Allopurinol for chronic kidney disease: a systematic review

Article history

The research reported in this issue of the journal was funded by the HTA programme as project number 12/45/01. The contractual start date was in December 2012. The draft report began editorial review in May 2013 and was accepted for publication in January 2014. The authors have been wholly responsible for all data collection, analysis and interpretation, and for writing up their work. The HTA editors and publisher have tried to ensure the accuracy of the authors’ report and would like to thank the reviewers for their constructive comments on the draft document. However, they do not accept liability for damages or losses arising from material published in this report.

Copyright statement

© Queen’s Printer and Controller of HMSO 2014. This work was produced by Fleeman et al. under the terms of a commissioning contract issued by the Secretary of State for Health. This issue may be freely reproduced for the purposes of private research and study and extracts (or indeed, the full report) may be included in professional journals provided that suitable acknowledgement is made and the reproduction is not associated with any form of advertising. Applications for commercial reproduction should be addressed to: NIHR Journals Library, National Institute for Health Research, Evaluation, Trials and Studies Coordinating Centre, Alpha House, University of Southampton Science Park, Southampton SO16 7NS, UK.

Description of health problem

The term chronic kidney disease (CKD) is used to describe abnormal kidney function (or structure) and is defined according to the presence or absence of kidney damage and level of kidney function. There are no obvious symptoms of decreasing kidney function and, hence, diagnosis often occurs when patients present for other conditions related to CKD, such as cardiovascular disease (CVD) and diabetes. As CKD is often asymptomatic in its early stages, patients are often not diagnosed until the disease has reached an advanced stage. 1 People with CKD have an increased prevalence of CVD and are more likely to die from a CVD-related cause than they are to progress to end-stage renal disease (ESRD). 1

Traditionally, serum creatinine measurements were the mainstay for initial identification of CKD. Higher levels of creatinine indicate a lower glomerular filtration rate (GFR) which indicates decreased renal function. However, in the early stages of CKD, creatinine levels may be within the normal range. Partly for this reason, it is commonly recommended that all clinical biochemistry laboratories should provide an estimated GFR (eGFR) calculated using a formula to identify affected people. In the UK, laboratories typically use the Modification of Diet in Renal Disease (MDRD) formula,2 which takes into account a patient’s age, sex and ethnicity alongside serum creatinine levels. It is recognised that eGFR is not a perfect measure. It is most likely to be inaccurate in people at extremes of body type such as those with muscle wasting disorders, amputees and malnourished people. 1 The MDRD formula tends to underestimate normal function3 and eGFR calculations assume that the level of creatinine in the blood is stable. Importantly, eGFR is only valid in adults over 18 years old, is not valid for pregnant women and has not been validated for all ethnic groups. 1 Despite this, eGFR is considered to be the ‘gold standard’ measure and it enables CKD to be classified from stage 1 (minimal or no symptoms) to stage 5 (ESRD) ( Table 1 ). Stage 3 CKD is commonly referred to as moderate CKD.

No specific treatment has been found to unequivocally slow the worsening of CKD, but the control of blood pressure helps5 and there is emerging evidence that treatment with sodium bicarbonate also slows down progression. 6 Thus, where an underlying cause can be identified, treatment tends to focus on this underlying cause in order to attempt to slow progression of renal failure.

Evidence from a large data set of 177,570 patients studied over 25 years (Kaiser Permanente)7 has reported uric acid level to be associated with ESRD. Clinical evidence from randomised controlled trials (RCTs) is now emerging that allopurinol (Zyloric), a drug commonly used to treat abnormally elevated levels of uric acid in patients with gout, may slow the progression of CKD. 8 Abnormally elevated levels of uric acid in the blood are known as hyperuricaemia, which develops when there is an excess production of uric acid, decreased excretion of uric acid or a combination of both. Uric acid is associated with CVD, CKD, metabolic syndrome and hypertension. High concentrations of uric acid in the blood may result in gout. The extent to which uric acid is a cause, effect or indeed a coincidental factor for these diseases remains unknown. 9–12

Evidence is also emerging from RCTs that allopurinol may be effective in reducing cardiovascular risk in people with CKD. 13,14 In the absence of the use of an angiotensin-converting enzyme inhibitor (ACEI) or angiotensin II receptor blocker (ARB) to control hypertension, there is also RCT evidence of a significant worsening of hypertension, a significant decline in kidney function and a significant increase in the urinary excretion of transforming growth factor beta-1 following allopurinol withdrawal. 15

Epidemiology

A systematic review of 26 population studies conducted worldwide (but none from the UK) and published in 2008 found that CKD prevalence varied widely among the study populations investigated, strongly depending on how GFR was measured or calculated. 16 However, in all populations, prevalence rates increased with age with the median prevalence rate being 7.2% in persons aged 30 years and over, while in persons aged 64 years and over it was between 23.4% and 35.8%. 16 The 2009 Health Survey for England has estimated the prevalence of stages 3–5 CKD in persons aged 16 years and over in England to be 6% (male, 5%; female, 7%), ranging from 1% of males and 2% of females aged 16–54 years to 31% of males and 36% of females aged 75 years and over. 17 According to the New Opportunities for Early Renal Intervention by Computerised Assessment project, based on data from three primary care trusts in England, the age-standardised prevalence of stages 3–5 CKD was 8.5% (male, 5.8%; female, 10.6%). 18 It has been recently found in the USA that the prevalence of CKD stages 3–5 is increasing most rapidly in those aged 60 years and over. Between 1988 and 1994 and 2003 and 2006, data from the National Health and Nutrition Examination Survey suggested a rise from 18.8% to 24.5% in this age group, whereas the prevalence of CKD stages 3–5 in people between the ages of 20 years and 39 years remained consistently below 0.5%. 19 To date, incidence estimates have been less commonly reported in the literature.

The incidence and prevalence of hyperuricaemia are not commonly studied. In a relatively recent study of the US population in 2007–8, prevalence rates for hyperuricaemia were reported to be 21.4% (men, 21.2%; women, 21.6%) compared with 3.9% for gout (men, 5.9%, women, 2.0%) in the same study. 20 A recent systematic review conducted in China has estimated the prevalence of hyperuricaemia to be similar to the US rate for men (21.6%) but not women (8.6%). 21

Description of technology under assessment

Allopurinol was developed in 1956 for use as an adjuvant in chemotherapy for leukaemia. 22 It was discovered that allopurinol inhibits conversion of hypoxanthine to xanthine to uric acid, resulting in a decrease in the production of uric acid. Allopurinol was first licensed in 1966. Cumulative sales figures estimate over 22 million patient-years for allopurinol since launch (Mary Cockburn, Medical Information Service, Aspen, 2013, personal communication).

In the UK, allopurinol is indicated for all forms of hyperuricaemia not controllable by diet (adults), secondary hyperuricaemia of differing origin (adults, children and adolescents) and in clinical complications of hyperuricaemic states, particularly manifest gout, urate nephropathy and for the dissolution and prevention of uric acid stones (adults). 23 For adults, it is also approved for the management of recurrent mixed calcium oxalate stones in concurrent hyperuricaemia when fluid, dietary and similar measures have failed. 23 In children and adolescents it is also indicated for uric acid nephropathy during the treatment of leukaemia, hereditary enzyme deficiency disorders, Lesch–Nyhan syndrome and adenine phosphoribosyltransferase deficiency. 24 In the USA, allopurinol is indicated for the management of patients with signs and symptoms of primary or secondary gout (acute attacks, tophi, joint destruction, uric acid lithiasis and/or nephropathy). 24 It is also approved for the management of patients with leukaemia, lymphoma and malignancies who are receiving cancer therapy (which causes elevations of serum and urinary uric acid levels) and the management of patients with recurrent calcium oxalate calculi whose daily uric acid excretion exceeds 800 mg/day in male patients and 750 mg/day in female patients.

The dose required to lower serum uric acid to normal or near-normal levels varies with the type of patient and severity of the disease. In the UK, according to the Medicines and Healthcare products Regulatory Agency (MHRA), a dose of 100–200 mg/day is recommended for mild conditions, 300–600 mg/day for moderately severe conditions and 700–900 mg/day in severe conditions. 23 In the USA, 200–300 mg/day is recommended for people with mild gout, 400–600 mg/day for those with moderately severe tophaceous gout and a maximum dose of 800 mg/day for those with severe conditions. 24 Allopurinol may be taken once daily, preferably after food. 23,24 However, if the daily dosage exceeds 300 mg/day, gastrointestinal intolerance may occur and, so, a divided dosage regimen may be appropriate. 23,24 A maximum dose of 400 mg/day is recommended for children and adolescents for the treatment of malignant conditions. 23

As allopurinol and its metabolites are primarily eliminated by the kidney, accumulation of the drug can occur in renal failure. 23,24 Hence, the US Food and Drug Administration (FDA) recommends that the dose of allopurinol should be reduced in people with CKD. 24 The MHRA, which granted the UK licence, is more explicit and states that in patients with impaired renal function, a starting dose of 100 mg/day should be employed and only be increased if the serum/urinary response is unsatisfactory. 23 Both the MHRA and FDA recommend a dose of 200 mg/day for people with a creatinine clearance of 10–20 ml/minute and when the creatinine clearance is < 10 ml/minute the dose should not exceed 100 mg/day. With extreme renal impairment (creatinine clearance < 3 ml/minute) the interval between doses may also need to be lengthened. 23,24

Usual care and guidelines for the management of chronic kidney disease

The recent National Institute for Health and Care Excellence (NICE) guideline makes recommendations for the treatment of people with CKD. 1 These echo the recommendations of the UK Renal Association,25 which do not differ significantly from guidelines published elsewhere in the world. In general, the aim is to control blood pressure and complications such as diabetes in people with CKD. Hence, it is recommended that treatment should consist of both lifestyle support (e.g. dietary advice, encouragement to exercise and smoking cessation interventions) and drugs, in particular, drugs to control hypertension. In addition, patients are encouraged to avoid nephrotoxic medications such as non-steroidal anti-inflammatory drugs. The precise regimen of treatment will, therefore, depend to some extent on a patient’s albumin creatinine ratio and comorbidities.

In addition to treatment for underlying causes and comorbidities, patients with more advanced stages of CKD may also require treatments for anaemia and bone disease. 1 Severe CKD requires renal replacement therapy, which may involve a form of dialysis or transplantation of a new kidney in suitable patients.

Commonly used to treat hyperuricaemia associated with gout, Lesch–Nyhan syndrome and tumour lysis syndrome (typically seen during the treatment of leukaemia), allopurinol is not commonly used to treat asymptomatic hyperuricaemia as may occur in CKD. Indeed, there are no specific guidelines for using allopurinol to treat people with CKD. According to NICE guidelines issued in 2008, there is insufficient evidence to recommend the routine use of drugs to lower uric acid levels in people with CKD who have asymptomatic hyperuricaemia. 1 Hence, patients are normally only treated if they also have gout and/or hyperuricaemia. As noted above in Chapter 1, Description of technology under assessment, in the UK a starting dose of 100 mg/day is indicated for patients with impaired renal function. 23 Professional guidelines also recommend a medication review for all people with CKD: ‘patients on analgesics, certain β-blockers (including atenolol), digoxin and allopurinol may all need their dose reducing’. 26

According to data presented to NICE in 2008, allopurinol is the most commonly used urate-lowering drug in the UK (89% of gout treatments), with most cases (98%) using doses of ≤ 300 mg/day. 27 Similarly, it was recently reported that in the USA over 90% of urate-lowering prescriptions are for allopurinol but, again, allopurinol is rarely (< 5% of patients) prescribed at doses exceeding 300 mg/day. 28 However, such doses may not be effective for the treatment of gout. 29 While it is generally recommended that doses be increased from a low dose,23,24 it has been reported that, in practice, this rarely happens. 27 According to the British Society for Rheumatology, dosing should be based on achieving a target level of serum uric acid of < 300 µmol/l. 30

Safety profile of allopurinol

Although mostly well tolerated, reports of serious adverse events (SAEs) appeared soon after allopurinol was approved for use and the first death directly related to allopurinol [from toxic epidermal necrosis (TEN)] was reported in 1970. 31 In general, harmful effects from allopurinol have been ascribed to toxicity (bone marrow suppression), hypersensitivity (rash, hepatic injury, renal injury, eosinophilia, leucocytosis), drug interactions (including with ampicillin, warfarin and certain cytotoxic agents such as azathioprine and mercaptopurine), idiopathic reactions and severe reactions resulting from the normal therapeutic effects of allopurinol. 32 Because skin reactions to allopurinol can be severe, and sometimes fatal, treatment with allopurinol should be discontinued immediately if a rash develops.

The term allopurinol hypersensitivity syndrome (AHS) has been coined to encompass the plethora of features caused by hypersensitivity to allopurinol. Singer and Wallace32 have proposed diagnostic criteria for the definition of AHS and following a review of 101 cases of AHS, Arellano and Sacristan33 modified the suggested criteria. Drug rash with eosinophilia and systemic symptoms (DRESS) syndrome is another term proposed by Bocquet et al. 34 to describe drug hypersensitivity syndromes. Table 2 summarises the diagnostic criteria proposed by each of these authors.

Existing data support the coexistence of three mechanisms for AHS: immunological factors, genetic predisposition and accumulation of the drug. 33 People with the HLA-B*5801 allele who are treated with allopurinol have been identified to be at increased risk of severe cutaneous adverse reactions (SCARs) [AHS/DRESS or Stevens–Johnson syndrome (SJS) or TEN]. 35 It is largely in order to avoid AHS/DRESS that reduced doses of allopurinol, based on creatinine clearance, were first recommended for people with CKD. 36 However, it has been reported that some patients with gout receiving allopurinol at creatinine clearance-adjusted doses do not benefit because the dosage may become too low to be able to effectively reduce serum uric acid levels. 37

According to adverse event (AE) data submitted to the FDA between the first quarter of 2004 and the first quarter of 2012, DRESS, SJS and TEN are some of the most commonly reported AEs, albeit at an incidence of ≤ 4% of all AEs. 38 Other similarly relatively common AEs (at a frequency of 1–2%) include renal failure, pyrexia and rash which, as reported in Table 2, are associated with AHS/DRESS.

Aims and objectives of the current review

The primary aim of this systematic review was to consider the clinical effectiveness of allopurinol for people with CKD. Primarily, the review considers the evidence from RCTs in terms of effects on mortality, progression of CKD, cardiovascular risk and the effects on blood pressure and a number of disease markers. Given the importance of AEs (common and rare), a secondary aim was to consider the evidence from observational studies describing AE and quality-of-life data.

Search strategy

The following databases were searched for relevant published literature on 7 January 2013.

• MEDLINE (1946 to 7 January 2013)

• EMBASE (1974 to 28 December 2012)

• The Cochrane Library (Issue 1, 2013)

• ClinicalTrials.gov (7 January 2013).

The search strategy is presented in Appendix 1 . Search terms included a combination of index terms (for the disease) and free-text words for the technologies involved (generic and trade names of the drugs). No methodological filters were employed to limit results to a specific study design. No date or language limits were applied to the search strategy, although it was recognised that studies in languages other than English would not be translated and, therefore, would be excluded. However, by quantifying the number excluded in this way, a simple if crude assessment of the likelihood of publication bias was made. In addition to the search of electronic databases, bibliographies of retrieved citations were also examined. Two manufacturers [GlaxoSmithKline who are listed as the manufacturer of Zyloric in British National Formulary (BNF) No. 64 (September 2012)39 and Aspen who were identified as the current marketing holder of Zyloric by GlaxoSmithKline] were approached for data. All the identified literature was held in the EndNote X5 software package (Thomson Reuters, NY, New York City).

Study selection

The citations identified by the search strategy were assessed for inclusion through two stages. First, two reviewers (NF and GP) independently screened all relevant titles and abstracts obtained via electronic searching to identify potentially relevant studies for inclusion in the review (screening stage 1). Second, full-text copies of the potentially relevant studies were obtained and assessed independently by the same two reviewers using the inclusion criteria outlined in Table 3 (stage 2). Any disagreements between reviewers were resolved by discussion at each stage. Studies that did not meet the inclusion criteria at stage 2 were excluded.

As is evident from Table 3 , systematic reviews were neither explicitly included nor excluded at stage 2. We used identified systematic reviews pragmatically, primarily to identify any relevant studies that we had not identified from our own searches. Data were not extracted from the systematic reviews but from the primary studies included.

Quality assessment and data extraction

Data extraction forms were developed and piloted on a sample of included studies and adapted to reflect the nature of both RCT and observational studies related to AEs and quality of life. Data were extracted on study design, population characteristics and outcomes by one reviewer (NF) and independently checked for accuracy by a second reviewer (GP). Differences were resolved through discussion.

All included RCTs were assessed for risk of bias using criteria based on Centre for Reviews and Dissemination guidance for undertaking reviews in health care. 40 These criteria include key aspects of RCT design and quality. Data relating to quality assessment were extracted by one reviewer (NF) and independently checked for accuracy by a second reviewer (GP). Differences were resolved through discussion.

No universally accepted standardised quality assessment tool exists for use in observational studies. There are a multitude of observational study designs and, so, even where tools exist, applying them is problematic. For example, a review of non-randomised studies conducted by Deeks et al. 41 described seven types of observational study and identified 194 tools that could be or had been used to assess non-randomised studies. Quality assessment of observational studies was planned and piloted using a modified version of the Newcastle-Ottawa Scale;42 however, owing to the heterogeneous study designs of included studies it was impossible to extract or compare information in a meaningful and relevant manner. Therefore, we made the pragmatic decision not to quality assess the observational studies that were used to identify AE data.

Evidence synthesis

The results of the data extraction and quality assessment for each study were presented in structured tables and as a narrative summary. All summary statistics were extracted for each outcome and where possible, data were pooled. 43 Meta-analysis was carried out using a fixed-effects model using Review Manager (RevMan, The Cochrane Collaboration, London, UK) and separating outcomes by the time points at which they were reported (6, 9, 12 and 24 months). Mean differences and confidence intervals (CIs) were presented.

Heterogeneity was explored through consideration of the study populations, methods and interventions and by visualisation of results. Heterogeneity was assessed in statistical terms, by the chi-squared test for homogeneity and the I ² statistic. 44 Where I ² was greater than 50%, a random-effects model was employed. Owing to the limited number of studies included, planned subgroup analysis based on stage of CKD, age, sex, ethnicity, concomitant medication and comorbidities were not performed. Limitations in the available data also meant that the planned sensitivity analyses, including the use of meta-regression, were not conducted.

Number of studies identified and included

Electronic searching of databases resulted in 1973 references when duplicates were automatically removed in EndNote. A further 123 duplicates were removed manually; thus, we identified 1850 unique references for screening at stage 1.

Initial screening identified 77 references to obtain as full-text papers, to which inclusion criteria were applied. Twenty studies (reported in 22 citations) were included at stage 2, from which an additional 12 references were identified via hand-searching bibliographies of included studies. Thus, a total of 32 studies (34 references) were included in the review and 55 references excluded (see Appendix 2 ). All of the included references were published in English.

The most common reason for exclusion was that the publication reported only single case reports (< 10 patients). While a number of excluded references were not published in English, only two were primarily excluded for this reason. One of these publications, published in German, appeared to be an overview of the published literature and, so, would have been excluded. 45 The other publication, from Hungary,46 reported on a review of all patients (n = 11) with AHS referred to a dermatology department over a period of 4 years. This paper may have been included had it been possible to translate it. As this small study is the only non-English study that could have been included from our searches, any publication bias resulting from the exclusion of non-English studies is likely to be small.

Included citations consisted of one overview of the literature,8 one systematic review and meta-analysis,35 four RCTs13,14,47,48 and 26 other studies. 32,33,36,49–71 As shown in Figure 1 , data were only synthesised from 21 studies. This is because patients included in the reviews of case reports by Lang,60 as well as the studies of patients by Lupton64 and McInnes et al. ,65 were also included in the reviews of case reports by Hande et al. 36 and Singer and Wallace. 32 The data from these reviews of case reports were also included in Arellano and Sacristan. 33 Hence, only data from Arellano and Sacristan33 were synthesised.

FIGURE 1.

Flow diagram showing number of citations found from search and included in review.

We also contacted GlaxoSmithKline, the manufacturer of Zyloric listed in BNF No. 64 (September 2012),39 to enquire if they had additional data on allopurinol. They informed us they did not and directed us to Aspen, the current manufacturer of Zyloric. Aspen also informed us they did not have any data, but did provide us with a search of the literature (EMBASE and MEDLINE) that they had conducted using the search terms ‘chronic kidney’ and ‘allopurinol’ from 2000 onwards (Mary Cockburn, Medical Information Service, Aspen, 2013, personal communication). Their search identified 24 unique references, all of which had been identified by our searches.

As references reporting on AEs were identified via hand-searching bibliographies of included studies which appeared to be a relatively high number of references, we conducted an additional search of Ovid MEDLINE and Ovid OLDMEDLINE 1946 to 7 January 2013 with Daily Update on 4 March 2013 to ensure that we had not failed to identify any more potentially relevant studies. For this search, we used the following search term: ‘Allopurinol/ae [Adverse Effects]’ with no limits for language, year or type of study. This broad search yielded 651 citations. However, no additional studies were identified for inclusion in the review.

Reviews

Key aspects of both reviews8,35 identified by our electronic searches are presented in Table 4 .

The overview by Kabul and Shepler,8 in part, had a remit similar to our systematic review. In this review, the methods used to conduct the literature search were described (a search of electronic databases and a search of primary studies in the reference lists of identified reviews), although not in sufficient detail to replicate the searches (e.g. search terms are listed but not how these are combined). No detail is given as to how studies were selected, data extracted or quality assessed (if at all).

The systematic review and meta-analysis conducted by Somkrua et al. 35 aimed to present evidence for the genetic association between the HLA-B*5801 allele and allopurinol-induced SJS/TEN. Bibliographic databases and terms used to search these are described, as are limits (e.g. limited to English-language studies). The authors of this review also examined bibliographies of included citations in order to identify additional studies. The manner in which studies were selected, data extracted and quality assessed was described. Methods for conducting the meta-analysis (DerSimonian and Laird method72 in which statistical heterogeneity was assessed via the Q-statistic and I ² tests44) and reasons for study exclusion were also described. The authors report that from 94 records, 88 citations were assessed for eligibility (as there were six duplicates). Six studies were included in the analysis.

Randomised controlled trials

Four RCTs13,14,47,48 involving 257 participants are included in the review and summarised in Table 5 . Two of these13,48 were also included in the review by Kabul and Shepler. 8 We also identified two further RCTs by Kao et al. 14 and Shi et al. 47 The former RCT resulted in three publications, including a full article in a peer-reviewed journal published in 201114 and two conference abstracts from 2010. 73,74 One of these abstracts74 included analysis on fewer patients than those reported by the other two citations;14,73 therefore, some findings slightly differ. The data reported in our systematic review are all taken from the most recent publication in 2011. 14

As reported in Table 5 , included RCTs were all relatively small in size, including between 40 and 113 randomised patients and relatively short in duration, with trials varying from 6 months47 to a follow-up of 23.4 months (hereafter referred to as 24 months). 13 All the primary and secondary outcomes prespecified for our systematic review were assessed by at least one trial.

Two ongoing studies75,76 were also identified by our review; one of these (NCT01575379)76 was also identified in the review by Kabul and Shepler. 8 Two further studies [CKD-FIX (Controlled trial of slowing of Kidney Disease progression From the Inhibition of Xanthine oxidase)77 and PERL (Preventing Early Renal function Loss )78] not registered at ClinicalTrials.gov were identified via an external peer reviewer. Key features of the four studies are summarised in Table 6 . Two of the identified ongoing trials are also relatively small, intending to recruit between 60 and 80 patients, with an expected follow-up of 12 weeks and 24 months. The final data collection dates for primary outcome measures in these trials are June 2015 (NCT01575379)76 and September 2015 (NCT01228903)75 respectively. It should be noted that NCT0157537976 is a pilot study to inform PERL,78 which, like CKD-FIX,77 is planned to be much larger in size and much longer in duration. In CKD-FIX,77 620 patients with CKD will be randomised 1 : 1 to either allopurinol or control. Treatment will be blinded and patients will be followed up for 2 years. In PERL,78 400 patients with diabetes will be randomised 1 : 1 to either allopurinol or control and followed up for 5 years. The final data collection dates for these two trials are not known.

The methodological quality of the included trials, as assessed by risk of bias, is summarised in Table 7 .

Three trials13,47,48 reported randomisation, but only one trial47 provided any information on treatment allocation. Kao et al. 14 provided no details about either. Kao et al. 14 was the only trial that claimed to be double blind although, aside from the assessors, it is unclear from reading the published paper who was blinded. In Goicoechea et al. 13 and Shi et al. 46 assessors were reported to be blinded.

In terms of dropouts from the study and follow-up, all trials13,14,47,48 accounted for dropouts and followed up > 80% of patients. It was not always clear in Goicoechea et al. 13 how many patients were included in each analysis for all outcomes. However, the principal author did provide this information on request (Dr Marion Goicoechea, Hospital General Universitario Gregori Marañón, 2013, personal communication).

Only Goicoechea et al. 13 reported to have conducted analysis on an intention-to-treat (ITT) basis, although information provided on request revealed the analyses were conducted on patients who completed the study (Dr Marion Goicoechea, Hospital General Universitario Gregori Marañón, 2013, personal communication). According to the Cochrane Collaboration,79 this may still be considered to be an ITT analysis when data can be assumed to be missing at random. There is no evidence to suggest this was not the case. Kao et al. 14 and Siu et al. 48 analysed data only from patients who received and completed their treatment. However, this may also be considered to be an ITT analysis in the same way that the analysis in Goicoechea et al. 12 could be considered to be an ITT. Shi et al. 47 may also be considered to have employed an ITT analysis as there were no reported dropouts.

All trials reported a variety of outcomes, but it was not clear how many of these end points were protocol defined and how many were exploratory. Goicoechea et al. 13 stated that their primary objective was to analyse the effect of allopurinol in patients with moderate CKD in the reduction of inflammatory markers and renal disease progression and, hence, had multiple primary outcomes. In Kao et al. ,14 the primary outcome was left ventricular mass index (LVMI). The primary outcome investigated by Shi et al. 47 was a change in renal function determined by changes in eGFR and by Siu et al,48 renal progression as defined by one of four states: stable, worsening, ESRD or death. Only Kao et al 14 presented details of power calculations required to derive sample size.

Two trials reported the results of subgroup analyses. 47,48 Shi et al. 47 reported repeated values of eGFR during the 6-month therapy for patients with eGFR< 60 ml/minute/1.73 m². Siu et al. 48 divided the treatment group by uric acid level at the end of the study and correlation analysis was performed among three groups (0.2–0.299 mmol/l, 0.3–0.399 mmol/l and 0.4–0.45 mmol/l) with regard to percentage decrease in systolic blood pressure and serum creatinine levels. It is not clear if any of the subgroup analyses were determined a priori.

Observational studies

Because AE and quality-of-life data may not be adequately captured by RCTs (although some data were available on AEs and hospitalisations in the RCTs), studies reporting non-RCT evidence were also considered to be eligible for inclusion into the systematic review (see Table 3 ). Twenty-six additional citations reporting on AEs were included as summarised in Table 8 . Three of these were not published papers but abstracts only. 57,59,66 No studies which measured the quality of life of patients were identified.

For reasons highlighted in Chapter 2, Quality assessment and data extraction, no attempt to assess the quality of observational studies was made.

Study characteristics

All RCTs were published between 2005 and 2012. Two trials were conducted in Europe (Madrid, Spain,13 and unspecified locations in the UK14) and two trials were conducted in China (Guangdong47 and Hong Kong48). The Kao et al. trial14 was sponsored by the British Heart Foundation and the Shi et al. trial47 by grants from Chinese authorities (the Scientific and Technologic Committee of Guangdong Province, Guangdong Province Health Office and Guangdong Natural Science Foundation); and the other two trials did not report on financial support.

As reported in Table 9 , although each trial had its own eligibility criteria, in general, similar patients were included across the trials with the exception that Shi et al. 47 included people with immunoglobulin A nephropathy (IgAN), serum creatinine < 265 μmol/l and excluded patients receiving ACEIs or ARBs. Patients with more progressive forms of CKD, many of whom were receiving concomitant antihypertensives at baseline, were included in the other three trials. 13,14,48

In Goicoechea et al. ,13 patients were given a 100 mg/day dose of allopurinol. In Kao et al. ,14 patients also started on 100 mg/day, but this dose was increased to 300 mg/day if tolerated. In Shi et al. ,47 patients started on 200 mg/day or 300 mg/day, and in Siu et al. 48 100 mg/day to 200 mg/day. In the last two trials,47,48 the starting dose depended on their serum creatinine levels and, in both trials, the dose was adjusted according to uric acid levels.

In total, 257 patients were analysed in the four RCTs:13,14,47,48 130 patients receiving allopurinol and 127 receiving control (usual treatment/placebo).

Participant characteristics

Key baseline demographic characteristics of participants are presented in Table 10 and key clinical markers at baseline are presented in Table 11 .

As can be seen from Table 10 , all of the RCTs included patients who were being treated for CKD; none of these studies stated that they included patients with asymptomatic hyperuricaemia. The mean age of patients in all trials was relatively similar between treatment and control groups. However, it was noticeable that in the mean age of patients across trials varied, with the mean age being > 70 years in two trials13,14 and being < 43 years in the other two trials. 47,48 There were proportionately more males in the allopurinol group than in the control group in the Kao et al. 14 and Shi et al. 47 trials, whereas in Siu et al. 48 there were proportionately more females. Conversely, there was a slight minority of males in the control groups in Kao et al. 14 and Shi et al. 47 and a slight majority of males in the control group in Siu et al. 48 Goicoechea et al. 13 did not report the sex of trial participants in the published paper, but provided these data on request (Dr Marion Goicoechea, Hospital General Univesitario Gregori Marañón, 2013, personal communication). The majority of patients were male in this trial, evenly spread between the allopurinol and control groups.

Although no differences between groups were reported to be statistically significant, an examination of renal pathology, comorbidities and concomitant medication suggests some marked differences between groups in each of the trials ( Table 10 ). These included differences in interstitial nephropathy, ischaemic cardiopathy, diabetes mellitus and concomitant medication in Goicoechea et al. ,13 diabetic nephropathy and concomitant medication in Kao et al. 14 and the patients with IgAN, focal segmental glomerulosclerosis, hypertension and concomitant medication in Siu et al. 47

Comparisons across trials are problematic because of the manner in which data were recorded. For example, no patient is reported to have hypertension in Goicoechea et al. 13 (although blood pressure levels were elevated in all trials) whereas it is described as a renal pathology in around half of the patients in Kao et al. 14 and as a comorbidity in 33% of all patients in Shi et al. 47 and 78% of all patients in Siu et al. 48 There also appears to be differences in concomitant medications received across trials. While the majority of patients in all trials received ACEIs or ARBs, the proportions of patients receiving a number of other concomitant medications such as diuretics and statins did appear to be more variable although it is unclear if the same concomitant medications were recorded by all trials. Certainly, based on the reporting of the use of ACEIs or ARBs, differences in the manner of reporting are apparent: Goicoechea et al. 13 appear to include ACEIs and ARBs under the term renin–angiotensin–aldosterone system blockers, Kao et al. 14 combine ACEIs and ARBs together whereas Siu et al. 47 report the use of ACEIs and ARBs separately. Shi et al. 47 do not report on concomitant medication, although it is known that patients receiving ACEIs or ARBs at study enrolment were excluded.

No trial reported the ethnicity of its participants. However, two studies were conducted in Europe13,14 and two were conducted in China. 47,48 It may be assumed, therefore, that the ethnicity of the participants were European and Asian respectively.

Only the trial by Kao et al. 14 presented data on smoking status, an important variable for cardiovascular risk. Smoking status was similar between the two groups (smoker: 15% vs. 12%; ex-smoker: 26% vs. 28%; non-smoker: 59% vs. 60% for allopurinol vs. control respectively; p = 0.118). The only other data relating to a patient’s lifestyle were reported by Goicoechea et al. 13 who noted that all patients were advised about their diet.

Perhaps the most marked differences across trials are in terms of key clinical parameters ( Table 11 ) where differences in baseline eGFR, serum creatinine and diastolic blood pressure are apparent between Shi et al. 47 and the other three trials. 13,14,48 In Shi et al. ,47 the mean eGFR was 69.5 ml/minute/1.73 m² [standard deviation (SD) 26.5 ml/minute/1.73 m²] in the allopurinol group and 63.6 ml/minute/1.73 m² (SD 27.5 ml/minute/1.73 m²) in the control group, suggesting many patients with mild CKD were included. These levels were markedly higher than in Goicoechea et al. 13 and Kao et al. 14 Similarly, levels of serum creatinine were markedly lower in Shi et al. 47 than in Goicoechea et al. 13 or Siu et al. 48 In addition, diastolic blood pressure appeared to be higher in both groups in Shi et al. 47 than in patients in the other three trials. 13,14,48 Levels of systolic blood pressure were, on the whole, relatively similar across all trials. 13,14,47,48 Three trials13,14,47 reported uric acid levels, which were similar across all trials and between treatment groups.

The only reported statistical difference between groups in any of the presented clinical markers at baseline was reported in diastolic blood pressure by Kao et al. 14 In this trial, diastolic blood pressure was higher in the control group (p = 0.036). However, in Siu et al. 48 an even larger difference in diastolic blood pressure between treatment groups was reported, but this was not reported to be significant (p = 0.25). Mean diastolic blood pressure was highest in both treatment groups in the trial of IgAN patients. 47 Levels of other clinical parameters were, on the whole, similar between groups.

On balance, we decided that the participant characteristics were sufficiently similar across three trials13,14,48 for these to be considered together for inclusion in meta-analyses. It was considered that Shi et al. ,47 on the other hand, was qualitatively different to these other three trials. 13,14,48 Therefore, we conducted sensitivity analyses by including and excluding this trial from the meta-analysis.

Results of evidence synthesis

Progression of chronic kidney disease

Changes in eGFR, were reported by three trials13,14,47 and are presented in Table 12 and Figure 2 . As noted in Table 12 , the method of calculating eGFR was not always the same across studies. No significant differences over time were reported in any study and no significant differences were reported between groups at any point in time except at the end of the study in Goicoechea et al. 13 (p < 0.001) Interestingly, a significant inverse correlation between uric acid levels and eGFR (r = –0375; p = 0001) was also reported at 24 months by Goicoechea et al. 13

TABLE 12 - Change in eGFR reported from RCTs

Estimated glomerular filtration rate calculated using the MDRD formula by Goicoechea et al. 13 and Chinese-abbreviated MDRD formula by Shi et al. ;47 method not specified by Kao et al. 14

p-values not reported.

Study	Treatment group	eGFR, mean (SD) (ml/minute/1.73 m2)a					Statistical test for change over time
		Baseline	6 months	9 months	12 months	24 months
Goicoechea et al. 201013	Allopurinol (n = 56)	40.8 (11.2)	41.1 (12.9)	Not measured	41.1 (13.2)	42.2 (13.2)	Not significantb
	Control (n = 57)	39.5 (12.4)	37.2 (14.3)		35.6 (13.4)	35.9 (12.3)	Not significantb
						Difference between groups, p < 0.001
Kao et al. 201114	Allopurinol (n = 27)	44 (11)	Not measured	Difference + 0.2 (6.9)	Not measured	Not measured	Not reported
	Control (n = 26)	46 (9)		Difference + 0.2 (5.5)
		Difference between groups, p = 0.427		Difference between groups, p = 0.997
Shi et al. 201247	Allopurinol (n = 21)	69.5 (26.5)	73.2 (34.8)	Not measured	Not measured	Not measured	p = 0.2
	Control (n = 19)	63.6 (27.5)	68.9 (36.6)				p = 0.9
		Difference between groups, p = 0.6	Difference between groups, p = 0.2

FIGURE 2.

Estimated glomerular filtration rate (eGFR) at 6, 12 and 24 months and difference in eGFR at 9 months (a) including Shi et al. 47 and (b) excluding Shi et al. 47

Siu et al. 48 reported patients who had stable and worsening of renal function, defined, respectively, as an increase in serum creatinine level at the end of study by < 40% compared with baseline and by > 40% compared with baseline, but not yet requiring dialysis. It was reported that significantly more patients in the control group showed deterioration in kidney function at the end of the study (stable disease, 84% vs. 54%; worsening disease: 12% vs. 42%, for allopurinol and control respectively; p = 0.015).

Regarding progression to ESRD, no trial reported any patient requiring transplantation. Goicoechea et al. 13 reported that one patient in each of the allopurinol and control groups required dialysis, as did Siu et al. 48

Change in blood pressure

Change in blood pressure was reported by all four RCTs. 13,14,47,48 The findings from the individual trials in terms of systolic and diastolic blood pressure are presented in Tables 14 and 15 .

TABLE 14 - Change in systolic blood pressure reported from RCTs

Study	Treatment group	Systolic blood pressure (mmHg), mean (SD)					Statistical test for change over time
		Baseline	6 months	9 months	12 months	24 months
Goicoechea et al. 201013	Allopurinol (n = 56)	147 (20)	145 (17)	Not measured	142 (16)	144 (15)	Not reported
	Control (n = 57)	146 (17)	144 (16)		141 (15)	143 (13)
Kao et al. 201114	Allopurinol (n = 27)	139 (14)	–4.9 (17.64)	–6.9 (14.4)	Not measured	Not measured	Not reported
	Control (n = 26)	145 (18)	–8.73 (21.36)	–5.1 (15.1)
		Difference between groups, p = 0.164	Difference between groups, p = 0.701	Difference between groups, p = 0.644
Siu et al. 200648	Allopurinol (n = 25)	138 (20)	Not measured	Not measured	127 (21)	Not measured	p = 0.02
	Control (n = 26)	135 (19)			135 (32)		p = 0.90
		Difference between groups, p = 0.68			Difference between groups, p = 0.31

TABLE 15 - Change in diastolic blood pressure reported from RCTs

Study	Treatment group	Diastolic blood pressure (mmHg), mean (SD)					Statistical test for change over time
		Baseline	6 months	9 months	12 months	24 months
Goicoechea et al. 201013	Allopurinol (n = 56)	77 (11)	76 (9)	Not measured	74 (9)	73 (10)	Not reported
	Control (n = 57)	76 (13)	77 (9)		75 (8)	74 (10)
Kao et al. 201114	Allopurinol (n = 27)	70 (8)	–1.85 (11.64)	–3.3 (8.6)	Not measured	Not measured	Not reported
	Control (n = 26)	75 (8)	–4.15 (12.95)	–2.5 (9.1)
		Difference between groups, p = 0.036	Difference between groups, p = 0.498	p = 0.741
Siu et al. 200648	Allopurinol (n = 25)	79 (10)	Not measured	Not measured	75 (10)	Not measured	p = 0.12
	Control (n = 26)	71 (14)			71 (13)		p = 0.89
		Difference between groups, p = 0.25			Difference between groups, p = 0.21

In Shi et al. ,47 diastolic blood pressure levels appeared to be higher in both groups than the other three studies. Changes in blood pressure in this trial were confined to the very small subgroup of patients with normal blood pressure at baseline (n = 17) and were reported as mean arterial pressure. In this subgroup, a significant reduction in mean arterial pressure over time was reported for the allopurinol group [n = 9; baseline: 92.9 mmHg (SD 10.1 mmHg); 6 months: 83.7 mmHg (SD 4.5 mmHg); p < 0.01] but not the control group [n = 8; baseline: 93.7 mmHg (SD 5.4 mmHg); 6 months: 93.8 mmHg (SD 4.1 mmHg); p = 0.9]). A strong correlation was observed between serum uric acid and mean arterial pressure (r = 0.388; p < 0.001).

As previously shown in Table 11 and also evident from Table 14 , baseline levels of blood pressure were reasonably similar across three trials13,14,48 which, as highlighted in Chapter 4, Participant characteristics, appeared to include broadly similar types of patients and so these were included in meta-analyses. The findings from these meta-analyses are presented in Figures 3 and 4 respectively. Overall, no significant differences between the treatment groups were found at any time point, both measures of blood pressure remaining largely unaltered over time.

FIGURE 3.

Systolic blood pressure at 6, 12 and 24 months and difference in systolic blood pressure at 9 months.

FIGURE 4.

Change in diastolic blood pressure at 6, 12 and 24 months and difference in diastolic blood pressure at 9 months.

Secondary outcomes

The following secondary outcomes were specified in the systematic review protocol:

• change in uric acid levels

• change in serum creatinine levels

• change in albuminuria levels

• number of patients with endothelial dysfunction

• number of patients with left ventricular hypertrophy

• change in number of blood pressure medications

• AEs

• quality of life.

Change in uric acid levels

Uric acid level was the only secondary outcome measured in all of the included RCTs. As summarised in Table 16 , changes over time were reported to be significantly improved in the allopurinol group by all four trials that reported this measure. 13,14,47,48 Pooled data from Goicoechea et al. 13 and Shi et al. 47 resulted in a significant difference in uric acid levels favouring allopurinol at 6 months [mean difference –0.07 mmol/l (95% CI –0.14 to –0.01 mmol/l)]. Pooled data from Goicoechea et al. 13 and Siu et al. 48 show a borderline significant improvement at 12 months [mean difference –0.17 mmol/l (95% CI –0.33 to 0.00 mmol/l)] ( Figure 5 ). However, a large amount of statistical heterogeneity was observed.

TABLE 16 - Changes in uric acid levels reported from RCTs

Reported in mg/dl in original paper and converted to mmol/l by multiplying by 0.0595.

Study	Treatment group	Uric acid levels (mmol/l), mean (SD)					Statistical test for change over time
		Baseline	6 months	9 months	12 months	24 months
Goicoechea et al. 201013	Allopurinol (n = 56)	0.47 (0.12)a	0.37 (0.09)a	Not measured	0.36 (0.11)a	0.36 (0.07)a	p < 0.001
	Control (n = 57)	0.43 (0.10)a	0.41 (0.10)a		0.44 (0.12)a	0.44 (0.1)a	Not significant
						Differences between groups, p < 0.016
Kao et al. 201114	Allopurinol (n = 27)	0.47 (0.7)a	Not measured	–0.18 (0.08)a	Not measured	Not measured	Not reported
	Control (n = 26)	0.46 (0.07)a		+0.02 (0.06)a
				Difference between groups in change over time, p < 0.001
Shi et al. 201247	Allopurinol (n = 21)	0.47 (0.07)a	0.33 (0.04)a	Not measured	Not measured	Not measured	p < 0.001
	Control (n = 19)	0.46 (0.07)a	0.44 (0.09)a				p = 0.03
		Difference between groups at start of study, p = 0.7	Difference between groups at end of study, p < 0.001
Siu et al. 200648	Allopurinol (n = 25)	0.58 (0.07)	Not measured	Not measured	0.35 (0.06)	Not measured	p < 0.001
	Control (n = 26)	0.59 (0.10)			0.60 (0.10)		Not significant

FIGURE 5.

Change in uric acid levels at 6, 9, 12 and 24 months (a) including Shi et al. 48 and (b) excluding Shi et al. 47

Significant differences in uric acid levels were also reported in Kao et al. 14 at 9 months and by Goicoechea et al. 13 at 24 months. Interestingly, Shi et al. 47 reported significant differences over time (6 months) in the control group as well as in the allopurinol group. However, differences between groups at the end of study were also reported to be significantly in favour of allopurinol.

Siu et al. 48 report results of a subgroup analysis in which the treatment group was divided into three categories according to uric acid level at the end of the study (0.2–0.299 mmol/l; 0.3–0.399 mmol/l; 0.4–0.45 mmol/l). In these very small subgroups, no clinical correlation could be shown for the three categories of target uric acid levels in relation to change in systolic blood pressure (p = 0.24). Similarly, no clinical correlation could be shown in relation to change in serum creatinine level (p = 0.32).

Change in serum creatinine levels

Siu et al. 48 was the only trial to report on changes in serum creatinine. In the allopurinol group, these increased over 12 months from 146 µmol/l (SD 56 µmol/l) to 176 µmol/l (SD 81 µmol/l). This increase was not statistically significant (p = 0.15). In the control group, serum creatinine levels increased from 164 µmol/l (SD 61 µmol/l) to 255 µmol/l (85 µmol/l), a change over time that was statistically significant (p = 0.003). However, the difference between groups at 12 months was not significant (p = 0.08).

Change in albuminuria levels

Change in albuminuria levels were reported only by Goicoechea et al. 13 In the allopurinol group the median decreased from 36 mg/day at baseline to 16 mg/day at 12 months. In the control group the median (interquartile range) value increased from 32 mg/day to 51 mg/day. Despite these apparent large differences over time and between groups, no significant differences were reported either in the change over time or between groups (p-values not presented).

Number of patients with endothelial dysfunction and number of patients with left ventricular hypertrophy

End points relating to endothelial dysfunction and left ventricular hypertrophy were addressed only by Kao et al. 14 In this trial, the primary outcome was change in LVMI at 9 months which in the allopurinol group was –1.42 g/m² (SD 4.67 g/m²) and in the control +1.28 g/m² (4.45 g/m²), a significant difference (p = 0.036). Endothelial dysfunction was measured by a change in flow-mediated dilatation (FMD) of the brachial artery. In the allopurinol group, the percentage change in FMD response to hyperaemia was 1.72% (SD 2.95%) at 6 months and 1.26% (SD 3.06%) at 9 months, compared with 0.03% (SD 2.84%) and 1.05% (SD 2.84%) respectively in the control group. At both points in time, differences between groups were statistically significant (p = 0.053 and p = 0.009 respectively). The authors state that no correlations were found between urate levels (either its baseline or its change) and the changes seen in LVMI or FMD.

Change in number of blood pressure medications

Three of the trials14,47,48 reported change in blood pressure medications ( Table 17 ). In Kao et al. ,14 a greater proportion of control group patients commenced medication and a greater proportion of allopurinol group patients stopped medication. However, no significant differences between groups were reported. In Siu et al. ,48 the number of patients who were receiving antihypertensives at the end of the study remained the same as that at the beginning of the study in both groups (although proportionately, there was an increase in the allopurinol group and a decrease in the control group); no significant differences between groups were reported at the beginning or end of study. In Shi et al. ,47 however, significant differences between treatment groups were reported at the end of the study (p = 0.003). In this trial, no patients commenced antihypertensive medication in the allopurinol group but 78% reduced treatment (including one patient who stopped) whereas in the control group, no patient reduced their medication but 33% increased it. However, the numbers of patients receiving medication in this trial are very small.

TABLE 17 - Change in blood pressure medications as reported in RCTs

Study	Measure of blood pressure medication
Goicoechea et al. 201013	Not reported
Kao et al. 201114	Allopurinol Antihypertensives commenced: n = 2 (7%) Antihypertensives stopped: n = 5 (18%)
	Control Antihypertensives commenced: n = 5 (19%) Antihypertensives stopped: n = 2 (8%)
	Difference between groups in commencing medication, p = 0.150
	Difference between groups in stopping medication, p = 0.258
Shi et al. 201247	Allopurinol Antihypertensive drugs reduced: n = 7 (78%) No change: n = 2 (22%) Antihypertensive drugs increased: n = 0
	Control Antihypertensive drugs reduced: n = 0 No change: n = 6 (67%) Antihypertensive drugs increased: n = 3 (33%)
	Differences between groups, p = 0.003
Siu et al. 200648	Allopurinol ACEI, beginning of study: n = 15 (58%) ACEI, end of study: n = 14 (56%) ARB, beginning of study: n = 8 (31%) ARB, end of study: n = 9 (36%)
	Control ACEI, beginning of study: n = 14 (56%) ACEI, end of study: n = 15 (58%) ARB, beginning of study: n = 5 (20%) ARB, end of study: n = 4 (15%)
	Difference between groups at start of study: ACEI, p = 0.90; ARB, p = 0.38
	Difference between groups at end of study: ACEI, p = 0.77; ARB, p = 0.11

Study characteristics

Adverse event data were available from 25 studies of patients being treated for various reasons, not just CKD: four RCTs13,14,47,48 and 21 observational studies (four prospective observational cohort studies,50,62,67,69 a hybrid prospective/retrospective study,63 eight retrospective case–control studies,51,53–56,59,68,70 six retrospective cohort studies,49,52,58,61,66,71 one review of case reports33 and a pharmacovigilance study57). In total, the 25 studies reported 2629 patients treated with allopurinol (including 433 patients chosen as allopurinol-tolerant controls) and 200 patients not treated with allopurinol. Study characteristics are summarised in Appendix 3, Table 23 .

Participant characteristics

In terms of the indications for allopurinol, these varied across and, in some cases, within studies. In seven studies,13,14,47,48,50,62,67 all patients had CKD and, in three of these,50,62,67 all patients also had gout. Gout was also an indication in nine other studies. 52,55,58,61,66,68–71 Hyperuricaemia was stated to be an indication in six studies,53–55,59,63,70 while patients with asymptomatic hyperuricaemia were known to be included in four others. 33,58,61 Two studies56,57 provided no information on the indication for allopurinol.

The mean age of patients varied considerably across the trials from 39 years in the allopurinol group of Shi et al. 47 to 74 years in Atzori et al. 49 The range of ages in studies also varied considerably across studies; the largest range was in Stamp et al. ,68 in which the age ranged between 23 and 92 years. Overall, the youngest patient was 8 years62 and the oldest patient was 96 years. 49

The proportion of males included in studies also varied by study. Most studies reported that the majority of patients were male, with males being a minority in only three studies 48,49,61 For studies that included a control group,13,14,47,48,51,53–55,59,68,70 the proportion of males was well balanced in both arms in only two. 13,68 The majority of studies had a greater proportion of males in the control group48,51,53–55,59,70 than the allopurinol group. 14,47

Thirteen studies50–56,58,61,63,68–70 explicitly reported the ethnicity of their subjects, although the ethnicity is implied (to be European) in another. 49 Four of these studies50,52,68,69 were conducted in New Zealand, where the ethnicity of patients varied across studies. In Stamp et al. ,69 the majority (82%) of patients were of European ancestry and this ethnic group also made up a large proportion of cases (48%) in Stamp et al. 68 In this latter study, there were significant differences between cases and controls, largely as a result of differences in the proportion of patients of Chinese origin (19% cases vs. < 1% controls) and Maori and Pacific Islanders (30% cases vs. 48% controls). While only 14% of patients were Maori and Pacific Islanders in Stamp et al. ,69 these accounted for 72% of patients in Dalbeth and Stamp (Maori: 26%; Pacific: 46%). 52 Reporting of ethnic origin was incomplete in Bowie et al. 50 Eight studies were conducted in Asian countries and consisted of patient populations primarily (> 90%) of Singaporean-Chinese origin in two studies,58,61 Han-Chinese in two studies, 51,53 Korean in two studies,54,55 and, in the remaining two, Thai70 and Japanese. 56 The remaining study which reported on ethnicity was conducted in Europe63 and the majority (> 85%) were described as European.

Baseline demographic characteristics are summarised in Appendix 3, Table 24 .

Adverse event data

A broad range of AE data were collected, as summarised in Table 18 .

Principal findings

We have conducted a systematic review of the literature to summarise the clinical effectiveness of allopurinol for people with CKD. Evidence for efficacy was determined solely from RCTs. The majority of patients in all trials started at a dose of 100mg/day as recommended by the MHRA. 23 This ensures that the trials are relevant to current clinical practice. All trials included patients with stage 3 CKD, although Shi et al. 47 appeared to include a majority of patients with milder CKD. Therefore, we conducted sensitivity analyses by including and excluding this trial from the meta-analysis. This did not result in differences in the overall findings.

None of the trials reported a worsening of CKD progression over time in the allopurinol group. Only Goicoechea et al. 13 reported a significant improvement in mean eGFR compared with the control group. It is therefore not possible to conclude that allopurinol is effective in decreasing the progression of the disease.

Only Goicoechea et al. 13 explicitly considered cardiovascular events as an end point. The findings from this study suggest that allopurinol may result in fewer cardiovascular events. However, such findings clearly need replicating in other studies.

No changes over time or between groups were reported in systolic or diastolic blood pressure, a risk factor for both CKD and CVD, from our meta-analysis of three trials. However, examination of the results also needs to consider that patients in the trials were receiving other treatments for their hypertension. It is therefore worth exploring changes in antihypertensive use further in future studies.

No significant differences were reported for the majority of other risk factors for CVD measured by the trials. However, significant (but small) improvements in left ventricular hypertrophy and endothelial dysfunction were reported in the allopurinol group in Kao et al. 14 Such improvements would unlikely have resulted in immediate differences in cardiovascular events in the short to medium term and their effects on CKD progression are unclear.

Given allopurinol is a drug which is designed to lower urate levels, the improvements in the allopurinol but not control groups in reducing uric acid levels in three trials13,14,48 is reassuring, suggesting that the drug was being taken at adequate doses and adequately adhered to. The improvement in uric acid levels reported in the control group in Shi et al. ,47 on the other hand, is surprising.

It is not clear how changes in uric acid levels are related to improvements in the other end points, if at all. Interestingly, in Siu et al. ,48 in a subgroup analyses, no clinical correlation could be shown between uric acid levels and change in serum creatinine or blood pressure while linear regression analysis conducted by Kao et al. 14 reported endothelial function to be independently related to uric acid levels. However, in both trials, subgroups were extremely small. Therefore, the mechanism of action through which allopurinol could result in the observed improvements in efficacy discussed above remains unexplained, mirroring the conundrum of whether uric acid is a cause, effect or indeed a coincidental factor for CKD and CVD. 9–12

Given the importance of AEs (both common and rare) and quality of life, in addition to the RCT evidence,13,14,47,48 we also considered evidence from 21 observational studies describing AE data;33,49–59,61–63,66–71 no studies were identified reporting on quality of life. Patients included in these studies had a variety of diseases (e.g. gout) and it is unclear whether or not the incidence and type of AEs and SAEs may therefore differ in people with CKD. Nevertheless, findings suggest that AEs may occur in 9.2% of patients and that no more than 2% of all patients who take allopurinol will experience a SCAR52 or, more specifically, AHS. 54 The evidence appears to be conflicting whether or not SCARs are dose related. A particularly interesting finding from the studies of SCARs was the association of the HLA-B*5801 allele with SCARs. However, as many patients with the HLA-B*5801 allele do not develop SCARs and, since it is not clear how to prevent SCARs, the utility of HLA-B*5801 testing is likely to be limited. This is particularly true in predominantly non-Asian populations, in which this allele is particularly rare.

Strengths and limitations

The main strength of our review is that we have systematically identified and presented evidence from RCTs assessing the efficacy of allopurinol in people with CKD. In addition, we have also examined evidence for the incidence of allopurinol-related AEs reported in the literature.

There are a number of limitations to our review. First, while the methodological quality of the included trials is acceptable, there are substantive limitations. None of the included studies included reported concealment of allocation, one of the greatest risks to study validity. There are relatively few (< 115) patients enrolled in any given trial, which limits the applicability of the results to the larger population in clinical practice. Even if a comprehensive meta-analysis could have been carried out, the total population of patients is only 257. In addition, the populations are clinically heterogeneous in terms of stage of disease, age, ethnicity and concomitant therapies. Trials other than Goicoechea et al. 13 are also limited by their relatively short follow-up period.

Arguably a greater limitation is that important confounding variables were not adequately described or controlled for. For example, information on diet and alcohol intake, which could influence all end points, is lacking. Although data are provided on concomitant medication, more insight may have been gained if more information had been provided on drugs which have been reported to lower levels of uric acid, such as losartan potassium,80 which may or may not have been taken by patients in studies (ARBs were permitted in three trials13,14,48). No trial attempted to adjust for these in factors their analysis.

In terms of end points, it would have been preferable if all trials had included eGFR to measure CKD progression as this is considered to be the best currently available measure. This would have enabled comparison of data from all included trials. With regard to cardiovascular risk, the end points measured by three trials,13,46,47 are surrogate measures and, in this respect, it is a limitation that only Goicoechea et al. 12 reported on cardiovascular events.

Finally, there is a dearth of studies examining information on quality of life using validated measures. In the absence of studies directly measuring quality of life in patients with CKD commencing allopurinol, we can only assume that the quality of life is not impaired unless they experience a SAE and/or are hospitalised. Studies utilising validated quality-of-life measures in people with CKD being treated with allopurinol would therefore be illuminating.

Given the above, the results from the ongoing CKD-FIX77 trial are eagerly awaited. Additional RCT evidence addressing many of the limitations of the studies identified by this systematic review and designed along the lines of CKD-FIX77 could also be warranted. The challenges of conducting such a trial have recently been explored by Badve et al. ,81 who argue that a pilot study should initially be conducted at a few sites in a single country. Based on power calculations, they estimate that this would need to include 620 patients (which is the same number of patients being enrolled into the CKD-FIX77 trial). Badve et al. 81 also argue that the pilot study should then be followed by an international multicentre trial of 7470 patients.

Alongside RCT evidence, additional supporting data are required from observational studies of patients with CKD and using allopurinol. If derived from, or linked to, a database or register such as the Kaiser Permanente data set used to assess risk factors for ESRD,7 such studies could collect data on the relationship between allopurinol and a number of risk factors and outcomes (efficacy and AEs).

Suggested research priorities

Given the uncertainties in the evidence base highlighted above, there is a need for a further RCT to be conducted, comparing allopurinol with usual care. Ideally, a double-blind trial design should be employed and, hence, usual care will also include placebo. The dose of allopurinol should be in accordance with guidelines for current practice. Ideally, such a trial would also be adequately powered to assess for CKD progression and also to consider stratification of key factors such as age, ethnicity, stage of CKD, comorbidities and concomitant medication (particularly other urate-lowering medications). However, the feasibility of enrolling enough patients to be included in a suitably large trial may be questioned given it has been estimated the sample size would need to be 620 patients for a pilot trial and 7470 patients for a gold standard trial. Nevertheless, a RCT larger in size than the trials conducted to date and included in this systematic review is required. This should include many of the same outcomes as previous RCTs and given the chronic nature of the disease, with a minimum overall follow-up of 24 months. As a minimum, end points should include measures of eGFR (primary outcome), cardiovascular events, cardiovascular risk factors and AEs (including SAEs, particularly SCARs). It could also include a composite end point (e.g. halving eGFR, ESRD requiring renal replacement therapy or renal death and a composite cardiovascular outcome). End points could also include changes in concomitant medication (e.g. antihypertensives) and, ideally, disease-specific quality-of-life measures. In order to inform analysis, it is important to collect information on the following baseline characteristics: age, sex, ethnicity, comorbidities and concomitant medication. The feasibility of collecting data on other lifestyle factors such as smoking, diet and alcohol intake (which are all cardiovascular risk factors and/or impact on levels of uric acid) should also be considered. Many of these requirements may be met by the ongoing CKD-FIX trial.

Additional supporting data are required from observational studies of patients with CKD and using allopurinol. Ideally observational data will include many of the data reported in RCTs, for example patient characteristics, allopurinol dose (and changes in dose), concomitant medication, change in eGFR, cardiovascular events, cardiovascular risk factors and AEs (including SAEs, particularly SCARs).

Contributions of authors

Nigel Fleeman Project management, data management, systematic review of clinical effectiveness data and preparation of report.

Gerlinde Pilkington Second reviewer contributing to processes of study selection, data extraction and quality assessment and provided comments on all drafts of the report.

Yenal Dundar Development of search strategies.

Kerry Dwan Statistical expertise.

Angela Boland Provided comments on the draft report.

Rumona Dickson Contributed to the development of the review protocol and provided comments on the draft report.

Hameed Anijeet Contributed to the development of the review protocol and clinical advice.

Tom Kennedy Contributed to the development of the review protocol and clinical advice.

Jason Pyatt Contributed to the development of the review protocol and clinical advice.

All contributors took part in the editing and production of this report.

Rider on responsibility for report

The views in the report are those of the authors and not necessarily those of the NIHR HTA programme. Any errors are the responsibility of the authors.

Disclaimers

This report presents independent research funded by the National Institute for Health Research (NIHR). The views and opinions expressed by authors in this publication are those of the authors and do not necessarily reflect those of the NHS, the NIHR, NETSCC, the HTA programme or the Department of Health. If there are verbatim quotations included in this publication the views and opinions expressed by the interviewees are those of the interviewees and do not necessarily reflect those of the authors, those of the NHS, the NIHR, NETSCC, the HTA programme or the Department of Health.

Ovid MEDLINE(R) In-Process and Other Non-Indexed Citations and Ovid MEDLINE(R) 1946 to 7 January 2013

EMBASE 1974 to 28 December 2012

The Cochrane Library Issue 1, 2013

ClinicalTrials.gov was searched using the term ‘allopurinol’.

As a relatively large number of studies of AEs were identified through the reference lists of studies already identified and included, an additional search of Ovid MEDLINE was conducted to identify AEs in order to ensure no studies had in fact been missed. The reason why so many AE studies had not been identified from the electronic searches already conducted was because the majority of AE studies were not specific to only patients with CKD, whereas the electronic searches conducted to date had the specific aim of only identifying patients with CKD. Therefore, the following broad search was conducted.

Ovid MEDLINE(R) In-Process and Other Non-Indexed Citations and Ovid MEDLINE(R) 1946 to 4 March 2013