Answers to the reviewer Date : 16 April 2014 MS: 1019561888118693 Authors: Ling Li et al. Title: " Is hyperuricaemia an independent risk factor for new-onset chronic kidney disease? A systematic review and meta-analysis based on observational cohort studies" Dear Editor, Thank you very much for considering our manuscript for publication. Your suggestions were very helpful to us, and we have incorporated those points into our revised manuscript. Reviewer 1: Stephen Juraschek’s comments Comment 1. Methods: What is the “related articles” function. That seems to be a specific reference to one database. Answer: “Related articles” function was frequently used in several database as OVID (Medline) and Pubmed. The search engine will guarantee users to find relevant article according to the key words of article you had found. In OVID (Medline), “related item” was used to broaden the search for the saurus. Comment 2. Methods: CKD (page 4) was defined as eGFR<60. This is not correct and more accurately should be specified as CKD stage III. Stages 1 and II are related to albuminuria in the setting of an egfr > 60. Answer: Thank you for the suggestion. We added 2 studies on stage 1-2 CKD patients in the revised manuscript. Comment 3. The paragraph on quality assessment is not clear, what is meant by domains? What are the domains? Answer: ‘Domains’ here means items in quality assessment scale. Some authors used this word in their quality assessment in meta-analysis. For example “George Thomas, Ashwini R. Sehgal, et al. Metabolic Syndrome and Kidney Disease: A Systematic Review and Meta-analysis. Clin J Am Soc Nephrol 2011; 6,1-10”. Comment 4. Discussion – Limitations paragraph: You should mention that hyperuricemia could reflect subclinical kidney disease. You should also dedicate more text to the alternative side of the debate that CKD is causal for hyperuricemia. Answer: Thank for your advices. We have add relevant content in the revised manuscript. New Manscript: Firstly, though we found that hyperuricemia could be a risk factor for CKD. For patients with subclinical kidney disease, hyperuricemia is the subsequence of decreased renal uric acid excretion, which can further in turn exacerbate kidney function. Therefore the causal relationship between hyperuricem and CKD is far more complicated than a simple cause and effect relation, which means we have to be cautious in interpreting the conclusion of the current study. Comment 5. Table 1: The adjustments are not readily interpreted with the number system. It would be better to list them out. Also, I think it is problematic to include cohorts where the adjusted covariates are not known/no reported. This is problematic as meta-analysis relies on the input studies for its conclusions. Answer: We have listed all adjustments out in the Table 1. in the revised manuscript. Mostly, for multivariate analysis, adjustment were automatically conducted by statistic software. The calculation of HR or OR has been considered for the influence of all relevant variables. In addition, for “Enter Method” for logistic and COX regression analysis, the results of adjustment would not be listed out in the results of calculation. Comment 6. Table 2: There is a lot of heterogeneity that is not explained by these subgroups. Cochrane recommends against pooling studies with excess heterogeneity. Answer: We found major heterogeneity was caused by different population and the statistic method for calculating OR(using continuous variable or dichotomous variable). In our modified manuscript, by introducing subgroup analysis on population and the way for calculating OR, the majority of heterogeneities were eliminated.(Figure 2-4, Table 3) Comment 7. Figure 2: I don’t think studies that do not specify CKD status at baseline should be included at all. Answer: We are sorry for not emphasizing that in our methods part, that only studies of non-CKD patients at baseline were include. Comment 8. Figure 3. This figure is not very high quality. Also there is a problem with the confidence interval of the Sonoda 2011 study…it seems cut off. Also, box sizes should be proportional to the study weight, not equal in size. Answer: By using Revman 5.1 instead of Stata 12.0 for graph drawing, the quality of figure has been improved and the box sizes were adjusted as well. Comment 9. Figure 6. Begg’s funnel plot is not particularly helpful with so few studies. Answer: Thanks for your suggestion When two new studies were included, study number is larger than 6, which allows for it to be used for vivid sight of symmetry of publication. Reviewer 2: Steven G Coca’s comments Comment 1. In what analyses did they use weighted mean differences? I don’t see in the text, figures or tables. Answer: Weighted mean difference was supposed to be used for description of the level of uric acid. However, as few article reported that, so the results didn’t present any weighted mean differences. We have corrected the error and deleted this sentence. Comment 2. They need to mention the publication in BMJ 2013;347:f4262 (Association of plasma uric acid with ischaemic heart disease and blood pressure: mendelian randomisation analysis of two large cohorts) in their discussion, a study that used genes that cause hyperuricemia as an instrumental variable and demonstrated most of the association between UA and heart disease is due to confounding. Answer: Thanks for your suggestion, we added it in our discussion. New Manuscript: For example, in a recent mendelian randomisation analysis of two large cohorts, though estimates confirmed known observational associations between plasma uric acid and hyperuricaemia with risk of ischaemic heart disease, however, when using genotypic instruments for uric acid and hyperuricaemia, the author saw no evidence for causal associations between uric acid, ischaemic heart disease.[29] Similarly, the association found by our meta-analysis does not necessarily lead to a causal relationship between SUA and CKD. Comment 3. Were they able to get lower heterogeneity for SUA and new-onset CKD when they accounted for factors that they identified through meta-regression that most-strongly influenced the heterogeneity? For example, for hyperuricemia, they reduced the heterogeneity to 0% by eliminating Kawashima. Answer: We have made further improvement on meta-regression analysis on reducing heterogeneity. Comment 4. Figures are labeled wrong. They have figures 4, 5 and 6 labeled as funnel plots. See Figure 2, 3 and Table 3 in detail. Answer: Sorry for our carelessness, we have corrected it. Reviewer 3: Seoyoung Kim’s comments Comment 1. Introduction: Page 3. 3rd paragraph, “A previous meta-analysis… “: The reference number is missing. Answer: Sorry, the missing reference number has be added [22]. Comment 2. Methods: Search strategy should be presented better. I performed a quick Pubmed search using the keywords that the authors listed in page 4 and found more than 1600 papers just in Pubmed. It is unclear how the authors got only 1208 relevant articles from all the data sources listed in the paper. Answer: We have provided a search strategy of each database and study selection in detail in our modified flow diagram. ( Figure 1. ) We sincerely appreciate the reviewer’s comment. New Manuscript: A literature search was performed using electronic databases Medline Ovid/Medline (1948 to present), Pubmed/Medline, Embase, and ISI Web/Web of Science for published studies from January 1970 to September 2013. Key words used were: (“hyperuricemia”, , “uric acid elevated”, “hyperuricaemia”, “uricacidemia”, and “uricacidaemia”) and “CKD”or “chronic kidney disease”or “Chronic renal dysfunction” or Chronic renal insufficiency) or. The “related items” function was used to broaden the search. The electronic search was up to July 2013 with no limitations regarding the type of publication. Comment 3. Methods: It is unclear whether the authors selected cohort studies that only included patients with no CKD at baseline. If not, how did the authors make sure that the outcomes in the individual studies were incident CKD? Answer: We did selected cohort studies that only included patients with no CKD at baseline. Thank you for the comment. Comment 4. Methods: If the authors included only cohort studies, why some of the cohort studies used ORs ? did those studies have the complete follow-up time? Answer: As patients in most observational cohort studies only received two or three blood sample test, and a number of individuals have asymptomatic chronic kidney disease. Thus the exact time for developing incident CKD in some studies would be unknown or unreliable. Thus, researchers in some studies use logistic regression analysis(OR) instead of COX regression(HR). Comment 5. Methods: It is unclear how the authors pooled the data when the source papers reported RR/OR per serum uric acid level. Was it per 1mg/dl? Or was it per 1 SD ? Answer: We analyze the summary OR calculated by continuous variable(uric acid increase per 1mg/dl) and dichotomous variable(hyperuricemia or not) respectively. As original studies has two types, some reported the increasing risk of new-onset CKD per 1mg/dl, some reported increasing risk of new-onset CKD in people with hyperuricemia. Comment 6. Methods: For the studies using quartiles or other types of categories of uric acid, how different the cut-offs were in each study? Answer: Studies using quartiles for categories of uric acid, the reference group were the first quartile, while most studies used population with normal uric acid level as reference for calculation of OR or HR. However OR calculated using quartiles were excluded from pooling as it exaggerated the OR value. Comment 7. Methods: For quality of studies, was ‘% followup or % lost to followup’ considered? It is particularly important in the studies with long-term duration. Answer: The percentage of population which were dropout were assessed according to modified New-Ottawa Scale, unclear rate of dropout or dropout rate higher than 10% were thought to be of low quality. We have showed them in Table 2.. Comment 8. Methods: Meta-regression: was it univariate only? Answer: Univariate was first used for meta-regression analysis, while all single variables could not explained all heterogeneity, multivariate meta-regression were used. In our modified manuscript, univariate meta-regression could fully explain the heterogeneity, so we didn’t further perform multivariate meta-regression. We thank the reviewer for the comment. Comment 9. Results: In all the results, the citations # should be added. Answer: Thanks for your suggestion. We have added citation number into all results accordingly. Comment 10. Results: In 8 studies that reported CKD risk in uric acid level (continuous), please be specific what the unit of uric acid was used. Was it 1mg/dl? Or was it per 1SD deviation? Answer: All were per 1mg/dl. We gave a more clear description towards that. Comment 11. Results: Page 7, in the 2nd paragraph, “The main limitation in the majority of studies was the absence of adequate follow up.” Please specify what ‘adequate follow up’ means. Answer: “adequate follow up” was one terms in NOS scale. When specified to our study, absence of adequate follow-up here means the time of follow-up was enough for observe the development of CKD. We define a follow-up of longer than 5 years as adequate follow-up. Comment 12. Results: Page 8, 2nd paragraph: was the meta-regression in 6 studies adjusted for a single variable at a time? How was the quality of studies used in the meta-regression? As a continuous score? Answer: Sorry for not mentioning that, the scores >5 points were thought to be of high quality. We have showed them out in Table 2.. Comment 13. Results: Table 1: does this table include only those studies analyzing the data by HU vs non-HU? If that is the case, it should have a separate table for the studies reporting CKD risk per uric acid. Answer: We used 1 table(Table 1) to show all results reported in each article. Five studies reported both risk estimates on HU vs non-HU and risk per mg/dl uric acid. All studies included were listed in the table. We used “&” to note AOR calculated using elevated uric acid level per 1mg/dl for incidence of onset of CKD and“ #” to note AOR calculated using hyperuricaemia individuals compared with normal individuals. i. The list of source studies is not ordered alphabetically or chronologically. Answer: We have updated the order in a chronological way. Thanks for you suggestion. ii. Should include % completion of follow up Answer: Most studies didn’t reported exact proportion of completion follow-up, in their study, only participants complete the follow-up were included for analysis. iii. For adjustments, please list the variable names instead of #. Answer: We have listed them out in the Table 1. in the revised manuscript. iv. Quality of the studies should be also shown individually. Answer: We have showed the quality assessment individually in Table 2. v. Ref # 31 and 32 were not adjusted for any? If that’s the case, why did the authors include the crude estimates? Answer: The studies were indeed adjusted, but some authors just did not mention it in the text. As for multivariate analysis, adjustments were often automatically conducted by statistic software. We have retrieved the variable used for adjustment in tables from original articles and filled it on our Table 1. Comment 14. Results: Figure 5, legend is incorrect. Figure legends do not seem detailed enough. Answer: Sorry for our carelessness, we have corrected it in modified manuscript. Comment 15. Discussion: Should mention some of the meta-analyses/studies on urate-lowering tx and CKD. Not all studies showed the effectiveness of allopurinol on reducing the risk of CKD. Answer: We have added some relative content in the discussion. We appreciate your recommendation. New Manuscript: A case-control study conducted by Kanbay et al. revealed that allopurinol intervention in individuals with eGFR >60 mL/min/1.73 m2 significantly increased eGFR at follow-up[26]. One recent meta-analysis made by Bose et al. demonstrated uric acid-lowering therapy with allopurinol could retard the progression of CKD, but had no effect on proteinuria and blood pressure[27]. We thank you again for your insightful comments on our paper. Sincerely yours, Ling Li, Chen Yang, Yuliang Zhao, Xiaoxi Zeng, Fang Liu, Ping Fu