The HOlland NEphrology Study (HONEST) Group did a randomised, double blind, placebo controlled, crossover trial between April 2006 and October 2009 in three medical centres. The primary outcome measure of the trial was proteinuria, and the secondary outcome measure was blood pressure. All participants gave written informed consent. The study sponsor provided trial drugs at no cost.
We screened consecutive patients with renal disease who visited the nephrology outpatient clinics for the presence of non-diabetic nephropathy, as confirmed by analysis of blood and urine or renal biopsy. Inclusion criteria were blood pressure above 125/75 mm Hg in combination with residual proteinuria above 1.0 g/day during ACE inhibition at maximal dose (lisinopril 40 mg/day), creatinine clearance of 30 mL/min or above, and age over 18 years. For safety reasons, we excluded patients with systolic blood pressure of 180 mm Hg or above, diastolic blood pressure of 110 mm Hg or above, or both. Other exclusion criteria were diabetes mellitus, renovascular hypertension, decrease of creatinine clearance by at least 6 mL/min in the previous year, a cardiovascular event in the previous six months, immunosuppressive treatment, regular use (>1 day/week) of non-steroidal anti-inflammatory drugs, pregnancy, or breast feeding.
During a run-in period of at least six weeks, patients received ACE inhibition at maximal dose (lisinopril 40 mg/day) and stopped all other renin-angiotensin-aldosterone system blockers. Additional antihypertensive drugs such as β blockers, α blockers, calcium channel blockers, and diuretics were allowed and kept stable during the study (table 1). No dietary intervention took place during the run-in period.
Table 1 Baseline* characteristics. Values are numbers of patients unless stated otherwise
After the run-in period, patients were treated during four treatment periods of six weeks with, consecutively, ACE inhibition at maximal dose (lisinopril 40 mg/day) plus placebo and ACE inhibition plus angiotensin receptor blockade at maximal dose (lisinopril 40 mg/day plus valsartan 320 mg/day). Both treatments were combined with, consecutively, a low sodium diet (target sodium intake 50 mmol Na+/day; approximately 1200 mg Na+/day or 3 g NaCl/day) and a regular sodium diet (target sodium intake 200 mmol Na+/day; 4800 mg Na+/day or 12 g NaCl/day). The drug interventions were double blind, whereas the dietary interventions were open label.
To prevent systematic errors resulting from the crossover design, the different treatment periods were done in random order. Because of this randomisation and the rather short half life of the interventions (lisinopril 12.6 hours, valsartan 9 hours, low sodium diet <1 week31
), the protocol did not include wash-out periods.
We defined four different treatment sequences as follows. (1) Placebo plus low sodium diet, valsartan plus low sodium diet, valsartan plus regular sodium diet, placebo plus regular sodium diet. (2) Placebo plus regular sodium diet, valsartan plus regular sodium diet, valsartan plus low sodium diet, placebo plus low sodium diet. (3) Valsartan plus regular sodium diet, placebo plus regular sodium diet, placebo plus low sodium diet, valsartan plus low sodium diet. (4) Valsartan plus low sodium diet, placebo plus low sodium diet, placebo plus regular sodium diet, valsartan plus regular sodium diet. An independent pharmacist randomised these sequences, using a computer program. We implemented the random allocation sequence by means of sequentially numbered containers of study drug. Physicians enrolled patients, and the pharmacist allocated the study drug sequentially to consecutive participants. The randomisation code remained secret during the entire study; all participants, investigators, and care providers were blinded, except for the pharmacist.
Physicians gave the participants a list of food products that are commonly consumed in the Netherlands, together with their sodium content, at the time of inclusion. Diverse professional dietitians gave further dietary counselling in various autonomous dietary practices in the community. Except for a request to achieve the particular sodium targets (that is, 50 mmol Na+/day during the low sodium diet and 200 mmol Na+/day during the regular sodium diet), dietitians did not receive extra training or a script for this study. Each patient had two to four dietary counselling sessions. Individualised counselling used the general principle of remaining as close as possible to the patients’ preferences and nutritional habits, to increase feasibility and compliance, taking into account adequacy of nutritional requirements as well as sodium content. For the periods on the regular sodium diet, the patients were advised to maintain their habits regarding sodium intake. For the periods on the low sodium diet, patients were advised not to add any salt to their food and to replace sodium rich products with sodium poor products. We monitored compliance by 24 hour urinary sodium excretion and informed the physician, patients, and dietitians of these results.
Measurements and calculations
At the end of each six week treatment period, patients collected 24 hour urine samples and blood pressure was measured and blood sampled after an overnight fast. Additionally, in the middle of every six week treatment period, patients collected 24 hour urine samples to monitor dietary compliance.
We measured proteinuria in 24 hour urine samples with a turbidimetric assay using benzethonium chloride (Modular, Roche Diagnostics, Mannheim, Germany). We measured blood pressure at one minute intervals with an automatic device (Dinamap, G E Medical Systems, Milwaukee, WI, USA) with the patient in a supine position. After 15 minutes of measurements, we used the mean of the last three readings for further analysis. We determined blood electrolytes, lipids, proteins, and urinary electrolytes by using an automated multianalyser (Modular, Roche Diagnostics, Mannheim, Germany). We assessed dietary sodium intake from urinary sodium excretion. We calculated creatinine clearance from creatinine concentrations in plasma and in 24 hour urine samples. We used the Maroni formula to assess dietary protein intake from urinary urea excretion.32 33
We assessed peripheral pitting oedema at the pretibial area of both legs by visual and manual examination and scored it as absent or present.
We expected that patients would present with a mean proteinuria of approximately 2 g/day during ACE inhibition. On the basis of previous studies, we assumed a 35% reduction in proteinuria by addition of a low sodium diet on top of ACE inhibition plus angiotensin receptor blockade and a standard deviation of 0.75 in log transformed proteinuria response.8 25 16 17 20
From these numbers, we estimated that 51 patients had to complete the crossover design to provide 90% power to detect a statistically significant difference. We used a significance level of α=0.0083 (rather than α=0.05) to adjust for six primary comparisons of interest. To account for a 10% dropout rate during the trial, we would need to include 56 patients (PASS 10, NCCS, East Kaysville, UT, USA). Of note, the sample size is smaller than would have been needed in a non-crossover study, as the same patient provides data for each treatment group and this increases power, owing to the smaller within patient variability than between group variability.34 35
We analysed data for the 52 patients who completed the trial, and we present these data here. Additionally, we analysed the data for all 54 patients who were included (intention to treat). As the effect estimates and confidence intervals were very similar and the statistical and clinical conclusions did not change, we have not shown these data. Before statistical testing, we natural log transformed skewed variables to obtain normality. We determined differences between the four different treatment sequences by using one way analysis of variance with Bonferroni’s post hoc tests and Pearson’s χ2 tests. We used paired t tests (which account for the same patients providing data for both treatment groups) to determine the effects of treatment. We did six comparisons for each parameter: ACE inhibition versus ACE inhibition plus angiotensin receptor blockade, ACE inhibition versus ACE inhibition plus low sodium diet, ACE inhibition versus ACE inhibition plus angiotensin receptor blockade plus low sodium diet, ACE inhibition plus angiotensin receptor blockade versus ACE inhibition plus low sodium diet, ACE inhibition plus angiotensin receptor blockade versus ACE inhibition plus angiotensin receptor blockade plus low sodium diet, and ACE inhibition plus low sodium diet versus ACE inhibition plus angiotensin receptor blockade plus low sodium diet. To allow for multiple testing, we set the type I error (α) at 0.0083 (Bonferroni correction) for analyses of the primary outcome (proteinuria). Furthermore, we did a linear mixed model analysis to check for carryover effects, with log transformed proteinuria as a dependent variable, participants as a random factor, and treatment and sequence as well as their interaction (treatment*sequence) as fixed factors.
We give data as mean with standard error (SE) when normally distributed or as geometric mean with 95% confidence interval when skewed. We report only unadjusted P values. We used SPSS 16.0 for Windows for all analyses.