Commentary on selected evidence

With advice from topic experts we selected 3 studies for further commentary.

Hyperphosphataemia in chronic kidney disease: cost-effectiveness of phosphate binders in children, young people and adults

For phosphate binders in children, young people and adults, we selected the network meta-analysis (NMA) by Palmer et al. (2016) for a full commentary. The reasons for selection were that it was highlighted by more than one topic expert as a significant addition to the body of evidence in this clinical area, and includes a large number of studies.

What the guideline recommends

For adults with CKD, NICE's guideline on hyperphosphataemia in chronic kidney disease recommends first line treatment with calcium acetate, to control serum phosphate in addition to dietary management. For adults with stage 4 or 5 CKD who are not on dialysis and who are taking a calcium-based binder, consideration of switching to a non-calcium-based binder is recommended if calcium-based phosphate binders are not tolerated, if hypercalcaemia develops, or if serum parathyroid hormone levels are low.

For adults with stage 5 CKD who are on dialysis and remain hyperphosphataemic despite adherence to the maximum recommended or tolerated dose of calcium-based phosphate binder, the guideline advises considering either combining with, or switching to, a non-calcium-based binder.

For adults with stage 5 CKD who are on dialysis and who are taking a calcium-based binder, if serum phosphate is controlled by the current diet and phosphate binder regimen but:

  • serum calcium goes above the upper limit of normal, or

  • serum parathyroid hormone levels are low 

the guideline recommends consideration of either combining with, or switching to, sevelamer hydrochloride or lanthanum carbonate, having taken into account other causes of raised calcium.

Methods

The NMA by Palmer et al. (2016) compared the phosphate binders sevelamer, lanthanum, iron, calcium, colestilan, bixalomer, nicotinic acid, and magnesium, in adults with CKD.

Prior to the NMA, a pair-wise meta-analysis was conducted, using a random effects model, to determine treatment efficacy. For the NMA, heterogeneity was explored using a common variance, and direct and indirect estimates were compared to evaluate consistency.

Prespecified sensitivity analyses were also conducted for studies focusing on dialysis, patients younger than 60 years, longer than 12 months follow-up, and low baseline serum phospherous. A post-hoc analysis was also conducted to compare all-cause mortality between sevalamer and calcium, incorporating a longer term follow-up study.

The inclusion criteria were:

  • parallel-group randomised clinical trials (RCTs)

  • a follow-up at least 4 weeks

  • adults with CKD

  • allocation to a phosphate binder, placebo, or standard care.

The analysis included a total of 12,562 patients from 77 studies, 62 (11,009 patients) of which were performed in a dialysis population. Risk of bias was assessed for the included studies according to methods outlined in the Cochrane handbook. The primary outcome was all-cause mortality. Additional outcomes were cardiovascular mortality, myocardial infarction, stroke, adverse events, serum phosphorus and calcium levels, and coronary artery calcification.

Results

Compared to placebo, The NMA found that no class of drug lowered mortality or cardiovascular events. All phosphate binders, except for colestilan, significantly lowered serum phosphorus levels, with iron performing best for this outcome (standardised mean difference −1.58, confidence interval [CI] −2.02 to −1.14). However, it should be noted that placebo-controlled trials were of short duration, with a maximum of 3 months.

Treatments were ranked according to their probability of being the best treatment for a specific outcome, and the surface under the cumulative ranking curve was estimated using the network rank command.

In terms of specific drug comparisons, the NMA results showed:

  • Sevalamer significantly reduced all-cause mortality compared to calcium (odds ratio [OR] 0.39, 95% CI 0.21 to 0.74, p=0.005). However, when the INDEPENDENT trial (n=466) was removed from the analysis, the results were no longer significant (OR 0.61, 95% CI 0.37 to 1.01).

  • Lanthanum was non-significant for all-cause mortality when compared to calcium (OR=0.78, 95% CI 0.16 to 3.72].

  • Iron was non-significant for all-cause mortality when compared to calcium (OR=0.37, 95% CI 0.09 to 1.60].

  • Colestilan was non-significant for all-cause mortality when compared to calcium (OR=0.55, 95% CI 0.07 to 4.43).

  • Sevelamer significantly reduced coronary artery calcification scores compared to calcium (standardised mean difference −0.20; 95% CI −0.40 to −0.01). However, the clinical significance of this result was reduced due to the short duration of trials.

  • There were no significant differences between non-calcium binders in terms of all-cause mortality.

Strengths and limitations

Strengths
  • The meta-analysis only included RCTs.

  • The network meta-analysis design enabled comparative analysis between specific phosphate-binder classes against each other or placebo, despite the lack of head-to-head trials. This included an assessment of heterogeneity, pre-specified sensitivity analysis and comparison of direct and indirect estimates to evaluate consistency.

  • The meta-analysis enhanced the evidence base in this area, building on a previous NMA (28 studies, n=8,335) by including a larger number of studies and analyses.

Limitations
  • The included placebo-controlled trials were of short duration, with a maximum of 3 months. The comparisons also varied considerably in terms of median follow-up, which may have influenced the reporting of outcomes.

  • The included studies were of variable quality, with a high or unclear risk of bias.

  • It was unclear whether attempts were made to identify unpublished studies.

  • Imprecise results of testing for heterogeneity may have caused some inconsistency.

  • The all-cause mortality benefit of sevelamer over calcium-based treatment was heavily dependent on a single large trial (INDEPENDENT). This benefit became non-significant when the trial was removed from the analysis.

  • Both calcium acetate and calcium carbonate were considered together as calcium-based binders and compared with non-calcium-based binders. This overlooked the finding in NICE guideline CG157 that calcium acetate has a far lower, and therefore safer, calcium load than calcium carbonate.

  • The impact of the results was limited by the exclusion of paediatric studies.

  • The study did not undertake any cost-effectiveness analysis, such as to explore whether the benefits of sevalamer could justify its higher cost.

  • The results reported in the abstract and the results section for adverse effects were discrepant with the adverse effects data presented in Figures 4 and 5.

Impact on guideline

While iron-based therapies were ranked first in the NMA for lowering serum phosphate, the guideline recommendations do not specify which non-calcium-based binder should be substituted or added if the serum phosphate is not controlled or adverse events, including hypercalcaemia, occur with calcium-based binders. The only exception to this is for children and young adults where the recommendations specify sevelamer for switching or as an add-on treatment. The exclusion of paediatric studies in the NMA by Palmer et al. (2016) therefore limited the impact of the results.

The results indicated that no drug class lowered mortality or cardiovascular events compared with placebo. However, topic expert feedback indicated that the durations of the trials appear too short to draw any definitive conclusions about treatment effects on mortality, cardiovascular events or vascular calcification.

The finding of an all-cause mortality benefit of sevelamer over calcium-based treatment was heavily dependent on the inclusion of the INDEPENDENT trial (n=466), which was one of few larger trials with a longer duration. When removed from the analysis, the effect of sevelamer on all-cause mortality was not significant, compared with calcium.

The findings are consistent with previous network and conventional meta-analyses in this area, which have been the subject of NICE Medicines Evidence Commentaries in 2014 and 2016.

Topic expert feedback indicated that further trials, with adequate power and duration covering both adults and children, are needed. This will determine whether phosphate binders reduce mortality, and if they do, whether any type of binder is superior for this outcome.

Cost effectiveness analysis was not included in the NMA by Palmer et al. Topic expert feedback indicated that sevelamer carbonate is available at considerably reduced cost compared to sevelamer hydrochloride as a generic version. There is therefore a potential need to revise the health economic modelling in NICE guideline CG157, and to consider sevelamer carbonate which was not included in the original guideline.

Chronic kidney disease in adults: investigations for chronic kidney disease – markers of kidney damage

For investigations for chronic kidney disease, we selected the individual patient data (IPD) meta-analysis by Tangri et al. (2016) for a full commentary. The reasons for selection were that it was highlighted by topic expert feedback, included a large overall sample, has a potential impact on guideline recommendations, and covers an emerging area of quantifying risk.

What the guideline recommends

NICE's guideline on chronic kidney disease in adults advises using the person's GFR and albumin:creatinine ratio (ACR) categories (see table 1 Classification of chronic kidney disease using GFR and ACR categories) to indicate their risk of adverse outcomes (for example, CKD progression, acute kidney injury, all-cause mortality and cardiovascular events) and discuss this with them. It does not currently make any recommendations on the use of equations for markers of kidney damage where CKD has been diagnosed.

Methods

The IPD meta-analysis by Tangri et al. (2016) evaluated the accuracy of kidney failure risk equations across a range of different geographic regions and patient populations. The equations, which were originally validated in a Canadian population, use demographic and laboratory data to predict progression of CKD to kidney failure. The four-variable equation included age, sex, estimated glomerular filtration rate (eGFR), and albumin to creatinine ratio. The eight variable equation used the same four variables plus laboratory measurements of calcium, phosphate, bicarbonate, and albumin.

Cohorts participating in the Chronic Kidney Disease Prognosis Consortium (CKD-PC) were selected for validation based on data availability. The CKD-PC is a collaborative research group integrating data from more than 70 cohorts spanning 40 countries and involving 2 million individuals from general, high-risk, or CKD populations. Available data on end stage renal disease from people in the cohorts participating in the CKD-PC was included. A random-effects model was used to perform meta-analysis across studies. The studies were selected to include patients with stages 3 to 5 CKD with an eGFR less than 60 ml/min/1.73 ml2 and an absence of kidney failure at baseline who had follow-up information on kidney failure, defined as treatment by dialysis or a kidney transplant.

The primary outcome was kidney failure, measured by treatment with dialysis or transplantation.

Calibration, defined as the difference between observed and predicted risk, was examined by comparing the observed 2-year and 5-year probability of kidney failure in individual cohorts to the predicted risk using the original and pooled risk equations.

In total, data from 31 cohorts were included covering 721,357 people with CKD, of which 23,829 experienced kidney failure. The median follow-up time was 4 years.

Results

Overall, the participants had a mean age of 74 years and a mean eGFR of 46 ml/min/1.73 ml2. In the North American cohorts, there were 41% of participants with albuminuria and a kidney failure incidence of 7.5 per 1,000 patient years. In the non-North American cohorts there were 40% of participants with albuminuria and a kidney failure incidence of 7.8 per 1,000 patient years.

The original risk equations were found to differentiate people who developed kidney failure from those who did not across all included cohorts (overall C statistic 0.90, 95% CI 0.89 to 0.92 at 2 years; C statistic 0.88, 95% CI 0.86 to 0.90 at 5 years).

Calibration was found to be poorer among some non-North American cohorts, where the original risk equations overestimated risk. A calibration factor was therefore added to address this. This reduced the baseline risk by 32.9% at 2 years and 16.5% at 5 years, resulting in improved calibration in non-North American cohorts at 2 years (12 of 15 studies p=0.04) and at 5 years (10 of 13 studies p=0.02).

Strengths and limitations

Strengths
  • The meta-analysis pooled data from a very large overall sample.

  • Meta-analysis was performed appropriately across studies using a random-effects model to allow for heterogeneity.

  • The study validated the risk equations in non-north American cohorts, including those in the UK setting, which are more representative of the population and setting considered in the guideline.

Limitations
  • The authors did not describe any search beyond the CKD-PC source. The studies within the CKD-PC may not reflect the entire set of existing studies, potentially introducing bias.

  • The equations only predict risk over 2 or 5 years, which may restrict their applicability, due to varying patterns of decline among patients.

  • Longer term risk of kidney failure beyond 5 years was not considered. This may affect other clinical decisions such as lifestyle modification, as acknowledged by the authors.

  • The authors acknowledged that the risk prediction equation cannot be used to predict risk among those with mild impairment in kidney function, because these patients were not included in the model.

  • The integrity of IPD was not fully reported, such as data consistency, baseline imbalance, and missing data.

  • Risk of bias assessment of the included cohort studies was not reported.

Impact on guideline

NICE guideline CG182 does not include recommendations for the use of risk equations in assessing risk of progression for people already diagnosed with CKD. This new evidence and topic expert feedback supports the use of the Tangri risk equations in predicting ESRD in CKD patients, and has a potential impact on the guideline to review the advice for determining the risk of progression and adverse outcomes.

Topic expert feedback further indicated that the equations have the potential to be used in primary and secondary care. In primary care, lower-risk patients could be managed without additional testing or treatment of CKD complications, whereas higher-risk patients could receive more intensive testing and early intervention.

However, the limitations of the model restrict its clinical value, and further evaluation and refinement may be necessary before implementation. For instance, the equations cannot be used currently to predict risk among those with mild impairment in kidney function, because these patients were not included in the model. Additionally the equations only predict risk over 2 or 5 years, which may restrict applicability, because the pattern of decline varies among patients. Topic expert feedback also highlighted that the equations cannot predict the risk of cardiovascular disease or death, and require validation in people with mild kidney disease. It was agreed that these factors will need to be considered during the update process.

Frequency of monitoring – defining progression by decline in eGFR

For frequency of monitoring, we selected the IPD meta-analysis by Coresh et al. (2014) for a full commentary. The reasons for selection were that it was recommended by topic expert feedback, was based on a very large data set, and highlights the potential value of smaller declines to indicate CKD progression over 1 to 3 years.

What the guideline recommends

NICE's guideline on chronic kidney disease in adults defines accelerated progression of CKD as a sustained decrease in GFR of 25% or more and a change in GFR category over 12 months, or a sustained decrease in GFR of 15 ml/min/1.73 ml2 per year (see recommendation 1.3.3).

Methods

The IPD meta-analysis by Coresh et al. (2014) examined the association of decline in eGFR with subsequent progression to ESRD, and explored the role of smaller declines in eGFR over longer periods as potential alternative outcomes for CKD progression.

Participants with available outcome data from cohorts in the CKD-PC with a repeated measure of serum creatinine were included. Studies with at least 10 events and participants aged over 18 years were included. ESRD cases before the baseline period were excluded. Baseline assessments in each cohort took place between 1975 and 2012.

In total, data were obtained from 35 cohorts (n=1,757,886) for analysis of mortality and change in eGFR.

The primary outcome was ESRD after the baseline period. This was defined as initiation of renal replacement therapy or death due to kidney disease other than acute kidney injury.

Results

Change in eGFR
  • Over a baseline period of 1 year (n=1,530,648) there were 12,344 ESRD events over a mean follow-up of 3.1 years.

  • Over a baseline period of 2 years (n=1,341,193), there were 8,532 subsequent ESRD events over a mean follow-up of 2.4 years.

  • Over a baseline period of 3 years (n=1,080,274) there were 5,159 subsequent ESRD events over a mean follow-up of 2.0 years.

In terms of ESRD risk, 52% of ESRD cases had a −30% change in eGFR over 2 years, whereas only 16% of ESRD cases reached a −57% eGFR change in the same period. A change in eGFR of −30% was associated with an adjusted Hazard ratio (HR) of ESRD of 5.4 (95% CI 4.5 to 6.4) for a baseline eGFR less than 60 ml/min/1.73 ml2. This result was similar over 1-year and 3-year baseline periods, as confirmed by sensitivity analysis. At a −57% decline, the risk was considerably higher (HR 32.1, 95% CI 22.3 to 46.3) for a baseline eGFR less than 60 ml/min/1.73 ml2.

Mortality
  • Over a baseline period of 1 year (n=1,757,886) there were 223,944 deaths from 27 cohorts.

  • Over a baseline period of 2 years (n=1,589,257) there were 158,603 deaths from 32 cohorts.

  • Over a baseline period of 3 years (n=1,259,477) there were 102,491 deaths from 34 cohorts.

From the mortality data analysed, there were considerably more people with an eGFR change of −30% or greater (cumulative prevalence of 7.1% (95% CI 6.6 to 7.7%) compared to a −57% change or greater (cumulative prevalence of 0.97% (95% CI 0.70 to 1.25%).

Compared to those with stable eGFR (0% change), the adjusted HR of all-cause mortality was higher with greater eGFR decline but was largely flat in the range of minimal decline (−10% change or less) or rise.

An association was found between a 30% decline in eGFR and a higher subsequent all-cause mortality risk, for both lower and higher baseline eGFR. The adjusted HR for a −30% change was 1.8 (95% CI 1.6 to 1.9) and this increased with greater declines.

Strengths and limitations

Strengths
  • The study included a very large overall sample size.

  • Meta-regression was conducted to assess variation in study characteristics and the robustness of findings. A large number of sensitivity analyses were also undertaken.

Limitations
  • The authors did not explain how the included cohort studies were selected from the CKD-PC or whether any attempts were made to search for other additional studies.

  • All of the included studies were observational in design, and methods for assessing risk of bias were not reported.

  • Standardisation of serum creatinine values may have varied across time and studies. Percent change in eGFR-based on a single first and single last eGFR is less precise than alternative designs where multiple measures are available at each time point.

  • Variation in design across cohorts introduced heterogeneity. However, the authors reported consistency across cohorts despite dramatic variation in design and populations.

Impact on guideline

The new evidence, based on a very large data set, highlights the potential value of smaller declines to indicate CKD progression over 1, 2 and 3 years. This is consistent with recommendations 1.3.3 and 1.3.5 in the NICE guideline on chronic kidney disease for 1-year follow up, which define increased risk of progression to ESRD as a sustained decrease in GFR of 25% or more over 12 months. The evidence reviewed for NICE guideline CG182 showed that a sustained drop in eGFR of 25% or a sustained drop of 15 ml/min/1.73 m2 over the period of 1 year was associated with an increased risk of mortality and progression to end stage kidney disease. There was more uncertainty of risk of progression with smaller declines in eGFR over longer periods.

Topic expert feedback indicated that the standard annual review in clinical practice fits with recommended 1-year follow up, as distinct from a 2- or 3-year follow up. However, additional topic expert feedback highlighted that CKD is a long-term condition, and that definitions of progression that are easy to apply in practice over a longer time period are needed.

The finding that a 30% change over 2 years is associated with a 5-fold increase in risk of ESRD was considered by topic experts to be very significant. It was considered that this has the potential to capture patients at high risk of ESRD who are likely to benefit from earlier referral and will not be highlighted as such currently. It may therefore impact on guideline recommendations and could potentially result in a change in practice.


This page was last updated: 24 April 2017