Association of Non-alcoholic Fatty Liver Disease with Chronic Kidney Disease: A Systematic Review and Meta-analysis

In a systematic review and meta-analysis, Giovanni Musso and colleagues examine the association between non-alcoholic fatty liver disease and chronic kidney disease. Please see later in the article for the Editors' Summary


Introduction
Chronic kidney disease (CKD) affects 4%-13% of the Western adult population and over 25% of individuals older than 65 years [1].CKD prevalence is continuously rising in concert with the rising epidemic of its risk factors including ageing, diabetes, obesity, metabolic syndrome, smoking, and hypertension [2][3][4].In the United States, over 400,000 people currently receive some form of renal replacement therapy, and this number is expected to reach 2.2 million by 2030 [2].Beside being a risk factor for end-stage renal disease (ESRD), CKD is an important cardiovascular disease (CVD) risk factor, and most patients with CKD die from CVD before any renal replacement therapy is initiated [5].
Early recognition and treatment of CKD aimed at reducing renal disease progression and CVD complications may limit its health-related burden [4].In particular, patients with stage 3 CKD benefit the most from early referral strategies [6].Despite these premises, CKD often goes unrecognized: in the Third National Health and Nutrition Survey (NHANES III), among all individuals with stage 3 CKD, the awareness was only 8.2% [7].
The high morbidity, mortality, and health care costs associated with CKD have led investigators to seek novel modifiable risk factors.Non-alcoholic fatty liver disease (NAFLD), the hepatic manifestation of the metabolic syndrome, affects 30% of the general adult population and up to 60%-70% of diabetic and obese patients [8].NAFLD encompasses a histological spectrum ranging from simple steatosis to non-alcoholic steatohepatitis (NASH), the latter with or without advanced fibrosis.NAFLD confers an increased risk of cirrhosis, largely limited to NASH, and of CVD, independently of metabolic syndrome and traditional risk factors and through mechanisms which remain unclear [9].Growing experimental and epidemiological evidence suggests that NAFLD and CKD share common pathogenic mechanisms and interactions [10].However, evidence of a link between NAFLD and CKD is uncertain due to the small study populations and the borderline associations between NAFLD and traditional risk factors for CKD in the published literature.A meta-analysis on the association of NAFLD and CKD has not been conducted to date.We therefore analysed the evidence regarding two research questions: (1) Does NAFLD affect the risk of CKD independent of major confounders?(2) Is NAFLD severity associated with the severity of CKD?

Data Sources and Searches
We searched English and non-English language publications on MEDLINE, Ovid MEDLINE In-Process, EMBASE, ISI Web of Science, and Cochrane Library, and abstracts from the annual American Association for the Study of Liver Disease (AASLD), the American Gastroenterological Association (AGA), the European Association for the Study of the Liver (EASL), the Digestive Disease Week (DDW), and the American Society of Nephrology (ASN) Kidney Week meetings from 1980 through January 31, 2014.Search terms were: chronic kidney disease OR CKD OR kidney function OR kidney failure OR renal disease OR renal insufficiency OR renal failure OR glomerular filtration rate (GFR) OR estimated glomerular filtration rate (eGFR) OR creatinine OR albuminuria OR microalbuminuria OR macroalbuminuria OR proteinuria OR kidney injury AND NASH OR NAFLD OR non-alcoholic steatohepatitis OR nonalcoholic fatty liver disease OR fatty liver OR liver fat OR steatosis OR liver enzymes OR transaminase OR ALT OR AST OR GGT OR severity of liver disease OR fibrosis.A full list of the search strategies in different databases is reported in Text S2.

Study Selection
Inclusion criteria.Criteria were observational studies including adult (age $18 y) populations of any sex or ethnicity, with a diagnosis of NAFLD and CKD.NAFLD had to be diagnosed by (1) liver histology, (2) imaging (ultrasound, computer tomography, magnetic resonance imaging, or spectroscopy), or (3) biochemistry (elevations in serum AST, ALT, or GGT).Competing causes of steatosis, including alcohol consumption and viral hepatitis infection had to be excluded according to standard guidelines [8].The presence of CKD had to be defined by (1) persistent (.3 months) GFR,60 ml/min/1.73m 2 , as estimated using the creatinine-based Modified Diet in Renal Disease (MDRD) or Chronic Kidney Disease Epidemiology Collaboration (CKD-EPI) equations [11,12] or cystatin C-based equation [13], (2) by creatinine clearance ,60 ml/min per 1.73 m 2 using 24-hour urinary studies, (3) persistent (.3 months) kidney damage (regardless of GFR), as defined by proteinuria (microalbuminuria or macroalbuminuria using albumin-to-creatinine ratio, 24-h albumin excretion rate, or proteinuria on fresh morning urine dipstick), (4) other abnormalities due to tubular disorders or structural abnormalities detected by electrolyte or urinary sediment alterations, histology, imaging, or (5) history of kidney transplantation [2].
Exclusion criteria.Excluded from the meta-analysis were nnon-human studies, letters/case reports, studies including fewer than ten individuals, articles not reporting outcomes of interest or primary data (editorials, reviews), or using inadequate case definitions.In particular, studies were excluded that did not adequately consider competing causes of hepatic steatosis including alcohol, or viral hepatitis, or that enrolled a mixed population of cirrhotic and non-cirrhotic individuals (due to the potential confounding effects of cirrhosis per se on GFR).

Outcome Measures
Primary outcome measures.Primary outcome measures were differences in the prevalence or incidence of CKD.We compared the risk of primary outcomes between individuals with NAFLD and without NAFLD as well as across the main histological subtypes of NAFLD, since NASH and advanced fibrosis carry a significantly worse prognosis than steatosis and milder fibrosis stages, respectively [9].The impact of NAFLD and of NAFLD histological subtypes (NASH, advanced fibrosis) on eGFR, treated as a continuous variable, and on proteinuria, was also examined.
Secondary outcome measures.The severity of CKD was the secondary outcome measure.We estimated the effect of the severity of liver disease in NAFLD, as defined by NASH or advanced fibrosis, on the stage of CKD.CKD stage was categorized by GFR according to recent guidelines into CKD stage 3b (eGFR 30-44 ml/min/1.73m 2 , CKD stage 4 (eGFR 15-29 ml/min/1.73m 2 ), and CKD stage 5 (eGFR,15 ml/min/ 1.73 m 2 ) [2].

Data Extraction and Quality Assessment
Data were extracted from each study independently and in duplicate by two authors (GM, RG), using a predefined protocol (supplied in Text S2) and a data extraction sheet based on the Cochrane Handbook for Systematic Reviews of Intervention [14].The analysis was carried out in concordance with the Cochrane Handbook of Systematic Reviews and reported according to PRISMA guidelines (Text S1) [15].The initial agreement between the two reviewers for selection and validity assessment of the studies was evaluated by the Kappa coefficient.Discrepancies between the reviewers were resolved by joint discussion and mutual agreement.
Methodological quality of studies was assessed by the 22-item STROBE score [16], with two items additionally incorporated into the checklist.For imaging assessment of NAFLD, examiners had to be blinded to clinical data, and the exam had to follow prespecified, standardized criteria to detect steatosis [17].For histological assessment of NAFLD, adequate biopsy specimens with a fragment length $1.5 cm with more than six portal tracts had to be obtained and scored by a blinded pathologist according to standard criteria [8].

Data Synthesis and Analysis
For all included studies, individual participant data (IPD) was solicited from principal investigators (PIs).PIs were asked to provide the most complete and updated data, even if the follow-up was longer than that used for their respective publications.The quality of the submitted IPD was assessed using pre-specified methods (see protocol in Text S2), and any inconsistencies were clarified with the PIs.
Data not available upon database closure, either because the IPD had not been provided or because full manuscripts had not been published, were not included in our analyses.
For all analyses, we combined studies providing IPD and studies providing aggregate data (AD) into a pooled effect measure using the two-stage method [18,19]: first, the available IPD were reduced to AD in each study, then these AD (from the IPD studies) were combined with the existing AD (from the AD studies) using standard meta-analysis techniques.
In reducing IPD to AD, for dichotomous outcomes we used multivariate logistic regression in cross-sectional studies to obtain log odds ratio (OR) with its standard error (SE), and Cox proportional hazard model in longitudinal studies (all providing time-to-event data) to obtain log hazard ratio (HR) and its SE separately for each study.We then combined individual ORs (for cross-sectional studies) and HRs (for longitudinal studies) and their 95% CIs from all included studies.Associations with continuous outcome variables were expressed as weighted mean differences (WMD) with 95% CI.Only the most adjusted risk estimates that were reported in the studies were included in the analysis.All measures of dispersion were converted to standard deviations (SDs).
The study-specific risk estimates were pooled using randomeffects model, because this approach provides a more conservative assessment of the average effect size than fixed-effects model.Significance was set at p = 0.05.
The I 2 statistic and its 95% CI [20]were calculated to assess statistical heterogeneity across studies: 0% suggests no heterogeneity, 0%-25% very low heterogeneity, 25%-50% low heterogeneity, 50%-75% moderate heterogeneity, and a value of .75% high heterogeneity [14].In case of I 2 values $50%, we explored individual study characteristics and those of subgroups of the main body of evidence.
We separately analyzed cross-sectional and longitudinal studies.Furthermore, for each outcome, the results of studies defining NAFLD by histology, imaging, or liver enzyme elevation are presented separately.
Sensitivity analysis was performed by repeating the metaanalysis after one study at a time was removed to assess whether any one study significantly affected pooled estimates.Additionally, a number of subgroup analyses were planned a priori.These subgroup analyses included repeated analysis after excluding studies not fulfilling each STROBE item, and separate analyses for the following items: diabetes: we examined the effect of NAFLD on CKD in non-diabetic versus diabetic individuals, to assess if the presence of diabetes affects the association of NAFLD with CKD; studies simultaneously adjusting versus studies not adjusting for all the following risk factors for CKD: age, body mass index (BMI), metabolic syndrome (overall or each of its components), hypertension, smoking status; study design: population-based versus hospital-based; ethnicity (Asian versus non-Asian population), as defined by the investigators.As highlighted by the recent report of the Third Asian Forum of Chronic Kidney Disease Initiatives, there are striking differences in risk factors for CKD between Asian ancestry and the remaining ethnicities: as an example, chronic glomerulonephritis due to IgA nephropathy is among the three leading causes of CKD in Asians, while it is far less common in the rest of the world.Different susceptibility loci involved in innate and adaptive immunity have been identified in recent genome-wide association studies (GWAS), but also environmental factors like infections and regular consumption of herbal remedies, may underlie these epidemiological differences [21][22][23].Similarly, in Asians NAFLD is often encountered in the absence of obesity and metabolic syndrome and a different genetic background (including different Apolipoprotein C3 gene variants as an example), has been proposed to account for these ethnic differences [24].
We therefore explored whether differences in epidemiology of NAFLD and CKD between Asian and non-Asian populations affect the association of NAFLD with CKD studies including exclusively non-cirrhotic patients versus studies including exclusively cirrhotic patients; methods used to estimate GFR; outcomes related to CKD: studies assessing both eGFR and proteinuria versus studies assessing solely eGFR or proteinuria; study data availability: studies providing IPD versus studies providing exclusively AD.
When eight or more comparisons were available, the effect of continuous variables including age, whole-body and abdominal obesity (as estimated by BMI and by waist circumference, respectively) [25], insulin resistance (estimated by homeostasis model assessment of insulin resistance [HOMA-IR] index), and duration of follow-up (for longitudinal studies) on the association between NAFLD and CKD was evaluated by meta-regression analysis (random effects model, within-study variance estimated with the unrestricted maximum-likelihood method).
Small study bias was examined by constructing funnel plots and by performing the Egger's test and the trim-and-fill analysis [26].
Additionally, for the primary end-point we separately performed a one-stage meta-analysis of studies providing IPD, to examine how the association of NAFLD with CKD was altered when individual patient level covariates were accounted for [19,20].In this analysis, data from all studies providing IPD were pooled together into a single dataset and effect estimates were calculated using multivariate logistic regression (cross-sectional studies) or Cox proportional hazard models (longitudinal studies).In these models, studies were incorporated as cluster and treated as random-effect, while covariates were treated as fixed-effect.The covariates entered in the models were age, BMI, metabolic syndrome, diabetes, hypertension, smoking status, ethnicity (Asian versus non-Asian population), presence of cirrhosis, waist circumference, HOMA-index, duration of follow-up (for longitudinal studies.We first analyzed the influence of each single pre-specified covariate on the association of NAFLD with CKD with NAFLD and covariate as fixed-effect and the study as random-effects.In a second step, we did a complete case multivariable analysis with respect to NAFLD and all pre-specified covariates. We used RevMan 5.2 (Nordic Cochrane Center) and SAS 9.2 (SAS Institute) for additional analyses that could not be done with RevMan.The trim-and-fill analysis was performed with Comprehensive Meta-analysis 2.0 (Biostat).

Results
The mean (standard deviation [SD]) agreement between the two reviewers for study selection and for quality assessment were 0.89 (0.02) and 0.91 (0.04), respectively.The flow of study selection is reported in Figure 1.
All studies included non-cirrhotic participants, except one crosssectional study comparing NASH-related cirrhosis with cirrhosis of other aetiologies, matched for Child-Pugh and Model for Endstage Liver Disease-(MELD) scores [30].
In both cross-sectional and longitudinal studies, the difference between NAFLD and non-NAFLD patients remained statistically significant even when considering eGFR as a continuous variable or when considering only proteinuria as outcome (Figures S2A, S2B, S3A, and S3B within Text S3).
In both cross-sectional and longitudinal studies, meta-regression analysis found no association between CKD and age (cross- The Egger's test found no strong evidence for small study bias and the trim-and-fill analysis did not appreciably attenuate the strength of the association (Figures S4A and S4B within Text S3).
NASH and advanced fibrosis were also associated with higher ORs for proteinuria and with a lower eGFR than steatosis and non-advanced fibrosis, respectively (Figures S5A, S5B The Egger's test found no strong evidence for small study bias and the trim-and-fill analysis did not appreciably attenuate the strength of the association (Figures S4A-S4D within Text S3).
NASH and advanced fibrosis were also associated with a higher OR for incident proteinuria and with more severe eGFR reduction than steatosis and non-advanced fibrosis, respectively (Figures S7A, S7B, S8A, and S8B within Text S3).The Egger's test found no strong evidence for small study bias and the trim-and-fill analysis did not appreciably attenuate the strength of the association (Figures S4E and S4F within Text S3).
The presence of serum creatinine elevation, configuring severely decreased renal function (CKD stage 4) or renal failure (CKD stage 5), was an exclusion criterion in cross-sectional studies, which focused on the association of NAFLD with clinically unrecognized (stage 1-3) CKD.
There was no heterogeneity in the meta-analysis of overall events, suggesting a consistent disease effect.S3).Furthermore, the main results remained largely unaltered after excluding the only crosssectional study including cirrhotic individuals (Figure S19 within Text S3), while no prospective study enrolled subjects with cirrhosis at baseline.Subgroup analyses are summarized in Table 3.

Subgroup Analyses
NAFLD Histological Subtypes and the Risk of CKD NASH/advanced fibrosis and prevalent CKD.The magnitude and direction of the effect were unaltered across studies fulfilling different STROBE score items (Figures S34 and S38 within Text S3) in non-diabetic versus diabetic individuals (Figures S35 and S39 within Text S3), in studies including Asian versus non-Asian individuals (Figures S36 and S40 within Text S3), in studies using CKD-EPI versus studies using the MDRD equation (Figure S41 within Text S3), after exclusion of studies assessing only eGFR or proteinuria (Figures S37 and S42 within Text S3), and in studies providing IPD versus studies providing exclusively AD (Figure S43 within Text S3).
All studies adjusted for traditional risk factors for CKD, were hospital-based and enrolled non-cirrhotic patients.
NASH/advanced fibrosis and incident CKD.The magnitude and direction of the effect remained unaltered in non-diabetic versus diabetic individuals (Figures S44 and S47

One-Stage Individual Participant Data Meta-analysis
Twenty studies (29,282 participants, 11 cross-sectional studies, nine longitudinal studies) were included in this analysis.We first analyzed the influence of each single pre-specified individual patient level covariate on the association of NAFLD with CKD with NAFLD and covariate as fixed-effect and the study as randomeffects.In a second step, we did a complete case multivariable analysis with respect to NAFLD and all pre-specified covariates.The covariates entered in the models were age, BMI, metabolic syndrome (presence versus absence), diabetes (presence versus absence), hypertension (presence versus absence), smoking status (current smokers versus non-smokers), ethnicity (Asian versus non-Asian population), cirrhosis (presence versus absence), waist circumference, HOMA-index, duration of follow-up (for longitudinal studies).
The magnitude of the effect of NAFLD on CKD remained largely unaffected after adjusting for the covariates separately and in the fully adjusted models (Table 4).

Discussion
The main results of our analysis are the following: (1) NAFLD was associated with an increased prevalence and incidence of CKD; (2) liver disease severity in NAFLD was associated with an increased risk and severity of CKD; (3) these associations remained statistically significant in diabetic and non-diabetic individuals, as well as in studies adjusting for traditional risk factors for CKD, and were independent of whole body/abdominal obesity and insulin resistance.No study Subgroup analysis was planned a priori to assess the impact of the following items on the association between NAFLD and CKD: (1) Fulfilment of STROBE items: we planned to repeat the analysis after excluding studies not fulfilling each STROBE item (different STROBE items are described in footnote to Table 1).( 2 The prevalence of CKD is rapidly growing and in the United States over 1.1 million individuals are estimated to have ESRD by the year 2015 [60].In addition to progressing to ESRD, CKD is also a major risk factor for CVD, and most individuals with CKD die from CVD before they develop ESRD.Therefore, the search for modifiable risk factors for CKD is attracting much attention. NAFLD is an emerging risk factor for end-stage liver disease and CVD: the frequency of NASH as the primary indication for liver transplantation has increased from 1.2% to 9.7% in the last decade, becoming the third most common indication for liver transplantation in the United States [61].Furthermore, the number of combined liver and kidney transplants has been increasing exponentially in the last 5 years [62], thereby challenging cost-effective resource utilization in the treatment of end-stage organ disease.For these reasons, establishing a link between liver and kidney injury would enhance earlier identification of kidney disease and allow for the selection of treatments targeting both liver disease and CKD progression in individuals with NAFLD, with potentially relevant preventive and therapeutic implications.Our analysis disclosed an association between the presence and severity of NAFLD and the risk and severity of CKD.This association remained robust in cross-sectional and longitudinal studies, across different definitions (imaging, histology, biochemistry) of NAFLD and after taking different confounders into account.Notably, heterogeneity across cross-sectional studies evaluating NAFLD by ultrasound was abated after excluding data from analysis of the NHANES III 1988-1994 cohort, which failed to find an association between NAFLD and CKD [40].This finding may be at least partially explained by the protocol used in that study: NHANES III was not originally designed to study hepatic steatosis and the authors diagnosed NAFLD retrospectively, on the basis of archived videotapes of gallbladder ultrasound examinations.In 2009-2010, trained ultrasound readers examined the protocol used in that study and found only modest intra-and inter-reliability for the presence of hepatic steatosis, i.e., 0.77 (95% CI 0.73-0.82)and 0.70 (95% CI 0.64-0.76),respectively [63][64][65].This flaw may have further diluted

Implications for Practice
Current guidelines do not recommend screening for CKD in the absence of traditional risk factors for CKD [66].Our data suggest that individuals with NAFLD should be screened for CKD by estimation of GFR and urinalysis even in the absence of classical risk factors for CKD, particularly if NASH and/or advanced fibrosis are suspected.Early recognition of impaired kidney function in NAFLD may also allow drug dosage adjustment, thus preventing drug accumulation, especially in those being treated for obesity-associated comorbidities.From a therapeutic standpoint, there is a considerable potential for improving the current care of NAFLD patients with CKD: with respect to lifestyle interventions, smoking cessation should be more vigorously pursued, as cigarette smoking is an established risk factor for CKD, and may also aggravate NAFLD [67,68].Among pharmacological options, preliminary data from the GREACE and FANTASY randomized trials suggest statins and angiotensin receptor blockers (ARBs) may improve both liver and kidney disease in NAFLD [56,[69][70][71].Beside statins and ARBs, other agents, including pentoxifylline and v-3 polyunsaturated fatty acids, improved surrogate markers of NAFLD and CKD in distinct NALFD-associated settings like obesity, diabetes, and hypertension, and their impact on CKD in NAFLD warrants future assessment [72][73][74][75][76].

Implications for Research
Further research is required to unravel the specific cascades linking NAFLD and kidney disease.NAFLD and CKD share common risk factors and therefore both liver and kidney injury may be driven by obesity-associated mechanisms of disease, including lipotoxicity, oxidative stress, enhanced pro-inflammatory cytokine, and renin-angiotensin-aldosterone system (RAAS) axis activation [10,[77][78][79][80].However, our analysis of longitudinal studies suggests NAFLD may promote CKD independently of coexisting risk factors.Consistently, recent data suggest the steatotic and inflamed liver may be a relevant source of proinflammatory, pro-fibrogenic, and anti-fibrinolytic molecules, including fetuin-A, fibroblast growth factor (FGF)-21, tumor necrosis factor (TNF)-a, transforming growth factor (TGF)-b, and plasminogen activator inhibitor-1, all having the ability to promote kidney injury [81][82][83][84][85]. Furthermore, fatty liver may damage the kidney through VLDL lipoprotein over-secretion and induction of atherogenic dyslipidemia [86,87], as triglyceride-rich lipoproteins and oxidized LDLs promote glomerular injury and mesangial cell proliferation [77].
We also found a cross-sectional association between NAFLD and CKD, implying that even mild renal dysfunction may promote liver disease, in a mutual negative loop with detrimental cardio-metabolic consequences Consistently, uni-nephrectomized rats developed body fat redistribution from adipose depots to nonadipose tissues, profound dysregulation of hepatic fatty acid metabolism, steatohepatitis, insulin resistance, hyperglycaemia, and dyslipidaemia early after uni-nephrectomy and long before glomerulosclerosis and chronic renal failure occur [88,89].Notably, all renal and metabolic changes were prevented Lastly, since NAFLD is an emerging risk factor for liver-related and cardiovascular complications, the hypothesis that the presence of CKD may represent a simple, cost-effective tool to predict increased liver-related and CVD risk in NAFLD is intriguing and warrants assessment in large-scale prospective studies.Currently, in fact, there is no validated non-invasive marker to predict the risk of both liver disease progression and future CVD, the main causes of death in NAFLD patients [9].
Our findings may also have therapeutic research implications.Future trials will need to evaluate the impact of experimental treatments on kidney-related outcomes.Notably, only two of the available randomized controlled trials in NAFLD reports the effect of drugs on eGFR and proteinuria and none has adequate size and duration to evaluate the impact of treatments on kidney-related clinical outcomes.
Our analysis has limitations, which are intrinsic to the nature of included studies and provide the basis for future research.Studies with biopsy-proven NAFLD were by their own nature less numerous and smaller than those adopting ultrasonographic/ biochemical definitions of NAFLD, leaving the possibility of small study bias that is not detected by current tests.Furthermore, these studies were performed in tertiary centers with the possibility of selection bias.Conversely, ultrasound/liver enzyme elevations are relatively insensitive to detect NAFLD, with possible misclassification of individuals with NASH/advanced fibrosis as healthy controls and underestimation of the strength of the association between NAFLD and CKD.However, there was no heterogeneity between studies: disease effect on CKD.
Finally, despite our best efforts IPD were unavailable from as much as 39% of relevant studies, representing 54% of participant population.While meta-analysis of AD has several  limitations, including ecological bias and study-level confounding, excluding such a substantial proportion of relevant literature from our analysis would have raised the concern of data availability bias [90].Furthermore, the methodological quality of the 13 studies providing exclusive AD was generally good, 77% of them adjusted for all potential confounders, and 69% of them allowed separate risk estimates for individual patient level covariates (including diabetes, cirrhosis, ethnicity).Balancing all these reasons, we presented all relevant literature evidence by combining IPD with AD in the main analysis, and separately analyzed studies providing IPD with a one-stage method: notably, the two analyses yielded similar results, further supporting the robustness of overall findings.
In conclusion, our analysis shows that the presence and severity of NAFLD are associated with an increased risk and severity of CKD and may be a target for the prevention and treatment of CKD.Future research should evaluate strategies and interventions to prevent renal disease progression in individuals with NAFLD.
NAFLD and prevalent/incident CKD.The magnitude and direction of the associations were unaltered across studies fulfilling different STROBE score items in non-diabetic individuals (Figures S10-S14 and S23-S26 within Text S3) versus diabetic individuals (Figures S15 and S27 within Text S3), when the analysis was restricted to studies adjusting for age and BMI and metabolic syndrome and hypertension and smoking status (Figures S16 and S28 within Text S3), in population-based versus hospital-based studies (Figures S17 and S29 within Text S3), in studies including Asian versus non-Asian individuals (Figures S18 and S30 within Text S3), in studies using CKD-EPI versus within Text S3), in Asian versus non-Asian individuals (Figures S45 and S48 within Text S3) and after exclusion of studies assessing only eGFR (Figures S46 and S49 within Text S3).All studies satisfied all STROBE score items, were hospital-based, enrolled non-cirrhotic individuals, adjusted for traditional risk factors for CKD, and used CKD-EPI equation to estimate GFR.

FB
SH ESB MA LH XG SF AV YY KDL MC RH HV MML RR SR GL LLW MM HCP KY MC.Enrolled patients: GM RG JHT ME SK MH RH HH SKY PC JG FB SH ESB MA LH XG SF AV YY KDL MC RH HV MML RR SR GL LLW MM HCP KY MC.

Table 1 .
Cross-sectional studies connecting NAFLD to chronic kidney disease included in the meta-analysis.

Table 1 .
Cont. defined by birth within boundaries delineated West by the Red Sea, the Suez Canal, the Dardanelles strait, the Bosphorus the Caucasus and the Urals and East by the Bering Sea, the Japan and Indonesian archipelagos.criteria,methods of selection, and follow-up described; (g) diagnostic criteria, outcomes, exposures, predictors, potential confounders, and effect modifiers for all variables clearly defined.Specifically, regarding the definition of NAFLD: for radiological assessment: radiological exam performed by radiologists blinded to clinical data and following pre-specified, standardized criteria to detect steatosis; for histological assessment of NAFLD: adequate biopsy specimen (fragment length $1.5cmwith .6 portal tracts) and liver biopsy processed and scored by blinded pathologist according to standard criteria; (h) sources of data and details of methods of measurement given for each variable of interest; (i) any efforts to address potential sources of bias described; (j) how the study size was arrived at clearly explained; (k) how quantitative variables were handled in the analyses clearly explained; (l) all statistical methods, how missing data and loss to follow-up were addressed, any sensitivity analyses clearly described; (m) numbers of individuals at each stage of study reported; (n) characteristics of study participants, number of participants with missing data, average, and total follow-up time clearly described; (o) outcome events or summary measures over time reported; (p) unadjusted and confounder-adjusted estimates and their precision (e.g., 95% CI) reported; (q) analyses of subgroups and interactions, and sensitivity analyses reported; (r) key results with reference to study objectives summarised; (s) limitations of the study discussed; (t) cautious overall interpretation of results given; (u) generalizability (external validity) of the study results discussed; (v) source of funding and role of the funders described.ACR, albumin-to-creatinine ratio; AER, albumin excretion rate; ALT, alanine aminotransferase; AST, aspartate aminotransferase; BP, blood pressure; CKD-EPI, Chronic Kidney Disease Epidemiology Collaboration; DM, diabetes mellitus; GGT, gamma-glutamyltransferase; HDL-C, high density lipoprotein cholesterol; HTN, hypertension; LDL-C, low density lipoprotein cholesterol; MELD, model for end-stage liver disease; Met Sy, metabolic syndrome; NA, not available; OSAS, obstructive sleep apnoea; Tg, triglycerides.doi:10.1371/journal.pmed.1001680.t001 Studies with different definitions of NAFLD (histology, imaging, liver enzyme elevation) were analyzed separately and are grouped together.a Asian ethnicity was b Modified 25-item STROBE score, with the item(s) not satisfied by the study indicated in parentheses: (a) title and abstract informative and balanced; (b) background/rationale stated in the introduction; (c) objective(s) specified in the introduction; (d) study design correctly and presented early in the paper; (e) setting, locations, and relevant dates described; (f) eligibility

Table 2 .
Longitudinal studies connecting NAFLD to chronic kidney disease included in the meta-analysis.

Table 2 .
Cont. defined by birth within boundaries delineated West by the Red Sea, the Suez Canal, the Dardanelles strait, the Bosphorus the Caucasus and the Urals and East by the Bering Sea, the Japan and Indonesian archipelagos.criteria,methods of selection, and follow-up described; (g) diagnostic criteria, outcomes, exposures, predictors, potential confounders, and effect modifiers for all variables clearly defined.Specifically, regarding the definition of NAFLD: for radiological assessment: radiological exam performed by radiologists blinded to clinical data and following pre-specified, standardized criteria to detect steatosis; for histological assessment of NAFLD: adequate biopsy specimen (fragment length $1.5cmwith .6 portal tracts) and liver biopsy processed and scored by blinded pathologist according to standard criteria; (h) sources of data and details of methods of measurement given for each variable of interest; (i) any efforts to address potential sources of bias described; (j) how the study size was arrived at clearly explained; (k) how quantitative variables were handled in the analyses clearly explained; (l) all statistical methods, how missing data and loss to follow-up were addressed, any sensitivity analyses clearly described; (m) numbers of individuals at each stage of study reported; (n) characteristics of study participants, number of participants with missing data, average, and total follow-up time clearly described; (o) outcome events or summary measures over time reported; (p) unadjusted and confounder-adjusted estimates and their precision (e.g., 95% CI) reported; (q) analyses of subgroups and interactions, and sensitivity analyses reported; (r) key results with reference to study objectives summarised; (s) limitations of the study discussed; (t) cautious overall interpretation of results given; (u) generalizability (external validity) of the study results discussed; (v) source of funding and role of the funders described.ACR, albumin-to-creatinine ratio; AER, albumin excretion rate; ALT, alanine aminotransferase; AST, aspartate aminotransferase; BP, blood pressure; CKD-EPI, Chronic Kidney Disease Epidemiology Collaboration; DM, diabetes mellitus; GGT, gamma-glutamyltransferase; HDL-C, high density lipoprotein cholesterol; HTN, hypertension; LDL-C, low density lipoprotein cholesterol; MELD, model for end-stage liver disease; Met Sy, metabolic syndrome; NA, not available; OSAS, obstructive sleep apnoea; Tg, triglycerides.doi:10.1371/journal.pmed.1001680.t002 Studies with different definitions of NAFLD (histology, imaging, liver enzyme elevation) were analyzed separately and are grouped together.a Asian ethnicity was b Modified 25-item STROBE score, with the item(s) not satisfied by the study indicated in parentheses: (a) title and abstract informative and balanced; (b) background/rationale stated in the introduction; (c) objective(s) specified in the introduction; (d) study design correctly and presented early in the paper; (e) setting, locations, and relevant dates described; (f) eligibility

Table 4 .
Adjusted effect estimates for non-alcoholic fatty liver disease, non-alcoholic steato-hepatitis, advanced (stage F3) fibrosis and prevalent/incident chronic kidney disease, based on individual participant data meta-analysis from 20 studies (29,282 participants).