Nephropathy After Administration of Iso-Osmolar and Low-Osmolar Contrast Media: Evidence from a Network Meta-Analysis

International Journal of Cardiology 172 (2014) 375–380

Contents lists available at ScienceDirect

International Journal of Cardiology

journal homepage: www.elsevier.com/locate/ijcard

Nephropathy after administration of iso-osmolar and low-osmolar contrast media: Evidence from a network meta-analysis

Giuseppe Biondi-Zoccai a,⁎,MarziaLotrionteb,HenrikS.Thomsenc,EnricoRomagnolid, Fabrizio D'Ascenzo e, Arturo Giordano f,g, Giacomo Frati a a Department of Medico-Surgical Sciences and Biotechnologies, Sapienza University of Rome, Latina, Italy b Heart Failure and Cardiac Rehabilitation Unit, Columbus Integrated Complex, Rome, Italy c Copenhagen University Hospital, Herlev, Denmark d Division of Cardiology, Policlinico Casilino, Rome, Italy e Division of Cardiology, University of Turin, Turin, Italy f Unità Operativa di Interventistica Cardiovascolare, Presidio Ospedaliero Pineta Grande, Castel Volturno, Italy g Unità Operativa di Emodinamica, Casa di Salute Santa Lucia, San Giuseppe Vesuviano, Italy article info abstract

Article history: Background/objectives: Contrast-induced nephropathy (CIN) may be a severe complication to the administration Received 20 December 2013 of iodine-based contrast media for diagnostic or interventional procedure using radiation exposure. Whether Accepted 18 January 2014 there is a difference in nephrotoxic potential between the various agents is uncertain. We aimed to perform a Available online 24 January 2014 systematic review and network meta-analysis of randomized trials on iodine-based contrast agents. Methods: Randomized trials of low-osmolar or iso-osmolar contrast media were searched in CENTRAL, Google Keywords: Scholar, MEDLINE/PubMed, and Scopus. Risk of CIN was appraised within a hierarchical Bayesian model Angiography computing absolute rates (AR) and odds ratios (OR) with 95% credibility intervals, and probability of being Contrast-induced nephropathy Mixed treatment comparison best (Pbest) for each agent. Network meta-analysis Results: A total of 42 trials (10048 patients) were included focusing on 7 different iodine-based contrast media. Systematic review Risk of CIN was similarly low with iodixanol (AR = 5.7% [2.2%–13.9%], Pbest = 18.8%), iomeprol (AR = 6.0% [2.2%–15.4%], Pbest = 24.8%), iopamidol (AR = 6.1% [2.2%–15.5%], Pbest = 21.5%), and ioversol (AR = 6.0% [2.1%–16.4%], Pbest = 31.3%). Conversely, CIN was twice as common with iohexol (AR = 11.2% [4.1%–29.5%], Pbest = 0.1%) and ioxaglate (AR = 11.0% [4.0%–26.9%], Pbest b 0.1%), with both proving less safe than iodixanol (respectively OR = 2.18 [1.22–3.92] and 2.05 [1.26–3.29]), iomeprol (OR = 2.08 [1.04–4.17] and 1.96 [1.06–3.48]) and iopamidol (OR = 2.04 [1.15–3.85] and 1.92 [1.06–3.45]). Data on iopromide were less conclusive (AR = 6.9% [2.6%–17.1%], Pbest = 3.6%). Conclusions: Iodixanol, iomeprol, iopamidol and ioversol are iodine-based contrast media with a similar renal safety proﬁle. Iohexol and ioxaglate have a poorer renal safety proﬁle, whereas further data may be required on iopromide. © 2014 Elsevier Ireland Ltd. All rights reserved.

1. Introduction the risk of adverse reactions [1,2]. One of the more common untoward effects associated with contrast media administration is contrast- Administration of iodine-based contrast media is one of the corner- induced nephropathy (CIN). Accordingly, momentous efforts have stones of modern diagnostic imaging, enabling accurate visualization of been applied to develop contrast agents with as little as possible risk minute anatomic structures and pathologic conditions during interven- of renal damage [3]. tional procedures, conventional radiography and computed tomogra- This has translated into a shift from old agents with high osmolarity phy (CT) scanning in radiology. Moreover, contrast media are crucial to new ones with lower osmolarity, as indeed this chemical property for several therapeutic procedures. Contrast agents have however has been identiﬁed as a potentially important factor in renal safety been considered, at least in the past, as potentially hazardous, given [4,5]. Several studies have focused on the comparison between contrast agents, many exploiting as comparator iodixanol, which is a iso-osmolar non-ionic dimer [6,7]. However, uncertainty persists on whether there is any meaningful difference in renal safety between iodixanol and the ⁎ Corresponding author at: Department of Medico-Surgical Sciences and other non-ionic agents. Systematic reviews incorporating network Biotechnologies, Sapienza University of Rome, Latina, Corso della Repubblica 79, 04100 Latina Italy. meta-analysis (i.e. mixed treatment comparison) can maximize the E-mail address: [email protected] (G. Biondi-Zoccai). information yield stemming from complex evidence networks,

0167-5273/$ – see front matter © 2014 Elsevier Ireland Ltd. All rights reserved. http://dx.doi.org/10.1016/j.ijcard.2014.01.075 376 G. Biondi-Zoccai et al. / International Journal of Cardiology 172 (2014) 375–380

2. Methods

This review was conducted in keeping with ongoing guidelines and is registered online at metcardio.org (Protocol #6-2012). All reviewing activities were conducted inde- pendently by two skilled systematic reviewers, with divergences resolved after consensus.

2.1. Search

MEDLINE/PubMed was searched according to Biondi-Zoccai et al. [9], combining a dedicated string for randomized trials as follows: (iodixanol OR iomeprol OR ioxaglate OR iopamidol OR iopromide OR iohexol OR iopamidol OR iopental OR iopentol OR iopramide OR ioversol) AND (nephropathy OR ((kidney OR renal) AND (insufﬁciency OR damage OR injury OR failure))). CENTRAL, Google Scholar and Scopus were also searched for additional citations. No language restriction was enforced. Queries were last updated on June 30, 2013.

2.2. Selection

Citations were initially screened at the title and abstract level, and discarded if clearly not pertinent. Potentially suitable citations were retrieved as full reports. They were then Fig. 1. Review proﬁle. included if reporting a randomized trial, on low-osmolar or iso-osmolar radiographic contrast agents, and reporting data on renal safety. Studies including high-osmolar contrast agents as comparators were conversely excluded a priori to minimize their confounding providing results which are more statistically precise and externally effect on the evidence network. valid than single randomized trials, or pairwise meta-analyses [8]. Accordingly, we aimed to perform a network meta-analysis of random- 2.3. Abstraction and appraisal ized trials comparing low-osmolar or iso-osmolar contrast agents Major design, patient, and procedural features were abstracted from the shortlisted focusing on renal safety. studies. Outcomes of interest were the short-term (≤30 days) rate of CIN, deﬁned as an

Table 1 Study features.

First author Year Acronym Study type 1st Rx 2nd Rx Sample Primary endpoint Follow-up

Andersen 1993 NA Open RCT Iodixanol Ioxaglate 74 Creatinine clearance decrease N25% 2 days Aspelin 2005 NEPHRIC Double-blind RCT Iodixanol Iohexol 129 Peak increase in creatinine 3 days Barrett 2006 IMPACT Double-blind RCT Iodixanol Iopamidol 153 Creatinine increase ≥25% or ≥0.5 mg/dL 3 days Bertrand 2000 VIP Double-blind RCT Iodixanol Ioxaglate 1314 Major adverse cardiac events 2 days Bolognese 2012 CONTRAST-AMI Single-blind RCT Iodixanol Iopromide 465 Creatinine increase ≥25% 1 month Carraro 1998 NA Double-blind RCT Iodixanol Iopromide 64 Creatinine increase ≥50% Hospital discharge Chalmers 1999 NA Open RCT Iodixanol Iohexol 102 Creatinine increase ≥10% 1 week Chen 2012 DIRECT Double-blind RCT Iodixanol Iopromide 562 Creatinine increase ≥50% 4 days Chuang 2009 NA Double-blind RCT Iodixanol Iohexol 50 Peak increase in creatinine 1 week Conroy 1994 NIOPAM Double-blind RCT Iodixanol Iopamidol 180 Imaging yield Hospital discharge Davidson 2000 COURT Double-blind RCT Iodixanol Ioxaglate 815 Major adverse cardiac events 1 month Dillman 2012 NA Double-blind RCT Iohexol Iopamidol 389 Peak increase in creatinine 3 days Feldkamp 2006 NA Double-blind RCT Iodixanol Iopromide 83 Creatinine increase ≥25% 2 days Grossman 1996 NA Double-blind RCT Iodixanol Iohexol 148 Imaging yield Hospital discharge Hardiek 2008 NA Double-blind RCT Iodixanol Iopamidol 102 Creatinine increase ≥25% 1 week Jakobsen 1996 NA Double-blind RCT Iodixanol Iohexol 16 Peak increase in creatinine 4 days Jo 2006 RECOVER Double-blind RCT Iodixanol Ioxaglate 275 Creatinine increase ≥25% or ≥0.5 mg/dL 1 week Juergens 2009 NA Double-blind RCT Iodixanol Iopromide 191 Creatinine increase ≥25% or ≥0.5 mg/dL 1 week Kuhn 2008 NA Double-blind RCT Iodixanol Iopamidol 248 Adverse events 2 days Laskey 2009 NA Double-blind RCT Iodixanol Iopamidol 417 Peak increase in creatinine 1 week Lee 1996 NA Double-blind RCT Iodixanol Iohexol 126 Adverse events 3 days Manke 2003 NA Double-blind RCT Iodixanol Iomeprol 328 Pain Hospital discharge Mehran 2009 ICON Double-blind RCT Iodixanol Ioxaglate 146 Peak increase in creatinine 1 month Nguyen 2008 NA Double-blind RCT Iodixanol Iopromide 117 Peak increase in creatinine 3 months Nie 2008 NA Double-blind RCT Iodixanol Iopromide 208 Creatinine increase ≥25% or ≥0.5 mg/dL 1 month Nossen 1995 NA Open RCT Iodixanol Iohexol 16 Pharmacokinetics Hospital discharge Poirier 1996 NA Double-blind RCT Iodixanol Iohexol 49 Imaging yield Hospital discharge Rosenblum 1996 NA Double-blind RCT Iodixanol Ioxaglate 46 Adverse events Hospital discharge Rudnick 2008 VALOR Double-blind RCT Iodixanol Ioversol 334 Creatinine increase ≥0.5 mg/dL 1 week Schmid (DSA) 2007 NA Double-blind RCT Iodixanol Iomeprol 96 Adverse events 1 week Schmid (Heart) 2007 NA Double-blind RCT Iodixanol Iomeprol 120 Adverse events 1 week Seraﬁn 2011 NA Double-blind RCT Iodixanol Iopromide 92 Peak increase in creatinine 3 days Shin 2011 NA Double-blind RCT Iodixanol Iopromide 420 Creatinine increase ≥25% or ≥0.5 mg/dL Hospital discharge Siegel 1996 NA Double-blind RCT Iodixanol Ioxaglate 54 Adverse events Hospital discharge Solomon 2007 CARE Double-blind RCT Iodixanol Iopamidol 414 Creatinine increase ≥0.5 mg/dL 1 week Sundgren 1996 NA Double-blind RCT Iodixanol Iohexol 199 Imaging yield Hospital discharge Thomsen 2008 ACTIVE Double-blind RCT Iodixanol Iomeprol 148 Creatinine increase ≥0.5 mg/dL 3 days Thorstensen 1994 NA Double-blind RCT Iodixanol Iohexol 147 Imaging yield Hospital discharge Tveit 1994 NA Double-blind RCT Iodixanol Ioxaglate 102 Peak increase in creatinine Hospital discharge Verow 1995 NA Double-blind RCT Iodixanol Iopamidol 134 Discomfort Hospital discharge Wessely (Dx only) 2009 CONTRAST Double-blind RCT Iodixanol Iomeprol 651 Peak increase in creatinine Hospital discharge Wessely (PCI) 2009 CONTRAST Double-blind RCT Iodixanol Iomeprol 324 Peak increase in creatinine Hospital discharge

DSA = digital subtraction angiography; Dx = diagnostic coronary angiography; NA = not available or applicable; PCI = percutaneous coronary intervention; RCT = randomized clinical trial; Rx = therapy. G. Biondi-Zoccai et al. / International Journal of Cardiology 172 (2014) 375–380 377

with the speciﬁc study deﬁnition. Other endpoints were short-term (≤30 days) rates of dialysis and all cause death. Study validity was appraised in keeping with the Cochrane Collaboration recommendations [10].

2.4. Analysis

Descriptive analyses relied on median values or percentages. Inferential analyses were conducted within a Bayesian framework using Markov chain Monte–Carlo methods, by means of a binomial likelihood model and logit link, and computing in the same model odds ratios (OR) and absolute rates (AR), both with 95% credibility intervals (CI) [11]. The probability that each agent has the lowest rate of events (Pbest) was also computed in such models. Each analysis was based on non-informative priors for effect sizes and precision. Convergence and lack of auto-correlation were checked and confirmed after a 50,000-simulation burn-in phase, and, finally, direct probability statements were based on an additional 100,000-simulation phase. Model fit and choice between random and fixed-effect models were based on computing and comparing estimates for deviance and deviance information criterion (DIC). Given the similar fitofrandomandfixed- effect models and the inherent more parsimonious features of the latter ones, results stemming from fixed-effect models are reported throughout. Sensitivity analyses were conducted for the primary endpoint limiting the scope to intravenous contrast administra- Fig. 2. Evidence network. tion only, intra-arterial contrast administration only, and studies published after 2003. Pairwise inconsistency and inconsistency between direct and indirect effect estimates were assessed with the I2 statistic, with I2 b 25%, 25% ≤ I2 ≤ 50%, and I2 N 50% representing absolute increase ≥0.5 mg/dL or a relative increase of ≥25% in serum creatinine. In addi- respectively mild, moderate and severe inconsistency. Extent of small study effects/publi- tion, the risk of ≥25% increase of in serum creatinine was separately abstracted as second- cation bias was assessed by visual inspection of funnel plots. Computations were ary endpoint. In case such outcomes were not reported, CIN was abstracted in keeping performed with WinBUGS 1.4.3 (MRC Biostatistics Unit, Cambridge, UK).

Table 2 Patient and procedural features.

First author Year Patient/procedure type Route Age Males Diabetes mellitus CRF Contrast volume Baseline creatinine (years) (mL) (mg/dL)

Andersen 1993 Coronary angiography and ventriculography IA NA 70% NA NA 107 1.0 Aspelin 2005 Coronary or aortofemoral angiography IA 71 59% 100% 100% 163 1.6 Barrett 2006 Computed tomography of liver or lower limb vessels IV 67 69% 24% 100% 598 1.6 Bertrand 2000 PCI IA 62 77% 18% NA 220 NA Bolognese 2012 Coronary angiography and PCI in STEMI IA 66 77% 20% 4% 157 1.1 Carraro 1998 Intravenous urography IV 68 86% 33% 100% 150 1.7 Chalmers 1999 Arteriography IA 63 71% 33% 100% 56 3.2 Chen 2012 Coronary angiography and PCI IA 70 67% 31% 100% 123 NA Chuang 2009 Intravenous pyelography IV 58 68% 38% 56% 59 1.4 Conroy 1994 Intravenous urography IV NA NA NA NA NA NA Davidson 2000 PCI IA 61 68% 27% NA NA NA Dillman 2012 Computed tomography IV 56 48% 12% NA 123 0.9 Feldkamp 2006 Coronary angiography IA NA 68% 100% NA 142 1.0 Grossman 1996 Intra-cranial computed tomography IV 48 45% NA NA 101 NA Hardiek 2008 Coronary angiography or PCI IA 65 59% 100% 0% 148 0.9 Jakobsen 1996 Aortic and abdominal angiography IA NA NA NA 100% NA NA Jo 2006 Coronary angiography or PCI in renal failure IA 67 56% 35% 100% 200 1.3 Juergens 2009 Coronary angiography or PCI in renal failure IA 70 24% 41% NA 108 1.6 Kuhn 2008 Coronary angiography and ventriculography IA NA NA NA NA NA NA Laskey 2009 Coronary angiography or PCI IA 70 67% 100% 100% 120 1.6 Lee 1996 Computed tomography IV 54 55% NA NA 125 NA Manke 2003 Femoral arteriography IA 64 82% 51% 12% 141 NA Mehran 2009 Coronary angiography or PCI IA 71 88% 46% 100% 208 1.8 Nguyen 2008 Computed tomography IV 65 71% 28% 100% 1.8 Nie 2008 Coronary angiography or PCI in renal failure IA 61 68% 27% 100% 154 1.5 Nossen 1995 Aortic and abdominal angiography IA NA NA NA NA NA NA Poirier 1996 Intra-cranial arteriography IA 55 47% NA NA 89 NA Rosenblum 1996 Aortic or peripheral arteriography IA 66 52% NA NA 194 NA Rudnick 2008 Coronary angiography or PCI in renal failure IA 71 71% 52% 100% 124 2.0 Schmid (DSA) 2007 Intra-arterial digital subtraction angiography IA 66 74% NA NA NA NA Schmid (Heart) 2007 Coronary angiography and ventriculography IA 61 69% NA NA NA NA Seraﬁn 2011 Intra-cranial arteriography IA 50 27% NA NA NA 0.8 Shin 2011 Coronary angiography or PCI IA 71 53% 45% 100% 173 1.4 Siegel 1996 Aortic and abdominal angiography IA 52 NA NA NA 170 NA Solomon 2007 Cardiac or coronary angiography IA 71 64% 41% 100% 135 1.5 Sundgren 1996 Intra-venous urography IV NA NA NA NA NA NA Thomsen 2008 Computed tomography of liver IV 66 70% 20% 100% NA 1.7 Thorstensen 1994 Femoral arteriography IA NA NA NA NA NA NA Tveit 1994 Cardioangiography IA 70 60% 68% 0 179 NA Verow 1995 Aorto-femoral angiography IA NA NA NA NA 92 NA Wessely (Dx only) 2009 Coronary angiography IA 73 64% 31% 100% 145 1.3 Wessely (PCI) 2009 PCI IA 74 72% 37% 100% 366 1.4

CRF = chronic renal failure; DSA = digital subtraction angiography; Dx = diagnostic coronary angiography; IA = intra-arterial; IV = intravenous; NA = not available or applicable; PCI = percutaneous coronary intervention; STEMI = ST-elevation myocardial infarction. 378 G. Biondi-Zoccai et al. / International Journal of Cardiology 172 (2014) 375–380

3. Results sequence, allocation concealment, masking of operators and patients, exclusions, and sample size computations were reported only in 4 From an initial set of 6235 citations, a final set of 42 randomized (9.5%) studies. trials were identified, including a total of 10048 patients (Fig. 1; online Preliminary pairwise analysis for CIN suggested no severe inconsis- only supplement). Specifically, trials were conducted between 1993 and tency (I2 = 18%), nor evidence of small study effects or publication 2012, ranging in sample size from 16 to 1314, and compared iodixanol bias. Similar results were found for the risk of ≥25% increase of in versus: iohexol (10 studies, 982 patients), iomeprol (6 studies, 1667 pa- serum creatinine (I2 = 35%). Network meta-analysis showed that tients), iopamidol (7 studies, 1648 patients), iopromide (9 studies, 2202 iodixanol, iomeprol, iopamidol, and ioversol were associated with sim- patients), ioversol (1 study, 334 patients) and ioxaglate (8 studies, 2826 ilarly low absolute risks of CIN (ranging from 5.7% [2.2%–13.9%] with patients; Table 1). In addition, 1 trial compared iohexol versus iopamidol iodixanol to 6.1% [2.2%–15.5%] with iopamidol), and similarly high in 389 patients. Accordingly, the evidence network was mostly star- probability of being the best treatments (ranging from 18.8% with shaped, with all trials but one using as common comparator iodixanol iodixanol to 31.3% with ioversol; Table 4)(Fig. 3). Conversely, iohexol (Fig. 2). The primary endpoint was typically peak increase in serum cre- and ioxaglate proved credibly worse than iodixanol, iomeprol, or atinine, or increase ≥25% or ≥0.5 mg/dL. However, some trials focused iopamidol. Specifically, iohexol had a probability of being the best treat- onmajoradversecardiaceventsoradverse effects, while follow-up ment of 0.1%, with a 11.2% (4.1%–29.5%) absolute risk of CIN, whereas was typically limited to 1 week or less. Several procedures or the corresponding figures for ioxaglate were b0.1%, and 11.0% (4.0%– patients were the focus of included studies, ranging from coronary 26.9%). Data on iopromide were less precise and cannot be considered angiography to computed tomography, with intravenous injection conclusive (AR = 6.9% [2.6%–17.1%], Pbest = 3.6%). of contrast being performed in 10 (23.8%) trials (Table 2). Accord- Network meta-analysis for the risk of ≥25% increase of in serum cre- ingly, 4 (9.5%) trials were limited to diabetics and 17 (40.5%) atinine, despite stemming from fewer trials and events, largely con- only included patients with chronic (albeit typically moderate) firmed the above findings (Table 5). Specifically, iomeprol was renal failure. associated with the lowest absolute risk (4.5% [0.8%–19.3%]) and the Study quality was variable (Table 3), but on average moderate to highest probability of being the best treatment (57.4%), and ioxaglate high, as a total of 38 (90.5%) trials reportedly used double blinding. proving credibly worse than iodixanol (absolute risk = 11.4% [4.0%– However, complete and precise details on generation of randomization 28.6%], Pbest b 0.1%, OR versus iodixanol = 2.12 [1.21–3.82]).

Table 3 Validity of included studies.

First author Year Randomization details Allocation concealment Blinding Excluded patients Sample size computation

Andersen 1993 Unclear Unclear Open 2.6% Not available Aspelin 2005 Unclear Unclear Double-blind 4.4% Adequate for 80% power Barrett 2006 Computer-generated sequence Third-party Double-blind 7.8% Not available Bertrand 2000 Unclear Unclear Double-blind 9.0% Not available Bolognese 2012 Computer-generated sequence Unclear Single-blind 2.1% Adequate for 8% non-inferiority margin Carraro 1998 Unclear Unclear Unclear NA Not available Chalmers 1999 Unclear Unclear Open 17.7% Not available Chen 2012 Unclear Third-party Double-blind 5.1% Adequate for 80% power Chuang 2009 Unclear Unclear Double-blind NA Not available Conroy 1994 Unclear Unclear Double-blind NA NA Davidson 2000 Unclear Third-party Double-blind 4.8% Not available Dillman 2012 Computer-generated sequence Third-party Double-blind 5.8% Not available Feldkamp 2006 Computer-generated sequence Sealed envelopes Double-blind NA Not available Grossman 1996 Computer-generated sequence Third-party Double-blind 0.7% NA Hardiek 2008 Unclear Third-party Double-blind 16.4% NA Jakobsen 1996 Unclear Unclear Double-blind NA NA Jo 2006 Unclear Unclear Double-blind 8.3% Adequate for 80% power Juergens 2009 Unclear Unclear Double-blind 5.2% Adequate for 80% power Kuhn 2008 Unclear Unclear Double-blind NA NA Laskey 2009 Unclear Unclear Double-blind 2.6% Adequate for 90% power Lee 1996 Unclear Unclear Double-blind NA NA Manke 2003 Unclear Unclear Double-blind NA Adequate for 90% power Mehran 2009 Computer-generated sequence Sealed envelopes Double-blind 0.0% Adequate for 80% power Nguyen 2008 Unclear Unclear Double-blind 7.1% Adequate for 95% power Nie 2008 Computer-generated sequence Unclear Double-blind 3.7% Adequate for 80% power Nossen 1995 Unclear Unclear Open NA NA Poirier 1996 Unclear Unclear Double-blind 2.0% NA Rosenblum 1996 Computer-generated sequence Third-party Double-blind NA NA Rudnick 2008 Computer-generated sequence Unclear Double-blind 11.3% Adequate for 90% power Schmid (DSA) 2007 Computer-generated sequence Third-party Double-blind NA NA Schmid (heart) 2007 Computer-generated sequence Third-party Double-blind NA NA Seraﬁn 2011 Computer-generated sequence Unclear Double-blind 16.7% Adequate for 90% power Shin 2011 Unclear Unclear Double-blind 2.1% Adequate for 80% power Siegel 1996 Computer-generated sequence Unclear Double-blind NA NA Solomon 2007 Computer-generated sequence Third-party Double-blind 3.3% Adequate for 80% power Sundgren 1996 Unclear Unclear Double-blind 3.0% NA Thomsen 2008 Unclear Third-party Double-blind 19.6% NA Thorstensen 1994 Unclear Unclear Double-blind NA NA Tveit 1994 Unclear Unclear Double-blind NA NA Verow 1995 Unclear Third-party Double-blind 4.3% NA Wessely (Dx only) 2009 Computer-generated sequence Sealed envelopes Double-blind 0.0% Adequate for 90% power Wessely (PCI) 2009 Computer-generated sequence Sealed envelopes Double-blind 0.0% Adequate for 90% power

DSA = digital subtraction angiography; Dx = diagnostic coronary angiography; NA = not available or applicable; PCI = percutaneous coronary intervention. G. Biondi-Zoccai et al. / International Journal of Cardiology 172 (2014) 375–380 379

Table 4 Risk of contrast-induced nephropathy. Bold type highlights odds ratios which are credibly far from equipoise.

Iodixanol Iohexol Iomeprol Iopamidol Iopromide Ioversol Ioxaglate

Iodixanol AR = 5.7% OR = 0.46 OR = 0.95 OR = 0.94 OR = 0.81 OR = 0.94 OR = 0.49 (2.2%–13.9%) (0.26–0.82) (0.66–1.37) (0.66–1.35) (0.60–1.10) (0.56–1.61) (0.30–0.79) Pbest = 18.8% Iohexol OR = 2.18 AR = 11.2% OR = 2.08 OR = 2.04 OR = 1.79 OR = 2.08 OR = 1.06 (1.22–3.92) (4.1%–29.5%) (1.04–4.17) (1.15–3.85) (0.96–3.45) (0.92–4.76) (0.50–2.33) Pbest = 0.1% Iomeprol OR = 1.05 OR = 0.48 AR = 6.0% OR = 0.99 OR = 0.85 OR = 0.99 OR = 0.51 (0.73–1.51) (0.24–0.96) (2.2%–15.4%) (0.59–1.67) (0.54–1.39) (0.52–1.89) (0.29–0.94) Pbest = 24.8% Iopamidol OR = 1.06 OR = 0.49 OR = 1.01 AR = 6.1% OR = 0.87 OR = 1.01 OR = 0.52 (0.74–1.52) (0.26–0.87) (0.60 –1.70) (2.2%–15.5%) (0.55–1.37) (0.52–1.92) (0.29–0.94) Pbest = 21.5% Iopromide OR = 1.23 OR = 0.56 OR = 1.17 OR = 1.15 AR = 6.9% OR = 1.16 OR = 0.60 (0.91–1.66) (0.29–1.04) (0.72–1.85) (0.73–1.83) (2.6%–17.1%) (0.63–2.17) (0.34–1.02) Pbest = 3.6% Ioversol OR = 1.06 OR = 0.48 OR = 1.01 (0.53–1.91) OR = 0.99 (0.52–1.91) OR = 0.86 (0.46–1.60) AR = 6.0% OR = 0.52 (0.62–1.79) (0.21–1.09) (2.1%–16.4%) (0.25–1.04) Pbest = 31.3% Ioxaglate OR = 2.05 OR = 0.94 OR = 1.96 OR = 1.92 OR = 1.67 OR = 1.92 AR = 11.0% (1.26–3.29) (0.43–1.99) (1.06–3.48) (1.06–3.45) (0.98–2.92) (0.96–4.07) (4.0%–26.9%) Pbest b 0.1%

AR = absolute rate (95% credibility interval); OR = odds ratio (row-item vs column-item; 95% credibility interval); Pbest = probability of being best treatment.

Sensitivity analyses for the primary endpoint largely confirmed the in urine output. Debate on which surrogate endpoint is the best remains overall results in direction and magnitude of effects when limiting the open, even if important advances have been made thanks to several col- scope of the review to studies published only after 2003 as well as laborative efforts [12,15]. those limited to trials on intra-arterial contrast administration (online A plethora of trials have compared modern low-osmolar or iso- only supplement). Conversely, the sensitivity analysis focusing on intra- osmolar contrast agents, aiming to identify which, if any, medium is venous administration only was limited by the fewer included studies safer than the others [7]. A common assumption in many trials has and patients (respectively 11 and 1590), without any such trial been that, in keeping with the NEPHRIC study [6], iodixanol is a safer reporting on ioversol or ioxaglate. Analyses for the risk of dialysis or agent, at least in those at higher risk of CIN, such as those with chronic death could not be performed due to the very rare occurrence of these renal failure due to diabetes mellitus. Our findings suggest however events (respectively 14/6728 [0.2%] and 84/5880 [1.4%]), with as that there is no difference between non-ionic low-osmolar agents and many as 18 (42.9%) studies reporting no patient requiring dialysis or the non-ionic dimer iodixanol, being associated with similar absolute dying during follow-up. risks of CIN or ≥25% increase of in serum creatinine, and also having comparative odds ratios. Conversely, some of the low-osmolar media have indeed an increased nephrotoxic potential compared to iodixanol. 4. Discussion In particular, iohexol and ioxaglate are associated with much higher rates of CIN, and might be avoided in high-risk patients. Given the un- This review is the first to exploit modern statistical methods to sum- certain estimates for iopromide, and the Pbest values which cannot con- marize and compare the renal safety of current iodine-based contrast clude in favor nor against this agent, no clear recommendation can be media, and has the following key implications: a) iodixanol has envisioned to date for this agent. established itself as the reference contrast agent as testified by its use as comparator in all but one of the included trials; b) despite this, low- osmolarity contrast agents appear, with the notable exception of iohexol and ioxaglate, as safe as iodixanol; c) accordingly, iodixanol, iomeprol, iopamidol, and ioversol have the lowest nephrotoxic potential but none of the agents completely avoid CIN. Modern medicine has diagnostic and interventional imaging as one of its cornerstone, and radiology with its ancillary or related subspecialty will play an ever increasing role in clinical practice. Contrast agents have always been a crucial component of imaging techniques, as they increase resolution and enable the identification of minute anatomic structures. However, their administration is not devoid of complications [12]. Notwithstanding the major improvements in contrast agents and the development of ever safer media, adverse events still occur in patients who have received them. The most feared complication in today's clinical practice is CIN, which may range from transient, benign, and asymptomatic decrease in glomerular filtration rate to irreversible renal failure eventually requiring dialysis [13]. Whereas the latter out- come is much more important and meaningful, its rare occurrence means that it cannot be realistically exploited in clinical research en- deavors aiming at comparing different contrast agents [14]. According- ly, all recent research focusing on renal safety of contrast media relies on surrogate renal safety endpoints, ranging from absolute increase in Fig. 3. Rankogram for the risk of contrast-induced nephropathy (the size of the circle serum creatinine, to relative increase in serum creatinine, to changes corresponds to the probability of being the best treatment for each competing therapy). 380 G. Biondi-Zoccai et al. / International Journal of Cardiology 172 (2014) 375–380

Table 5 Risk of ≥25% increase of in serum creatinine. Bold type highlights odds ratios which are credibly far from equipoise.

Iodixanol Iohexol Iomeprol Iopamidol Iopromide Ioxaglate

Iodixanol AR = 5.7% OR = 0.75 OR = 1.30 OR = 0.98 OR = 0.85 OR = 0.47 (2.2%–13.9%) (0.33–1.70) (0.34–5.88) (0.59–1.59) (0.62–1.16) (0.26–0.83) Pbest = 13.1% Iohexol OR = 1.33 AR = 7.5% OR = 1.72 OR = 1.30 OR = 1.14 OR = 0.63 (0.59–3.00) (2.2%–22.4%) (0.38–11.11) (0.63–2.70) (0.48–2.70) (0.24–1.72) Pbest = 7.9% Iomeprol OR = 0.77 OR = 0.58 AR = 4.5% OR = 0.76 OR = 0.65 OR = 0.36 (0.17–2.92) (0.09–2.66) (0.8%–19.3%) (0.16–3.13) (0.14–2.63) (0.08–1.47) Pbest = 57.4% Iopamidol OR = 1.02 OR = 0.77 OR = 1.32 AR = 5.9% OR = 0.87 OR = 0.48 (0.63–1.70) (0.37–1.58) (0.32–6.42) (2.0%–15.8%) (0.49–1.56) (0.24–1.04) Pbest = 17.3% Iopromide OR = 1.17 OR = 0.88 OR = 1.54 OR = 1.15 AR = 6.7% OR = 0.56 (0.86–1.62) (0.37–2.10) (0.38–7.32) (0.64–2.04) (2.5%–16.6%) (0.29–1.04) Pbest = 4.3% Ioxaglate OR = 2.12 OR = 1.60 OR = 2.75 OR = 2.10 OR = 1.80 AR = 11.4% (1.21–3.82) (0.58–4.13) (0.68–13.07) (0.96–4.22) (0.96–3.49) (4.0%–28.6%) Pbest b 0.1%

AR = absolute rate (95% credibility interval); OR = odds ratio (row-item vs column-item; 95% credibility interval); Pbest = probability of being best treatment.

Despite the inherent strengths of our work, which include the com- [2] Kwok CS, Pang CL, Yeong JK, Loke YK. Measures used to treat contrast-induced nephropathy: overview of reviews. Br J Radiol 2013;86:20120272. prehensiveness and the robust statistical methods, this review has also [3] Nunag M, Brogan M, Garrick R. Mitigating contrast-induced acute kidney injury as- limitations. First, being a study-level meta-analysis, within-study inter- sociated with cardiac catheterization. Cardiol Rev 2009;17:263–9. actions or key moderators cannot be appraised in detail. Moreover, any [4] Fedan JS, Dowdy JA, Johnston RA, Van Scott MR. Hyperosmolar solution effects in guinea pig airways. I. Mechanical responses to relative changes in osmolarity. J study-level systematic review is only as good as the included trials, and Pharmacol Exp Ther 2004;308:10–8. thus the fair internal validity of most included studies translate into an [5] Sendeski MM. Pathophysiology of renal tissue damage by iodinated contrast media. acceptable but not absolute internal validity of the review. The likeli- Clin Exp Pharmacol Physiol 2011;38:292–9. hood of experiencing other adverse effects may also play an important [6] Aspelin P, Aubry P, Fransson SG, et al. Nephrotoxic effects in high-risk patients un- dergoing angiography. N Engl J Med 2003;348:491–9. role in choosing between one agent and the others. Finally, non- [7] From AM, Al Badarin FJ, McDonald FS, Bartholmai BJ, Cha SS, Rihal CS. Iodixanol iodinated contrast agents (e.g. carbon dioxide or paramagnetic media) versus low-osmolar contrast media for prevention of contrast induced nephrop- can also be used in selected cases and for selected indications when athy: meta-analysis of randomized, controlled trials. Circ Cardiovasc Interv 2010;3:351–8. contrast-enhanced imaging is envisioned. Whenever this is possible [8] Biondi-Zoccai G, Landoni G, Modena MG. A journey into clinical evidence: from case and imaging yield can be considered acceptable, they may be an appeal- reports to mixed treatment comparisons. HSR Proc Intensive Care Cardiovasc Anesth ingly safer choice with regard to the kidney. 2011;3:93–6. [9] Biondi-Zoccai GG, Agostoni P, Abbate A, Testa L, Burzotta F. A simple hint to improve In conclusion, iodixanol, iomeprol, iopamidol and ioversol are con- Robinson and Dickersin's highly sensitive PubMed search strategy for controlled trast media with a similar renal safety profile and can thus be used rou- clinical trials. Int J Epidemiol 2005;34:224–5. tinely for diagnostic and interventional imaging. Iohexol and ioxaglate [10] Higgins JPT, Green S, editors. Cochrane handbook for systematic reviews of interven- tions version 5.1.0 [updated March 2011]. The Cochrane Collaboration; 2011 [Avail- are contraindicated instead, whereas further data are required on the able from: www.cochrane-handbook.org (last accessed on July 31, 2013)]. risk-benefitprofile of iopromide. [11] Biondi-Zoccai G, Lotrionte M, Landoni G, Modena MG. The rough guide to systematic reviews and meta-analyses. HSR Proc Intensive Care Cardiovasc Anesth 2011;3:161–73. Appendix A. Supplementary data [12] Stacul F, van der Molen AJ, Reimer P, et al. Contrast induced nephropathy: updated ESUR Contrast Media Safety Committee guidelines. Eur Radiol – Supplementary data to this article can be found online at http://dx. 2011;21:2527 41. [13] Weisbord SD, Palevsky PM. Contrast-induced acute kidney injury: short- and long- doi.org/10.1016/j.ijcard.2014.01.075. term implications. Semin Nephrol 2011;31:300–9. [14] Assareh A, Yazdankhah S, Ahmadzadeh A, Yadollahzadeh M, Nasehi N, Haybar H. Defining the at risk patients for contrast induced nephropathy after coronary angi- References ography; 24-h urine creatinine versus Cockcroft–Gault equation or serum creatinine level. J Res Med Sci 2012;17:859–64. [1] Seeliger E, Sendeski M, Rihal CS, Persson PB. Contrast-induced kidney injury: mech- [15] Budano C, Levis M, D'Amico M, et al. Impact of contrast-induced acute kidney injury anisms, risk factors, and prevention. Eur Heart J 2012;33:2007–15. definition on clinical outcomes. Am Heart J 2011;161:963–71.