A randomized controlled trial of different serum phosphate ranges in subjects on hemodialysis

Large observational studies have identified hyperphosphatemia as an independent risk factor for cardiovascular disease and mortality in dialysis patients [1‐3]. Such studies cannot establish causality, but current guidelines encourage careful control of serum phosphate in the uncertain expectation that it will improve outcomes. Studies of serum phosphate focus on licensing requirements for the binder [4‐11] rather than the benefits of lowering serum phosphate per se, and have not addressed the fundamental questions of why, and to what target, serum phosphate should be controlled [12, 13]. Despite significant investment in expensive oral phosphate binders [14‐19] (up to $4500 per month per patient) [20], large pill burden (up to 15 pills daily) and significant rates of non-adherence [18], there remains no evidence that lowering serum phosphate definitely improves clinical outcomes.

Randomising subjects to a binder or placebo is unlikely to gain ethical approval for anything other than a short-term study, but we propose that a randomised controlled study of different concentrations of serum phosphate would answer the crucial questions around the value of controlling serum phosphate. However, whether it is possible to achieve and maintain low concentrations of serum phosphate in the range required to enable comparison with a higher concentration, and whether it is possible to maintain the phosphate concentration separation between the groups over an extended time period, is unknown. Furthermore, the consent rate for a trial in which subjects would be randomised to a treatment target above the current guidelines needs to be assessed.

We report the results of a study to examine these factors and to determine whether a large scale, long term randomized controlled outcome trial (RCT) may be possible. The Serum Phosphate Intervention in Renal Replacement Therapy (SPIRiT) trial randomised 104 hemodialysis patients to two groups with different serum phosphate target ranges; a Lower Range Group (LRG) and a Higher Range Group (HRG). Initial screening involved all patients from the two adult renal services of Greater Manchester in England, a conurbation with a total population of 2.8 million people.

All trial procedures adhered to the Declaration of Helsinki and the trial protocol was reviewed and approved by the National Regional Ethics Committee East Midlands Derby, REC Ref No. 13/EM/0042.

Details of the SPIRiT trial objectives, design and methods have been published previously [21]. It is an open-label, dual center, randomized controlled trial in which all hemodialysis patients from two large adult renal centers in Manchester (centers 1 and 2) had their electronic medical records screened. All in-center HD patients aged 30 years or above, on regular dialysis for at least 6 months (to ensure no recovery of renal function), meeting UK Renal Association standards for quality of dialysis, with a 3-month mean serum phosphate > = 1.4 mmol/L (4.3 mg/dL) despite binders, serum PTH of < 900 pg/L, and able to consent, were deemed eligible. All eligible subjects were approached for consent. Written Informed consent was obtained from subjects who agreed to enter the study and baseline characteristics recorded. Their current phosphate binder was discontinued and serum phosphate concentration checked after a 3 week washout period. The washout period ended if the phosphate concentration rose to > = 1.7 mmol/L (5.3 mg/dL) but could continue for a maximum of 5 weeks. Those whose serum phosphate concentration did not reach 1.7 mmol/L (5.3 mg/dL) were excluded from randomisation. All blood samples were collected before the commencement of a dialysis session (pre-dialysis samples). All subjects received a single dietetic review and advice regarding control of dietary phosphate intake.

The percentage of study participants achieving, and being maintained within, the higher and lower target ranges for serum phosphate, over the duration of the maintenance phase of the study.

Randomisation was to the lower range group (LRG with treatment serum phosphate target of 0.8 to 1.4 mmol/L - 2.5-4.3 mg/dL) or higher range group (HRG with treatment target of 1.8 to 2.4 mmol/L - 5.6–7.4 mg/dL). The ‘titration phase’ comprised 2 months following randomization during which the subjects were prescribed variable doses of Lanthanum or Sevelamer (if necessary), in order to achieve study-group targets. Serum phosphate concentration was measured once a week during this phase with dose-adjustment of phosphate binder.

Primary end-point is the percentage of study participants achieving, and being maintained within, the higher and lower target ranges for phosphate, over the maintenance phase of the study [21]. Forty-seven subjects completed the titration phase in the HRG and 50 subjects in the LRG. At the end of titration, 29 HRG (61%) and 19 LRG subjects (38%) had achieved the target serum phosphate concentration, resulting in a mean phosphate of 2.00 ± 0.4 mmol/L (6.2 +/− 1.2 mg/dL) (HRG) and 1.63 ± 0.4 mmol/L (5.0 +/− 1.2 mg/dL) (LRG) (p < 0.01).

Figure 1 shows the percentage of subjects achieving target serum phosphate concentrations at sequential time-periods after randomization.

At the start of washout, mean serum phosphate concentrations in the two groups was similar – 2.00 +/− 0.42 mmol/L (6.2 +/− 1.3 mg/dL) in the HRG versus 1.93 +/− 0.32 mmol/L (6.0 +/− 1.0 mg/dL) in the LRG. Following the washout period and prior to randomization mean serum phosphate concentrations in the two groups were not different (p = 0.15) – 2.1 +/− 0.36 mmol/L (6.5 +/− 1.1 mg/dL) in HRG Vs 2.2 mmol/L +/− 0.4 mmol/L (6.8 mg/dL +/− 1.2 mg/dL) in the LRG, but separated with a statistically significant difference at the end of titration, and maintained a separation of approximately 0.34 mmol/L (1.1 mg/dL) throughout the maintenance phase (Fig. 2), despite a large range.

Seven hundred ninety-eight subject records were screened and 555 subjects excluded because they did not meet the screening criteria (Fig. 3). Fifty-three percent of the subjects approached gave consent (129/243). There was a difference in recruitment between the centers. Fifteen percent of the screened population was randomized at center 1 (76/494) versus 9% at center 2 (28/304). One hundred four subjects were randomized resulting in a recruitment rate of 13% of the dialysis population.

Table 7 shows serum concentration of PTH in HRG and LRG for the duration of the study. There is no significant difference in the median PTH between the two groups at any point during the study. However, one patient was taking oral Cinacalcet 30 mg a day at the start of the study – the patient continued with this dose for the study duration. Another patient in the HRG was started on Cinacalcit 30 mg during the maintenance phase – this patient underwent a deceased donor kidney transplantation and exited the study after a month of starting the Cinacalcit.

Four hundred thirteen blood samples were collected in the HRG during the maintenance period, i.e., visit 10 onwards. Individual patients’ serum phosphate concentration varied quite widely, but considering all blood results collectively, 55% were within the treatment target range of 1.8 to 2.4 mmol/L (5.6 to 7.4 mg/dL). Serum phosphate concentration was below the target in 134/413 samples and above the target in 55/413 blood samples. Only one patient remained within in the target range for the entire duration of the study.

Four hundred forty-nine blood samples were collected in the LRG during the maintenance period, i.e., visit 10 onwards. Forty-five percent of all serum phosphate results were within the treatment target range of 0.8 to 1.4 mmol/L (2.4 to 4.3 mg/dL). Serum phosphate concentration was below the target in 8/449 samples and above the target in 246/449 blood samples. Only two patients remained within the target range for the entire duration of the study.

Despite obtaining a clear separation in the serum concentration of phosphate levels between the two groups, mean serum calcium concentrations remained steady with no significant differences between the two groups (Fig. 6) – it remained steady at 2.3 +/− 0.2 mmol/L in both the groups. Serum albumin showed no significant difference between the two groups and the mean remained at 35 +/− 4 g/L through the maintenance period. Serum Cholesterol dropped significantly ion the LRG from 4.1 +/− 0.9 mmol/L at randomization to 3.6 +/− 0.9 mmol/L at the end of titration. At the end of titration, mean cholesterol concentrations in the HRG had remained the same as that at randomization, 4.0 +/− 1 mmol/L. Hence, at visit 10, there was a statistically significant lowering of cholesterol in the LRG with p < 0.04. But the statistical significance did not persist beyond visit 10.

This study demonstrates that it is possible to recruit and randomize dialysis patients to two different serum phosphate concentrations, and maintain them over 10 months, despite the HRG being above the range recommended by current guidelines. Approximately 50% of the SPIRiT subjects in either group achieved the trial target range of serum phosphate at multiple-time points in the trial and a difference of 1.1 mg/dL (0.34 mmol/L) between the HRG and the LRG was achieved and maintained during the 10-month follow-up. Having two mutually exclusive target ranges rather than target concentrations ensured separation of the groups’ mean phosphate values. Block and colleagues [3] studied retrospective data for 40,538 in-center hemodialysis subjects sampled from the Fresenius medical care North America patient statistical profile system, and stratified serum phosphate concentrations into 8 categories of 1 mg/dL increments. With multivariable adjustment, the relative risk of death increased from 1.0 for serum phosphate concentration of 4.0 to 5.0 mg/dL to 1.1, > 1.2, 1.4, 1.7 and 2.0 at serum phosphate concentrations of 5–6, 6–7, 7–8, 8–9 and > 9 mg/dL respectively. The phosphate separation achieved in the SPIRiT study was 0.34 mmol/L - equivalent to 1.05 mg/dL, and based on such retrospective data, would be adequate to show a difference in clinical outcomes, if any, in a large-scale RCT.

We did not achieve our goal of serum phosphate between 0.8 to 1.4 mmol/L (2.5–4.3 mg/dL) as less than 1/2 subjects in that group achieved that goal and the mean level was 1.5 +/− 0.4 mmol/L (4.6 +/− 1.2 mg/dL). We did achieve a difference between the 2 groups as the high phosphate group had a mean of 2.0 +/− 0.5 mmol/L (6.2 +/− 1.5 mg/dL) and still did not have a high percentage in the range. There was only one other study which attempted to randomize a similar number of patients to two different serum phosphate ranges [23]. This study was started after we started to recruit for the SPIRiT study, and had a much shorted follow-up period of 26 weeks. The investigators further used calcium carbonate, a calcium containing phosphate binder. While it can be argued that there is no RCT evidence showing increased mortality in dialysis patients with calcium containing binders, there is evidence that they increase the risk of vascular calcification [24]. This is a compelling reason to not use these in the larger study. Further, there was no washout period which makes it possible that patients were included in the study who did not need to be on the phosphate binders in the first place.

Whilst it may be possible to achieve specific concentrations of phosphate over a specified time period, a larger study will depend on the willingness of physicians and patients to participate. Consequently, the secondary end points of this study examine this. Of the 22 local nephrologists, only 2 raised concerns regarding their patients taking part in the study, and this was not a clinical concern but related to the effect higher phosphate concentrations might have on their unit’s adherence to national targets. They were anxious that a deterioration in reported results might attract criticism – an understandable concern in the current target-driven culture. They agreed to take part on condition that that year’s UK Renal Registry report carried a comment about the possible effects of the SPIRiT study on compliance with targets.

Willingness of patients to participate, once approached, was high with 53% (129/243) of the eligible participants consenting. Local research practices and trial personnel appear to have had a significant effect on recruitment with 16% of screened subjects at center 1 consenting versus 9% at center 2. This was most likely due to the study clinician being primarily based at center 1. In a much larger study it is unlikely that the majority of centers would have a full time research clinician available and therefore the lower recruitment rate may be more representative of ‘real life’. In addition, the pace of recruitment at the two centers varied (8.4 subjects per week recruited versus 4.3 per week respectively) probably for similar reasons.

At randomization, 29.4% of the HRG had diabetes versus 20.8% of the LRG. More HRG had previous coronary artery disease (29.4% of HRG versus 18.9% of LRG) and more vascular disease (31.4% HRG versus 22.6% LRG). We did not stratify for the cardiovascular risk factors at randomization because of the small sample size. In a larger study, it is likely that an adequate sample size will resolve these discrepancies. However, as the EVOLVE [25] study demonstrated, a large sample size may not always guard against bias, and the larger study would need stratified randomization for established cardiovascular risk factors in dialysis patients.

The overall trial drop-out rate of 27% per annum was comparable to other interventional RCTs in dialysis patients. The HEMO study randomized 1846 patients to high flux or low flux dialysis and followed them up for a mean of 2.84 years. Five hundred ninety patients dropped out over the course of the study which gives a drop-out of 33% during the study [26]. On the other hand, the EVOLVE study randomized 3883 dialysis patients with secondary hyperparathyroidism to cinacalcet group or placebo with to examine differences in clinical outcomes. Seventy-eight percent of the subjects in the cinacalcet group and 61% in the control group stopped the study medication for non-protocol-specific reasons resulting in a crossover between the groups [27]. In the SPIRiT study, the drop-out rate in the LRG and HRG was 7/51 (13%) and 10/53 (19%) respectively. Consent withdrawal was higher in the LRG (4 vs 9). Although subjects did not give clear explanations for their withdrawal it seems likely that pill burden may have been a factor, with all LRG consent withdrawals happening in the first 4 months of the trial. Interestingly during the washout period at the start of the study 2 patients said they felt so much better having stopped their binders that they refused to restart and were withdrawn from the study.

Only non-calcium containing phosphate binders were used in this study, and at the start of the trial in 2013, availability was limited to lanthanum and sevelamer [13]. Any subject requiring high dose phosphate binders and intolerant of lanthanum was limited to sevelamer tablets which could result in a pill-burden of up to 15 tablets a day. Since the completion of this study, other non-calcium binders have become available which might make a larger study easier by providing subjects with more choice [28] and lower pill burden.

One patient in the HRG was on Cinacalcet 30 mg a day at the start of the trial and stayed on it for the duration of the study. One other patient needed to be started on Cinacalcet during the study. Hypercalcemia was not a particular problem since only non-calcium containing phosphate binders were used. The dose of 1.25 alfa–calcidol was determined by the clinical care team and the presence or absence of alfa-calcidol therapy did not make an impact on the calcium and phosphate concentrations in the trial. However, the availability of alfa calcidol ensured that hypocalcaemia was adequately treated. With the availability of non-calcium containing phosphate binders, it is possible to deliver effective phosphate binding without the need for calcium loading; It is also possible to control PTH with calcimimetics despite a higher serum phosphate concentration – the availability of medication to address these mineral disorders in isolation makes it possible to design and conduct such studies.

Although mortality in the HRG (9/51) was substantially higher than in the LRG (2/53), the overall mortality for the study was 10.6%, which compares favourably with the UK Renal Registry annual mortality rate of 11% [29]. A similar rate would be expected in a small trial such as this, with standard exclusion criteria possibly creating a selection bias towards subjects with fewer comorbidities. However, as the study progressed the difference in death rate between LRG and HRG increased noticeably, causing the Safety Monitoring Committee to scrutinise the data in ever greater detail each month. The Committee satisfied itself that the difference in mortality was likely to have occurred by chance plus the notable imbalance in risk factors at randomisation. Prevalence of diabetes, pre-existing coronary artery disease and vascular disease were significantly higher in the HRG. The trial was not powered to examine clinical outcomes such as mortality and cardiovascular events.

LRG and HRG achieved the treatment target ranges just about 50% of the time. Only 3 patients stayed in their respective target ranges for the entire duration of the study. A high pill burden which can enhance poor adherence is very likely the cause for the LRG patients not always achieving their target ranges. It is unclear why the serum phosphate concentrations were below the target range in the HRG over 30% of the time (134/413 samples in the maintenance period). This happened despite careful screening and washout ensuring that no patient who did not need a phosphate binder at baseline, was inadvertently recruited. One possibility is that the years of educating the dialysis staff and the patients about the benefit of low phosphate inadvertently introduced a bias where in the HRG patients practiced extra care about maintaining a low phosphate diet during the study period, since they were not on phosphate binders. In the design of the larger study, it is important to include a regular diet diary monitoring to see if patient and provider biases play a role in tighter phosphate control in the HRG.

Oral phosphate binder tablets can have significant gastro-intestinal side effects such as bloating which can lead to reduced oral intake and poor nutrition. We monitored serum albumin as a surrogate marker of nutrition, and did not find any significant reduction in the LRG which had the largest pill burden. As expected, the LRG had a lower serum cholesterol concentration (though not statistically significant). Sevelamer is a polymer which is known to cause a reduction in cholesterol levels.

This study demonstrates that it is possible to randomise dialysis patients to different concentrations of phosphate control – a hope first suggested in the introductory sections of guidelines published as long ago as 2003 (KDOQI) and 2009 (KDIGO). Though the TARGET study showed the possibility of randomization, SPIRiT study demonstrated that a clinically significant difference in serum phosphate concentrations can be maintained in the two groups over a prolonged period of time (1 year) which is necessary to study differences in clinical outcomes in the two groups. Both physicians and patients agreed to participate in numbers that suggest a larger scale study is possible. Despite this being an interventional RCT, 43% of the eligible target population were successfully consented, and the drop-out rate was comparable to published large scale, long-term dialysis studies. A clinically significant separation of mean serum phosphates was achieved between the groups representing a possible increased relative risk of 0.2 based on published observational data.

Whilst the number of deaths in the HRG was noticeably higher than in the LRG the study is under-powered to evaluate this outcome.

These results suggest that a similar but larger 2 year study is indeed possible. Based on a lower conversion rate of 10% from screening to randomisation, screening of 17,160 patients would be expected to result in 1716 consents; 858 patients in each group. One thousand two hundred fifty-three patients would remain after 12 months and approximately 915 in follow-up after 24 months. Such a study would counter “the lack of patient-centered outcomes as end points in the majority of trials in this field” as highlighted by KDIGO.