-1670), Karolinska Institutet Center for Gender-based Research, the Swedish Heart and Lung Foundation (PS), the Swedish Medical Research Council (PS), Scandinavian Clinical Nutrition AB, Westmans Foundation (PS) and Martin Rind’s Foundation (PS).

References

  1. Top of page
  2. Abstract.
  3. Introduction
  4. Patients and methods
  5. Results
  6. Discussion
  7. Conflicts of interest
  8. Acknowledgements
  9. References