Short-Term Effects of Very-Low-Phosphate and Low-Phosphate Diets on Fibroblast Growth Factor 23 in Hemodialysis Patients

Study Design

Figure 1 illustrates the study design and outcome assessments. Between January 3, 2018 and June 8, 2018 we conducted a randomized, active-controlled trial with a crossover design in which we randomly assigned the participants with an allocation ratio of 1:1 to two interventions. The trial protocol is available in Supplemental Appendix 1. The study was approved by the institutional review board at Far Eastern Memorial Hospital (FEMH-106108-F) and was registered online before study initiation (Clinicaltrials.gov identifier NCT03367338). At the time of study initiation, the difference in change-from-baseline FGF23 levels between the two diets was designated as the primary outcome measure, without specifying the assay type. After the study participants were enrolled, we specified the primary outcome (FGF23) measured with intact assay, removed 1,25-dihydroxyvitamin D₃ from the list of secondary outcomes, and added C-terminal FGF23 as the secondary outcome. The rationale for the changes are that intact FGF23 is biologically active and can more precisely reflect dietary response than C-terminal FGF23, and 1,25-dihydroxyvitamin D₃ measurement was unavailable at our central laboratory.

• Download figure
• Open in new tab
• Download powerpoint

Figure 1.

Study design and outcome assessments. Participants in group A received a 2-day diet with a PPR of 8 mg/g, followed by a 5-day washout period, and then received a 2-day diet with a PPR of 10 mg/g. Those in group B received the diets in the opposite order. A total of four repeated measurements for the primary and secondary outcomes were obtained before dialysis sessions. Before each study phase, each participant performed a 3-day dietary record to estimate the nutrient content of his or her usual diet. During the study periods, dietary adherence was assessed by a 2-day dietary record. PPR, phosphate-to-protein ratio; R, randomization.

Study Population

Participants were recruited from a hemodialysis unit of a tertiary teaching hospital and asked to participate if they met the following inclusion criteria: (1) aged >20 years, (2) having ESKD and having undergone thrice-weekly hemodialysis for >3 months, (3) having adequate dialysis (urea reduction ratio ≥65%), (4) most recent serum phosphate level >5.5 mg/dl or between 3.5 and 5.5 mg/dl with regular phosphate binder use, and (5) serum intact parathyroid hormone (PTH) level <800 pg/ml. In accordance with available study diets restricted to between 42.5 and 67.5 kg body wt, we did not enroll participants with a dry weight outside of this range. The exclusion criteria were serum albumin level <2.5 g/dl, hospitalization within the past 4 weeks, psychiatric disorders, mental retardation, dislike of the study meals, or poor dietary adherence. All participants provided written informed consent.

Randomization

After recruitment, one of the investigators equally stratified the participants by intact PTH level and dialysis shift before randomization. The intact PTH level is used as one of the stratification factors because it determines whether response to a dietary intervention. Participants are stratified by dialysis shift because there is concern about postprandial hypotension, whereby participants with the afternoon shift consume study lunch during hemodialysis and are therefore unable to adhere to the study protocol, leading to an unwanted dropout rate and the subsequent variability in the analyses. Within each stratum, a random allocation sequence was generated by computer-based randomization and used to randomly allocate the participants to the interventions. The investigator sent the random allocation sequence to the dietitian, who prepared individualized study meal boxes for the participants according to assigned group. The study meals were prepared by the same person and with similar natural food materials, appearance, and taste, so that the participants were blinded from their allocation group.

Characteristics of Study Diets

The characteristics and preparation of the study diets are described in Supplemental Appendix 2. In brief, both study diets (5) were made up of natural food sources without any phosphate-containing additives, boiling meats to reduce phosphorus amount (16), and increased proportion of plant-based foods (17). Before patient enrollment, the compositions of the study diets were subjected to chemical analysis; the measurement methods are described in Supplemental Appendix 3. Table 1 lists the nutrient composition of each diet. The two diets were designed to have minimal variation between them in total calorie, protein, and calcium contents, while differing in phosphate and phosphate-to-protein ratio contents. As shown in Supplemental Table 1, the study diet was provided in three meals per day. The nutrient compositions of the study diets categorized according to body weight are provided in Supplemental Table 2.

Dietary Interventions

As shown in Figure 1, each participant consumed two low-phosphate diets, each for 2 days, and diet order was randomized. Participants in group A consumed a very-low-phosphate diet, with a phosphate-to-protein ratio of 8 mg/g, which is equivalent to 10 mg phosphate per kg of body wt, for 2 days, which was followed by a 5-day washout period in which they adhered to their usual diets. Then, they followed a 2-day low-phosphate diet, with a phosphate-to-protein ratio of 10 mg/g, which is equivalent to 12.6 mg phosphate per kg of body wt. Those in group B consumed the study diets in the opposite order. No additional food was allowed during each study period.

The delivery of study meals was assimilated into the routine hemodialysis schedule. During the intradialytic period, study meal boxes were provided at the hemodialysis unit. During the interdialytic period, the packaged study meal boxes were retrieved at the hospital cafeteria or delivered to the residences via a preexisting home delivery service and consumed by the participants as outpatients. Delivery of the meals and consumption of the study diets were verified by telephone.

We made considerable efforts to enhance dietary adherence; these efforts are summarized in Supplemental Appendix 4. To avoid interference with the effects of low-phosphate diets, the following prescriptions were not allowed to change during the entire study period: dialysis duration, dialysis frequency, dialysis shift, dialysate calcium concentration, and dosages of medications, including phosphate binder, vitamin D analogs, and iron agents. Adherence with phosphate binders was assessed before and after each study period.

Dietary Assessment

Before each study period, participants prospectively maintained a dietary record of their daily intake for 3 days, including a dialysis weekday, a nondialysis weekday, and a nondialysis weekend day, allowing us to estimate the nutrient content of their usual diet as well as dietary adherence. During the study periods, the participants maintained a 2-day dietary record of their consumption of portions of the assigned study diets and foods outside of the study diets. The completeness, consistency, and clarity of the food diaries were reviewed by the dietitians.

Data Collection

The following data were recorded: age, sex, dry weight, body mass index, duration of dialysis therapy, history of parathyroidectomy, interdialytic weight gain, dialysis unit BP, type of arteriovenous shunt, dialysate calcium concentration, urea reduction ratio, hemoglobin, ferritin, alkaline phosphatase, albumin, 25-hydroxyvitamin D, glucose, and the amount, frequency, and type of medications, including phosphate-binding agents and vitamin-D analogs. The phosphate-binder doses among study participants were compared by calculating phosphate-binding equivalent dose as described by Daugirdas et al. (18).

Laboratory Measurements

Nonfasting venous blood samples were drawn at around 7:30 am for participants with the morning shift and at around 12:30 pm for those with the afternoon shift. At the baseline and the beginning of each study period, blood samples were drawn before the second dialysis session of the week. At the end of each study period, blood samples were drawn before the third dialysis session of the week. Standard assays for serum phosphate and calcium were performed using automated analyzers. Intact PTH was analyzed in serum using an immunoradiometric assay (ELSA-PTH; Cisbio Bioassays, Codolet, France). Intact FGF23 was assessed in serum using an ELISA (Kainos Laboratories, Tokyo, Japan). The C-terminal fragments of FGF23 were assessed in EDTA-plasma using a sandwich ELISA (Immutopics, San Clemente, CA) according to the manufacturer’s instructions. Each sample was run in duplicate, and mean values are presented.

Outcomes

The primary outcome measure was mean difference in change-from-baseline intact FGF23 level between the two dietary interventions. Secondary outcomes were difference in change-from-baseline serum phosphate, intact PTH, and C-terminal FGF23 levels between the two dietary interventions. We also measured difference in change-from-baseline serum calcium level between the two dietary interventions. As depicted in Figure 1, a total of four repeated measurements for each participant were obtained. The laboratory technicians who performed the outcome measurements were blinded to the allocation sequence.

Sample Size Determination

On the basis of the results from our meta-analysis (1), a total sample size of 29 was required to achieve 90% power and a type 1 error of 0.05, assuming a standardized mean difference in FGF23 levels between two low-phosphate diets of 0.74 and a dropout rate of 25%.

Statistical Analyses

Continuous measures were evaluated as means (±SD) or median (first and third quartiles), and categorical variables as counts and percentages. For non-normally distributed data, Wilcoxon signed-rank tests were performed to evaluate between-group differences. For normally distributed data, paired t tests were performed. Effect estimates for the primary and secondary outcomes were presented with Cohen d values, which was calculated as the difference between two means divided by a sd for the data. The sign of Cohen d indicates the direction of the effect. The Cohen d values of 0.2, 0.5, and 0.8 are suggested to correspond to small, medium, and large effects, respectively. We used mixed-effects models to examine the difference in treatment effect of the low-phosphate diets on outcomes and the presence of potential bias resulting from carryover and period effects. In each mixed-effects model, the dependent variable was a primary or secondary outcome, the participant was included as a random effect, and the independent variables were diet, group, and study period. To take account of the stratified randomization, the stratification factors were included as covariates in the model. A two-sided P value of <0.05 indicated statistical significance. All analyses were performed with SAS version 9.4 software (SAS Institute, Cary, NC).