A New Preprocedure Risk Score for Predicting Contrast-Induced Acute Kidney Injury

Accepted Manuscript A New Pre-procedure Risk Score for Predicting Contrast-Induced Acute Kidney Injury Chongyang Duan, MD, Yingshu Cao, MS, Yong Liu, MD, Lizhi Zhou, MD, Kaike Ping, MS, Ming T. Tan, PhD, Ning Tan, MD, FACC, FESC, Jiyan Chen, MD, FACC, FESC, Pingyan Chen, MS PII:

S0828-282X(17)30030-2

DOI:

10.1016/j.cjca.2017.01.015

Reference:

CJCA 2358

To appear in:

Canadian Journal of Cardiology

Received Date: 28 October 2016 Revised Date:

23 January 2017

Accepted Date: 23 January 2017

Please cite this article as: Duan C, Cao Y, Liu Y, Zhou L, Ping K, Tan MT, Tan N, Chen J, Chen P, A New Pre-procedure Risk Score for Predicting Contrast-Induced Acute Kidney Injury, Canadian Journal of Cardiology (2017), doi: 10.1016/j.cjca.2017.01.015. This is a PDF file of an unedited manuscript that has been accepted for publication. As a service to our customers we are providing this early version of the manuscript. The manuscript will undergo copyediting, typesetting, and review of the resulting proof before it is published in its final form. Please note that during the production process errors may be discovered which could affect the content, and all legal disclaimers that apply to the journal pertain.

ACCEPTED MANUSCRIPT A New Pre-procedure Risk Score for Predicting Contrast-Induced Acute Kidney Injury Short title: Pre-procedure Risk Score for CI-AKI

Duan

Chongyang Duan, MD1, Yingshu Cao, MS1, Yong Liu, MD2, Lizhi Zhou, MD1, Kaike Ping,

RI PT

MS1, Ming T. Tan, PhD1,3, Ning Tan, MD, FACC, FESC2, Jiyan Chen, MD, FACC, FESC2*, Pingyan Chen, MS1*

State Key Laboratory of Organ Failure Research, National Clinical Research Center for

SC

1

Kidney Disease, Guangzhou, China; Department of Biostatistics, School of Public Health,

2

M AN U

Southern Medical University, Guangzhou 510515, China

Department of Cardiology, Guangdong Cardiovascular Institute, Guangdong General

Hospital, Guangdong Academy of Medical Sciences, Guangzhou 510100, China 3

Department of Biostatistics, Bioinformatics & Biomathematics, Georgetown University,

*

TE D

4000 Reservoir Rd NW, Washington DC, USA (Tan)

Address for correspondence:

EP

Prof. Pingyan Chen, State Key Laboratory of Organ Failure Research, National Clinical

AC C

Research Center for Kidney Disease, Department of Biostatistics, School of Public Health, Southern Medical University, 1838 North Guangzhou Avenue, Guangzhou, 510515, P.R. China. E-mail: [email protected]; Tel/Fax: +8620-61648320. Dr. Jiyan Chen, MD, FACC, FESC, Department of Cardiology, Guangdong Cardiovascular Institute, Guangdong General Hospital, Guangdong Academy of Medical Sciences, Guangzhou 510100, China. E-mail: [email protected]; Tel: +8620-83827812-10528; Fax: +8620-83851483.

1

ACCEPTED MANUSCRIPT Word count: 5614

Brief Summary: We developed a simple and accurate tool for early predicting contrast-induced acute kidney injury in patients after coronary angiography or percutaneous

RI PT

coronary intervention at pre-procedure. It could help clinicians to assess the risk of contrast-induced acute kidney injury before contrast exposure, plan and initiate the most

AC C

EP

TE D

M AN U

SC

appropriate disease management in time.

2

ACCEPTED MANUSCRIPT Abstract Background: Most of the risk models for predicting contrast-induced acute kidney injury (CI-AKI) are available only for post contrast exposure prediction, while the prediction at pre-procedure is more valuable in practice. The study aims to develop a risk scoring system

angiography or percutaneous coronary intervention (PCI).

RI PT

based on pre-procedure characteristics for early predicting CI-AKI in patients after coronary

Methods: We prospectively recruited 1777 consecutive patients who were randomized in an

SC

approximate 3:2 ratio to create a development dataset (n=1076) and a validation dataset (n=701). A risk score model based on pre-procedure risk factors were developed using the

M AN U

stepwise logistic regression. Validation was performed by bootstrap and split-sample methods. Results: The occurrence of CI-AKI were 5.97% (106/1777), 5.95% (64/1076) and 5.99% (42/701) in overall, development and validation dataset, respectively. The risk score developed by five prognostic factors (age, serum creatinine, NT-proBNP, high sensitive

TE D

C-reactive protein and primary PCI), ranging from 0 to 36, was well-calibrated (Hosmer-Lemeshow χ²=4.162, P=0.842). Good discrimination was obtained both in development and validation datasets (C-statistics 0.809 and 0.798, respectively). The risk

EP

score was highly and positively associated with CI-AKI (P for trend <0.001), in-hospital and

AC C

long-term outcomes.

Conclusions: The novel risk score model we developed is a simple and accurate tool for early/pre-procedure prediction of CI-AKI in patients undergoing coronary angiography or PCI. This tool allows assessment of the risk of CI-AKI before contrast exposure, allowing for timely initiation of appropriate preventive measures. Keywords: contrast-induced acute kidney injury; risk score; coronary angiography; NT-proBNP; C-reactive protein

3

ACCEPTED MANUSCRIPT INTRODUCTION Contrast-induced acute kidney injury (CI-AKI) is one of the common complications of coronary diagnostic and interventional procedures, which is significantly associated with prolonged hospitalization, health cost and increased mortality1, 2. Identifying patients at risk

RI PT

of CI-AKI or contrast-induced nephropathy (CIN) easily and accurately would allow the administration of prophylactic interventions to those at high risk3, 4. International Society of Nephrology (ISN)’s 0by25 (zero preventable deaths by 2025) initiative aims to eliminate

SC

preventable deaths from acute kidney injury (AKI) by 2025 by calling for global strategies that permit timely diagnosis and treatment of potentially reversible AKI. As Mehta et al.

M AN U

addressed in the Future Perspective Section “the effect of AKI on morbidity and mortality will be shaped by advances in methods to detect AKI earlier in the disease course”5. Several clinical risk models for prediction of CIN or CI-AKI have been identified, which can be roughly classified into post contrast exposure models and pre-procedural or

TE D

pre-contrast exposure models. The majority of those proposed models were post-contrast models that be used when patients have been exposed to contrast medium, and the accuracy of these models yielded C-statistics ranging from 0.61 to 0.924, 6-14. Mehran’s risk score

EP

model8 yielding a C-statistics of 0.69, is one of the most influential models, so is the

AC C

Bartholomew score9. However, those post-contrast models may be a little time-insensitive and to develop a risk model that using less easily calculated variables which can achieve quick applicability of CIN or CI-AKI risk before patients being subjected to contrast exposure may be more clinically attractive9. Though Maioli15, Chen16 and Inohara17 proposed preprocedural scores for predicting the risk of CI-AKI and CIN with seven or more factors, both of them were a lack of quantitative biomarkers that may serve as important predictors in risk score models and have no data on the association of developed score with long-term outcomes. 4

ACCEPTED MANUSCRIPT Thus, there is an need for a more objective and simple identification tool which can be readily available for prediction of CI-AKI on admission. The goal of this study is to establish a novel simple and accurate risk score that can early predict CI-AKI before procedure by considering biomarkers such as N-terminal pro-brain natriuretic peptide (NT-proBNP) and

RI PT

high sensitive C-reactive protein (Hs-CRP).

METHODS

SC

Subjects

This prospectively designed observational study included all consecutive patients (n=3957)

M AN U

who underwent coronary angiography or PCI between January 2010 and October 2012 in Guangdong General Hospital according to the institutional protocol. As in a previous study18 and the PRECOMIN (Predictive Value of Contrast Volume to Creatinine Clearance Ratio, ClinicalTrials.gov NCT01400295) study, we included patients aged ≥18 years who agreed to

TE D

stay in the hospital for 2–3 days after coronary angiography. The exclusions were identified according to the updated European Society of Urogenital Radiology Contrast Media Safety Committee guidelines3 (Figure 1).

EP

Finally, 1777 patients were included for analysis. Follow-up events were carefully

AC C

monitored and recorded by trained nurses through office visits and telephone interviews after coronary angiography. Outcomes and MACEs were reviewed by an event judgment group. The mean follow-up time was 2.36 ± 0.81 years (median, 2.24 years; interquartile range, 1.72–2.91 years). The institutional Ethics Research Committee approved the study, and all patients provided written informed consent to participate. Coronary angiography Coronary angiography was performed according to standard clinical practice, using standard guide catheters, guidewires, balloon catheters, and stents via the femoral or radial approach. 5

ACCEPTED MANUSCRIPT The contrast dose was left to the discretion of the interventional cardiologist. All patients received nonionic, low-osmolarity contrast agents (either Iopamiron or Ultravist, both 370 mg I/mL). Subjects were treated according to AHA/ACCF guidelines19, 20. According to the local institutional protocol18, serum creatinine concentrations were measured in all patients at

RI PT

hospital admission and on days one, two, and three after coronary angiography. Outcome definition

The primary endpoint was CI-AKI, defined as an increase in serum creatinine of

50% or

SC

0.3mg/dl from baseline within 48 hours (patients who had missed day two serum creatinine,

M AN U

only day one values were used)21. Additional endpoints were CIN25 and CIN0.5. CIN25 and CIN0.5 were defined as an increase in serum creatinine of

25% and

0.5mg/dl

respectively from baseline within 72 hours of contrast exposure (the rule same with CI-AKI for missing serum creatinine was applied). We also recorded the in-hospital clinical outcomes, including renal replacement therapy, acute heart failure, re-AMI, use of IABP, arrhythmia,

TE D

stroke, bleeding and death, and long-term major adverse clinical events (MACE), including death, non-fatal myocardial infarction, target vessel revascularization, CIN requiring renal

EP

replacement therapy, stroke, and re-hospitalization. Since we expected missing data for measurements of serum creatinine made later than day one due to usual practice was

AC C

relatively high (55% in this study), we performed sensitivity analysis in the validation of the risk score applying the dataset that having serum creatinine values on both day 1 and day 2. Furthermore, because there is a lack of consensus on how to define CIN or CI-AKI21, we also validated our risk score by applying to predict CIN25 and CIN0.5. Candidate predictors Candidate predictors included clinical variables such as age (years), gender, hypertension, hypotension, diabetes mellitus, anemia, pre-existing CKD defined as pre-admission CrCl (calculated by Cockcroft–Gault formula) < 60ml/min/1.73m2, left ventricular ejection 6

ACCEPTED MANUSCRIPT fraction (LVEF) <40%, CHF, ALB < 35 g/L, uric acid (> 417 µmol/L for male, 357 µmol/L for female), serum creatinine <1.5 µmol/dl, blood urea nitrogen (BUN), Hemoglobin (HG), NT-proBNP (pg/ml), and Hs-CRP (mg/L). All these variables were assessed and recorded on admission. The baseline serum creatinine measured before procedure were used to develop

RI PT

the risk (score) model for early prediction of CI-AKI, and those in 48h after the procedure were used to assess the development of CI-AKI. Sample size and missing data

SC

Sample size determination for the observational study is rather difficult, especially in multiple regression models. We applied the rule of thumb. Namely, events per variable (EPV) being

M AN U

10 or greater under this circumstance. We expected about six significant clinical factors in the risk model. Thus, the EPV value was 10.7 (64/6) in our study.

There were 312 patients with incomplete measurements, 2 for diabetes mellitus, 4 for hypotension, 21 for anemia, 15 for CHF, 5 for CKD, 231 for LVEF, 40 for ALB, 22 for BUN,

TE D

and 21 for HG. We assumed missing data occurred at random depending on the clinical variables and utilized multiple imputations. All analyses were performed based on imputed complete cases.

EP

Statistical analysis

AC C

All eligible patients were randomized with a 3:2 ratio to create a development (training) dataset and a validation dataset, respectively. Based on the development dataset, a risk model and a risk score model were developed sequentially. Candidate predictors that were significant at P<0.05 in univariate analysis and were biological interest or clinical important were included in the development of the risk model. The clinical meaningful interactions between predictors were also examined. A bootstrap method was used to select the best subset of risk factors to avoid overfitting. Variables that were selected in at least 70% of 1000 bootstrap repeats were included in the final multivariate models. For the risk score model, the 7

ACCEPTED MANUSCRIPT scoring method similar to Sullivan et al.22, was employed based on the developed risk model. For scoring purpose, the continuous variables were divided into categories regarding clinical significance, such as the cut-off of 400, 800 and 1500 pg/ml for NT-proBNP23, 1, 3 and 7 for Hs-CRP24, respectively. The risk score of CI-AKI was categorized into four levels (low-risk,

RI PT

moderate-risk, high-risk, very high-risk) to enhance the clinical utility of the risk score model. Pearson's contingency coefficient was used to measure the degree of association between the score levels, and the risk of CI-AKI and the Cochran-Armitage test was applied to examine

SC

the trend.

The predictive accuracy of the risk model and the risk score model were assessed by both

M AN U

discrimination measured by C-statistic and calibration evaluated by Hosmer-Lemeshow χ² statistic and calibration plot. The bootstrap method with 1000 replications was used to perform an internal cross-validation of the risk score model. The average of C-statistic and χ² statistic was reported. Furthermore, split-sample validation of the risk score model was

TE D

conducted to assess the stability of the model. The area under ROC curves was compared using the nonparametric approach of DeLong et al.25. The details of multiple imputation methods, scoring methods, and other statistical analysis

EP

methods are provided in Supplementary file part 1. The report of our study strictly followed the Transparent Reporting of a multivariable prediction model for Individual Prognosis Or

AC C

Diagnosis (TRIPOD) statement26, 27.

RESULTS

All patients A total of 1777 consecutive patients (mean age: 63 ± 11 years, 429 female; the flow of participants is presented as Figure 1) who underwent coronary angiography (mean contrast volume 124.92 ± 67.25 mL, mean CrCl 71.72 ± 26.90 mL/min) were included in the final 8

ACCEPTED MANUSCRIPT analysis (details of the baseline characteristics are shown in Supplementary Table S1). Among the 1777 patients, 106 (5.97%) patients developed CI-AKI. Patients of 5.95% (64/1076) developed CI-AKI in the development dataset, while 5.99% (42/701) in the validation dataset. All the clinical factors considered were balanced between the development

RI PT

and validation dataset (Supplementary Table S2). Table 1 shows that a total of 14 variables were significantly associated with the development of CI-AKI in univariate analysis. After including biological interest gender and clinically important diabetes mellitus, 16 variables

SC

were finally used in risk model development and no significant two-way interaction was founded.

M AN U

Risk model development

After variable selection by a bootstrap technique (the details are provided in Supplementary Figure S1-S3, Table S3-S4), age >75 yrs., serum creatinine value >1.5 mg/dl, primary PCI, log-transformed Hs-CRP and log-transformed NT-proBNP were selected as the best subset of

TE D

risk factors to develop the risk model (Table 2). The risk model had excellent discriminative power with a C-statistic of 0.809 (95%CI: 0.749-0.870) and was well calibrated with Hosmer-Lemeshow χ² statistic of 2.016 (P=0.981).

EP

Risk score development

AC C

Scores for all predictors are presented in Table 3, and an online calculator http://www.echobelt.org/ciakirisk/ is also provided. The total risk score ranges from a minimum value of 0 (lowest risk) to a maximum value of 36 (highest risk), with corresponding predicted probabilities of developing CI-AKI ranging from 0.85% to 65.11%. Validation of risk score Applying the risk score model to the development dataset gave good discrimination with a C-statistic of 0.809 (95%CI: 0.753-0.866) and high calibration with a χ² statistic of 4.162 (P=0.842) (Figure 2). Also, the bootstrapping internal validation yielded an average 9

ACCEPTED MANUSCRIPT C-statistic of 0.809 (bias-corrected 95%CI: 0.751~0.860). The split-sample validation based on validation dataset yielded results similar to those from the bootstrapping validation with a C-statistic of 0.798 (95%CI: 0.727-0.870) and a calibration χ² statistic of 6.390 (P=0.495). The ROCs (Figure 2) and the CI-AKI proportions (Figure 3A) from both development and

RI PT

validation datasets were concordant with each other. Sensitivity analysis based on a dataset that having serum creatinine values on both day one and day two showed that the risk score still yielded very high discrimination power (Figure 2), and the observed incidence of CI-AKI

SC

was nicely consistent with the predicted ones (Figure 3B). We also validated our risk score in Coronary Angiography only population and PCI population separately (Supplementary

M AN U

Figure S4), and good discriminative power was demonstrated both in development dataset (with a C-statistic of 0.791 and 0.819) and validation dataset (with a C-statistic of 0.840 and 0.779). When applying our risk score to the other definitions of CIN (CIN0.5 and CIN25%), good discriminative power was demonstrated, and the split-sample validation based on

TE D

validation dataset yielded results similar to those from the development dataset (Supplementary Figure S5).

Clinical implications of the risk score model

EP

We arbitrarily categorized the risk scores into four levels to enhance the clinical use of risk

AC C

score model. Four possible definitions of low-risk, moderate-risk, high-risk, and very high-risk groups were provided (details provide in Supplementary Table S5-S6). As Supplementary Figure S6 (for the first definition classified by quartering the risk score) shown, the risk scores is highly and positively associated with the risk of developing CI-AKI (Pearson's contingency coefficient =0.326, P for trend <0.001), other definitions of CIN, in-hospital death and MACE (P for trend <0.001) in both development and validation sets. (For the other three definitions, details were provided in Supplementary Figure S7-S9.) Risk score and long-term clinical outcomes 10

ACCEPTED MANUSCRIPT Patients with a high-risk score presented a worse survival rate than patients with low-level risk score (log-rank analysis P<0.001) (Figure 4A for definition 1), so did the follow-up MACE (log-rank analysis P<0.001) (Figure 4B for definition 1). After adjusting for baseline clinical factors, the higher risk score group remained a significant risk factor for long-term

RI PT

clinical outcomes (Table 4). We also noted the wide 95% CIs of high-risk and very high-risk groups due to the small number of patients falling into those two categories. However, the trend of higher score linking to a higher risk of long-term clinical outcomes was obvious.

SC

(For the other three definitions, details were provided in Supplementary Figure S11-S13 and Table S7.)

M AN U

Risk score, NT-proBNP, V/CrCl, Chen’s score, Inohara’s score and Mehran’s score We validated the discrimination of NT-proBNP, contrast volume to creatinine clearance ratio (V/CrCl), Chen’s score, Inohara’s score and Mehran’s score by using our data. The predictive value of our risk score is significantly higher than the preprocedural NT-proBNP (C-statistic:

TE D

0.809 vs. 0.760, P=0.041), V/CrCl (C-statistic: 0.809 vs. 0.690, P=0.001), Chen’s score (C-statistic: 0.809 vs. 0.635, P=0.034) and Inohara’s score (C-statistic: 0.809 vs. 0.765, P=0.002). When compared with the widely used Mehran’s CIN score, our risk score achieves

EP

similar predictive value (C-statistic: 0.809 vs. 0.784, P=0.335). (Further details are shown in

AC C

Supplementary Figure S14 and Table S8)

DISCUSSION

In the present study, we developed and validated a risk score model for prediction of CI-AKI based on five readily available factors (age >75 years, serum creatinine value >1.5 mg/dl, NT-proBNP, Hs-CRP and primary PCI) at pre-procedure, and we showed that an increasing score confers exponentially increased CI-AKI risk, in-hospital adverse outcomes and long-term death. The proposed risk score showed high discrimination and well calibration 11

ACCEPTED MANUSCRIPT both in development dataset and validation dataset. Because a quantitative score allows different potential decision, we provided four possible definitions (Supplementary Table S5) of low, moderate, high-risk and very high-risk groups for CI-AKI for physicians' choices of cut points that are more or less stringent as needed. This proposed simple risk score allows

RI PT

for immediate identification of high-risk patients before the procedure and appropriate (and timely) risk allocation.

Of note, together with well-known risk factors, the proposed score also includes two

SC

biomarkers, NT-proBNP and Hs-CRP. NT-proBNP, an easily available biomarker, is especially noteworthy since it was such a strong predictor of outcomes in our study. This is

M AN U

not entirely surprising since CHF or elevated right sided pressures are clearly associated with increased risk of CI-AKI. NT-proBNP is a better measure of volume overload and thus should help enhance the predictive ability on the development of CI-AKI. Studies28-30 has shown that it is associated with most of the risk factors included in the Mehran’s score and other

TE D

common known risk factors of CI-AKI. Elevated NT-proBNP levels have also been shown to be a powerful predictor of short-term and long-term outcomes31, and the measurement of serum BNP on admission could help identify patients who are at risk for developing CIN

EP

after primary PCI30. Hs-CRP, as a marker of systemic inflammation, has been proven to be

AC C

associated with an increased risk for CI-AKI in patients undergoing PCI24. Previous studies suggested that various inflammatory-related factors were important contributors to the development of CI-AKI32 and among them, Hs-CRP emerged as a powerful one33. Since the treatment of CI-AKI is rather limited and the prognosis of CI-AKI always ends up with prolonged hospital stay, unfavorable in-hospital and long-term clinical outcomes, a practical and effective solution to this complication is its prevention34. So the advances in methods to detect CI-AKI earlier are urgently needed. Before this study, Maioli15, Chen16 and Inohara17 developed pre-procedural scores for risk of CIN and CI-AKI with seven or more 12

ACCEPTED MANUSCRIPT factors yielding a C-statistics of 0.86, 0.82 and 0.80, respectively. In the present study the predictive power of Chen’s score was far lower than what they have reported (C-statistic: 0.635 vs. 0.82), and significantly lower than our risk score (C-statistic: 0.635 vs. 0.809, P=0.034). The predictive power of Inohara’s score is also significantly lower than our risk

RI PT

score (C-statistic: 0.765 vs. 0.809, P=0.002). We could not compare with Maioli’s score because our study did not consider of two factors (one procedure effected within the past 72 h, pre-procedure creatinine ≥ baseline creatinine) in Maioli’s score. However, the other five

SC

factors in Maioli’s score is similar to what is used in the Chen score. Thus we inferred that the Chen and Maioli’s scores might have similar predictive power. Furthermore, the quantitative

M AN U

biomarkers used in our risk model may be more objective than that used in those three scores. Liu et al.30 showed that NT-proBNP alone may be a promising tool for predicting the risk of CIN, while in our risk score the predictive power is significantly improved by incorporating the other four factors with preprocedural NT-proBNP (C-statistic: 0.809 vs. 0.760, P=0.041).

P=0.001).

TE D

Our score also shows a better predictive power than the V/CrCl (C-statistic: 0.809 vs. 0.690,

Compared to the widely used Mehran’s CIN score, our risk CI-AKI score achieves similar

EP

predictive power but uses readily available pre-procedural factors, and timely pre-procedural

AC C

risk prediction makes many potential benefits in general. It can help the interventional team to establish more individually tailored procedures such as appropriately reduce the dose of contrast volume following with sufficient hydration or combined with pharmacological prophylaxis or to adopt some other precautionary measures on those patients more prone to develop CI-AKI3. Patients at high risk can also choose alternative imaging methods or opt out of the further investigation. It can also allow clinical trials and quality improvement interventions to target patients most likely to benefit from the system based quality improvement efforts provided by Brown and colleagues35 which reduced the rate of CIN by 13

ACCEPTED MANUSCRIPT 20% in consecutive patients with PCI at multiple centers through a multifaceted intervention. Samuel et al.36 recommended that the clinicians should consider using the scores that do not include contrast volume to estimate a patient’s risk of CIN. Thus our risk score is clinically attractive. (Further work to detect whether combining NT-proBNP and Hs-CRP with the

RI PT

Mehran’s score can increase the discrimination is shown in Supplementary Figure S15). Study limitations

The current study has several limitations. First, because this prospective, observational study

SC

was conducted in a single center with a relatively small sample size, the risk score model needs to validate and recalibrate further for more widespread use. Second, 55% (970/1777)

M AN U

patients had no serum creatinine recorded on day two due to about half patients discharged on day one after the procedure. This discordance between the scientific definition of CI-AKI and the clinical practice is an issue whose solution may goes beyond the aims of this study. It is true that any missing may affect the precision of proposed models. However, the sensitivity

TE D

analysis based on dataset having serum creatinine values on both day 1 and day 2 showed that the risk score still yielded very high discrimination power, and the observed incidence of CI-AKI was nicely consistent with the predicted ones. It suggested that the proposed risk

EP

score was valid and stable to overcome the effect of the miss of serum creatinine on day 2.

AC C

Furthermore, when using dataset that have only day 1 values, relatively smaller but still good discrimination power also demonstrated (development set AUC=0.742, validation set AUC=0.721). Third, this model was validated using the split-sample method, which is a relatively simple way to test for overfitting but does not determine the generalizability to independent cohorts. Fourth, the proposed risk score might need further investigation for high-risk patients (severe renal dysfunction, advanced age) and primary PCI patients because of inadequate patients included in the present study, though our model showed excellent performance in these patients. Finally, a larger proportion of patients lost to follow-up may 14

ACCEPTED MANUSCRIPT affect the results regarding adverse clinical outcomes during follow-up.

CONCLUSIONS The generated novel risk score model is a simple and accurate tool for early prediction of

RI PT

CI-AKI in patients after coronary angiography or PCI at pre-procedure. It can be used for both clinical and investigational purposes. It could help clinicians to assess the risk of CI-AKI before contrast exposure, plan and initiate the most appropriate disease management

SC

in time.

M AN U

Acknowledgements:

We are grateful for the many contributions of the participants, physicians, nurses, coordinators, and professional staffs at the clinical center in China.

TE D

Funding:

This work was supported by the National Natural Science Foundation of China [No.

AC C

Disclosures:

EP

81673270/81273191].

None.

References:

1. Best PJM, Lennon R, Ting HH, et al. The impact of renal insufficiency on clinical outcomes in patients undergoing percutaneous coronary interventions. J Am Coll Cardiol. 2002;39:1113-1119. 2. McCullough PA, Adam A, Becker CR, et al. Epidemiology and Prognostic Implications 15

ACCEPTED MANUSCRIPT of Contrast-Induced Nephropathy. The American Journal of Cardiology. 2006;98:5-13. 3. Stacul F, van der Molen AJ, Reimer P, et al. Contrast induced nephropathy: updated ESUR Contrast Media Safety Committee guidelines. Eur Radiol. 2011;21:2527-2541. 4. McCullough PA, Adam A, Becker CR, et al. Risk Prediction of Contrast-Induced

RI PT

Nephropathy. The American Journal of Cardiology. 2006;98:27-36.

5. Mehta RL, Cerda J, Burdmann EA, et al. International Society of Nephrology's 0by25 initiative for acute kidney injury (zero preventable deaths by 2025): a human rights case for

SC

nephrology. Lancet. 2015.

6. McCullough PA, Wolyn R, Rocher LL, Levin RN, O, Neill WW. Acute renal failure

M AN U

after coronary intervention: incidence, risk factors, and relationship to mortality. The American journal of medicine. 1997;103:368-375.

7. McCullough PA, Manley HJ. Prediction and Prevention of Contrast Nephropathy. J Interv Cardiol. 2001;14:547-558.

TE D

8. Mehran R, Aymong ED, Nikolsky E, et al. A simple risk score for prediction of contrast-induced nephropathy after percutaneous coronary intervention. J Am Coll Cardiol. 2004;44:1393-1399.

EP

9. Bartholomew BA, Harjai KJ, Dukkipati S, et al. Impact of nephropathy after

AC C

percutaneous coronary intervention and a method for risk stratification. The American Journal of Cardiology. 2004;93:1515-1519. 10. Marenzi G, Lauri G, Assanelli E, et al. Contrast-induced nephropathy in patients undergoing primary angioplasty for acute myocardial infarction. J Am Coll Cardiol. 2004;44:1780-1785. 11. Wykrzykowska JJ, Garg S, Onuma Y, et al. Value of Age, Creatinine, and Ejection Fraction (ACEF Score) in Assessing Risk in Patients Undergoing Percutaneous Coronary Interventions in the 'All-Comers' LEADERS Trial. Circulation: Cardiovascular Interventions. 16

ACCEPTED MANUSCRIPT 2011;4:47-56. 12. Tziakas D, Chalikias G, Stakos D, et al. Development of an easily applicable risk score model for contrast-induced nephropathy prediction after percutaneous coronary intervention. Int J Cardiol. 2013;163:46-55.

RI PT

13. Gurm HS, Seth M, Kooiman J, Share D. A Novel Tool for Reliable and Accurate Prediction of Renal Complications in Patients Undergoing Percutaneous Coronary Intervention. J Am Coll Cardiol. 2013;61:2242-2248.

SC

14. Park HS, Kim CJ, Yi J, et al. Contrast Volume/Raw eGFR Ratio for Predicting Contrast-Induced Acute Kidney Injury in Patients Undergoing Percutaneous Coronary

M AN U

Intervention for Myocardial Infarction. Cardiorenal Med. 2015;5:61-68.

15. Maioli M, Toso A, Gallopin M, et al. Preprocedural score for risk of contrast-induced nephropathy in elective coronary angiography and intervention. J Cardiovasc Med. 2010;11:444-449.

TE D

16. Chen Y, Fu N, Xu J, et al. A simple preprocedural score for risk of contrast-induced acute kidney injury after percutaneous coronary intervention. Catheter Cardio Inte. 2014;83:E8-E16.

EP

17. Inohara T, Kohsaka S, Abe T, et al. Development and Validation of a Pre-Percutaneous

AC C

Coronary Intervention Risk Model of Contrast-Induced Acute Kidney Injury With an Integer Scoring System. The American Journal of Cardiology. 2015;115:1636-1642. 18. Tan N, Liu Y, Zhou Y, et al. Contrast medium volume to creatinine clearance ratio: A predictor of contrast-induced nephropathy in the first 72 hours following percutaneous coronary intervention. Catheter Cardio Inte. 2012;79:70-75. 19. Wright RS, Anderson JL, Adams CD, et al. 2011 ACCF/AHA Focused Update of the Guidelines for the Management of Patients With Unstable Angina/Non–ST-Elevation Myocardial

Infarction

(Updating

the

2007 17

Guideline).

J

Am

Coll

Cardiol.

ACCEPTED MANUSCRIPT 2011;57:1920-1959. 20. Jneid H, Anderson JL, Wright RS, Adams CD, Bridges CR, Casey DE, Ettinger SM, Fesmire FM, Ganiats TG, Lincoff AM, Peterson ED, Philippides GJ, Theroux P, Wenger NK, Zidar JP, Anderson JL. 2012 ACCF/AHA Focused Update of the Guideline for the

RI PT

Management of Patients With Unstable Angina/Non-ST-Elevation Myocardial Infarction (Updating the 2007 Guideline and Replacing the 2011 Focused Update): A Report of the American College of Cardiology Foundation/American Heart Association Task Force on

SC

Practice Guidelines. Circulation. 2012;126:875-910.

21. Azzalini L, Spagnoli V, Ly HQ. Contrast-Induced Nephropathy: From Pathophysiology

M AN U

to Preventive Strategies. Can J Cardiol. 2016;32:247-255.

22. Sullivan LM, Massaro JM, D'Agostino RB. Presentation of multivariate data for clinical use: The Framingham Study risk score functions. Stat Med. 2004;23:1631-1660. 23. Januzzi JL. NT-proBNP testing for diagnosis and short-term prognosis in acute

TE D

destabilized heart failure: an international pooled analysis of 1256 patients: The International Collaborative of NT-proBNP Study. Eur Heart J. 2005;27:330-337. 24. Gao F, Zhou YJ, Zhu X, Wang ZJ, Yang SW, Shen H. C-Reactive Protein and the Risk

EP

of Contrast-Induced Acute Kidney Injury in Patients Undergoing Percutaneous Coronary

AC C

Intervention. Am J Nephrol. 2011;34:203-210. 25. DeLong ER, DeLong DM, Clarke-Pearson DL. Comparing the areas under two or more correlated receiver operating characteristic curves: a nonparametric approach. Biometrics. 1988;44:837-845.

26. Collins GS, Reitsma JB, Altman DG, Moons KG. Transparent Reporting of a multivariable prediction model for Individual Prognosis Or Diagnosis (TRIPOD): the TRIPOD Statement. Br J Surg. 2015;102:148-158. 27. Moons KGM, Altman DG, Reitsma JB, et al. Transparent Reporting of a multivariable 18

ACCEPTED MANUSCRIPT prediction model for Individual Prognosis or Diagnosis (TRIPOD): explanation and elaboration. Ann Intern Med. 2015;162:W1. 28. Maisel AS, Clopton P, Krishnaswamy P, et al. Impact of age, race, and sex on the ability of B-type natriuretic peptide to aid in the emergency diagnosis of heart failure: results from

RI PT

the Breathing Not Properly (BNP) multinational study. Am Heart J. 2004;147:1078-1084. 29. Pfister R, Sharp S, Luben R, et al. Mendelian randomization study of B-type natriuretic peptide and type 2 diabetes:

evidence of causal association from population studies. Plos

SC

Med. 2011;8:e1001112.

30. Liu Y, He YT, Tan N, et al. Preprocedural N-Terminal Pro-Brain Natriuretic Peptide

M AN U

(NT-proBNP) Is Similar to the Mehran Contrast-Induced Nephropathy (CIN) Score in Predicting CIN Following Elective Coronary Angiography. J Am Heart Assoc. 2015;4:e1410. 31. Patel UD, Garg AX, Krumholz HM, et al. Preoperative Serum Brain Natriuretic Peptide and Risk of Acute Kidney Injury After Cardiac Surgery. Circulation. 2012;125:1347-1355.

TE D

32. Goldenberg I. Nephropathy induced by contrast media: pathogenesis, risk factors and preventive strategies. Can Med Assoc J. 2005;172:1461-1471. 33. Toso A, Leoncini M, Maioli M, et al. Relationship between inflammation and benefits of

EP

early high-dose rosuvastatin on contrast-induced nephropathy in patients with acute coronary

AC C

syndrome: the pathophysiological link in the PRATO-ACS study (Protective Effect of Rosuvastatin and Antiplatelet Therapy on Contrast-Induced Nephropathy and Myocardial Damage in Patients With Acute Coronary Syndrome Undergoing Coronary Intervention). JACC Cardiovasc Interv. 2014;7:1421-1429. 34. Wright RS, Anderson JL, Adams CD, et al. 2011 ACCF/AHA Focused Update Incorporated Into the ACC/AHA 2007 Guidelines for the Management of Patients With Unstable

Angina/Non–ST-Elevation

Myocardial

2011;57:e215-e367. 19

Infarction.

J

Am

Coll

Cardiol.

ACCEPTED MANUSCRIPT 35. Brown JR, Solomon RJ, Sarnak MJ, et al. Reducing Contrast-Induced Acute Kidney Injury Using a Regional Multicenter Quality Improvement Intervention. Circulation: Cardiovascular Quality and Outcomes. 2014;7:693-700. 36. Silver SA, Shah PM, Chertow GM, Harel S, Wald R, Harel Z. Risk prediction models for

AC C

EP

TE D

M AN U

SC

RI PT

contrast induced nephropathy: systematic review. BMJ. 2015;351:h4395.

20

ACCEPTED MANUSCRIPT Figure titles and legends Figure 1 Flow of enrollment of the study participants. (CIN0.5=CIN defined as an increase in serum creatinine of ≥0.5 mg/dl; CIN25%/0.5=CIN defined as an increase in serum creatinine of ≥25% or 0.5 mg/dl).

RI PT

Figure 2 Receiver operator characteristic curves showing the area under the curve and calibration plot with Hosmer-Lemeshow test for goodness of fit result for the risk score. (Sensitivity analysis refers to the validation based on dataset including serum creatinine

SC

values on both day 1 and day 2.)

Figure 3 (A) Increasing risk of contrast-induced AKI with increasing risk score is evident in

M AN U

the development and validation datasets. (B) The observed incidence of CI-AKI was nicely consistent with the predicted ones using the risk score based on dataset having serum creatinine values on both day 1 and day 2.

Figure 4 (A) Cumulative hazard of mortality as a function of time and risk score groups in

TE D

the development dataset. (B) Cumulative hazard of MACE as a function of time and risk

AC C

EP

score groups in the development dataset.

21

ACCEPTED MANUSCRIPT Table 1 Association of pre-procedure characteristics and CI-AKI (development dataset, univariate analysis) Development Dataset Characteristic* CI-AKI(n = 64)

p

RI PT

No CI-AKI(n = 1012)

70 ± 11

Age >75 yrs.

26(40.6%) 18 (28.1%)

Smokers (%) Diabetes (%) Hypertension (%)

LVEF (%)

AC C

LVEF <40%

EP

Previous MI (%)

Serum creatinine (mg/dl) Serum creatinine >1.5 mg/dl CrCl (mL/min)* CKD BUN (mmol/L) Hemoglobin (g/L)

124(12.3%)

<0.001†

236 (23.3%)

0.380†

27(42.2%)

403(39.8%)

0.708

18(28.1%)

251(24.8%)

0.552†

45(70.3%)

590(58.3%)

0.058

5(7.8%)

14(1.4%)

<0.001†

8(12.5%)

158(15.6%)

0.504

7(10.9%)

114(11.3%)

0.936

51.10±13.25

58.64±12.71

<0.001

15(23.4%)

106(10.5%)

0.001†

1.37±0.74

1.00±0.35

<0.001

21(32.8%)

62(6.1%)

<0.001†

60.71±44.66

73.29±26.24

0.029

42(65.6%)

330(32.6%)

<0.001†

6.67±3.54

4.98±2.03

<0.001†

123.14±21.26

133.30±15.11

<0.001†

TE D

Hypotension (%) Dyslipidemia (%)

<0.001

M AN U

Female (%)

63 ± 11

SC

Age (yrs.)

ACCEPTED MANUSCRIPT 56(87.5%)

908(89.7%)

0.572

Anemia (%)

33(51.6%)

335(33.1%)

0.003†

CHF (%)

26(40.6%)

135(13.3%)

<0.001†

Primary PCI

16(25.0%)

81(8.0%)

<0.001†

Uric acid (µmol/L)

406.20±117.65

377.05±105.07

0.048

Hyperuricemia

26(40.6%)

292(28.9%)

0.045†

ALB (g/L)

31.51±5.07

35.67±4.45

<0.001

ALB <35

48(76.2%)

399(40.4%)

<0.001†

M AN U

SC

RI PT

ACEIs /ARBs (%)

1599.0(527.1~6320.2)

249.5(67.2~964.1)

Log NT-proBNP

7.40±2.00

5.52±1.77

Hs-CRP (mg/L)

15.75(4.07~45.30)

3.30(1.36~9.05)

Log Hs-CRP

2.54±1.53

1.30±1.46

<0.001†

2.76±1.19

2.57±0.89

0.338

0.87±0.31

0.94±0.28

0.113

LDL-C (mmol/L) HDL-C (mmol/L) *

TE D

NT-proBNP (pg/ml)

<0.001†

EP

Data are presented as the mean value ± SD or percentage of subjects. MI =

AC C

myocardial infarction; LVEF = left ventricular ejection fraction; CrCl = the creatinine clearance calculated by applying the Cockcroft-Gault formula to the serum creatinine concentration; BUN = blood urea nitrogen; ACEIs/ARBs = angiotensin-converting enzyme inhibitor/angiotensin receptor blocker; PCI = percutaneous coronary intervention; CHF = congestive heart failure; NT-proBNP = N-terminal pro-brain natriuretic peptide; Hs-CRP = High sensitivity C reactive protein; LDL-c = low densith lipoprotein-cholesterol; HDL-c = high densith lipoprotein-cholesterol.

ACCEPTED MANUSCRIPT

EP

TE D

M AN U

SC

RI PT

Variables that were finally used in risk model development.

AC C

†

ACCEPTED MANUSCRIPT Table 2 Univariate and multivariable logistic regression analysis of risk factors which be selected to develop the risk model for predicting CI-AKI (development dataset, n=1076)

Variable

Multivariable analysis

OR

95%CI

P value

RI PT

Univariate analysis *

OR

95%CI P value

7.48

4.18~13.39 0.000

2.85

1.42~5.70 0.003

Primary PCI

3.83

2.08~7.05 0.000

1.76

0.87~3.55 0.117

Age >75 yrs.

4.90

2.88~8.35 0.000

3.24

1.81~5.81 0.000

Log NT-proBNP†

1.86

Log Hs-CRP†

1.73

M AN U

*

SC

Serum creatinine >1.5 mg/dl

1.58~2.20 0.000

1.37

1.12~1.67 0.002

1.45~2.06 0.000

1.34

1.08~1.65 0.007

Because the variables included in our risk model have no missing data, we didn’t

TE D

combine the results of analyses carried out on the ten imputed data sets. CI = confidence interval; OR = odds ratio; Log NT-proBNP = Log-transformed N-terminal

EP

pro-brain natriuretic peptide; Log Hs-CRP = Log-transformed High sensitivity C-reactive protein.

The natural logarithmic transformations of NT-proBNP and Hs-CRP were made

AC C

†

because of their extreme positive skewness.

ACCEPTED MANUSCRIPT Table 3 Risk scores for all of the predictors Risk factors*

Score

0

>75 yrs.

8

Baseline serum creatinine value 0

>1.5 mg/dl

7

M AN U

≤1.5 mg/dl

NT-proBNP <400 pg/ml

0

400~800 pg/ml

2

<1 mg/L

AC C

1~3 mg/L

EP

>1500 pg/ml

4

TE D

800~1500 pg/ml

Hs-CRP

SC

≤75 yrs.

RI PT

Age

8

0 3

3~7 mg/L

6

>7 mg/L

9

Primary PCI

*

No

0

Yes

4

NT-proBNP = N-terminal pro-brain natriuretic peptide; Hs-CRP = High sensitivity C

ACCEPTED MANUSCRIPT

AC C

EP

TE D

M AN U

SC

RI PT

reactive protein.

ACCEPTED MANUSCRIPT Table 4 Univariate and multivariable Cox regression analysis of candidate risk factors for follow-up death and MACE (development dataset, n=1076) Univariate analysis HR

95%CI

p value

HR

≥27

99.99

21.90~456.56

<0.001

44.30

18~26

35.81

8.08~158.69

<0.001

9~17

10.66

2.44~46.62

0.002

Female

0.86

0.40~1.87

CHF

2.66

CKD

8.17

Anemia

2.08

Diabetes

1.49

Death Risk score

2.93~76.89

0.001

7.74

1.68~35.74

0.009

0.705

0.53

0.23~1.23

0.141

1.37~5.15

0.004

0.66

0.31~1.43

0.292

3.76~17.73

<0.001

2.38

0.98~5.79

0.056

1.12~3.87

0.021

0.51

0.20~1.26

0.143

0.77~2.90

0.234

1.00

0.48~2.07

1.000

SC

<0.001

TE D

M AN U

15.01

1.55

0.21~11.28

0.666

1.25

0.14~11.10

0.842

1.74

0.87~3.49

0.117

0.96

0.43~2.11

0.911

AC C

Hypertension

p value

7.97~246.28

EP

Hypotension

95%CI

RI PT

Variables

Multivariable analysis

*

Hyperuricemia

3.15

1.69~5.87

<0.001

2.74

1.39~5.42

0.004

ALB <35

3.79

1.89~7.60

<0.001

1.09

0.50~2.36

0.837

BUN

1.21

1.15~1.28

<0.001

1.07

0.99~1.15

0.091

Hemoglobin

0.97

0.96~0.98

<0.001

0.97

0.94~0.99

0.017

MACE Risk score

ACCEPTED MANUSCRIPT 1.98

<0.001

2.05

1.03~4.08

0.041

18~26

2.04

1.32

0.001

1.31

0.77~2.25

0.320

9~17

1.31

0.94

0.108

1.09

0.76~1.58

0.634

Female

0.78

0.54

0.179

0.64

0.43~0.95

0.026

CHF

1.88

1.34

<0.001

1.48

CKD

1.64

1.23

0.001

1.23

Anemia

1.44

1.07

0.016

0.93

Diabetes

1.41

1.03

0.031

Hypotension

1.41

0.52

0.500

Hypertension

1.19

0.88

Hyperuricemia

1.22

ALB <35

1.48

BUN

1.09

Hemoglobin

0.99

RI PT

3.57

0.047

0.86~1.74

0.257

0.63~1.37

0.718

0.92~1.80

0.142

1.62

0.58~4.48

0.356

0.249

1.03

0.75~1.42

0.865

0.9

0.200

1.15

0.83~1.59

0.407

1.1

0.009

1.01

0.71~1.43

0.968

1.05

<0.001

1.03

0.97~1.08

0.336

0.98

<0.001

0.99

0.98~1.00

0.029

SC

1.01~2.17

1.29

M AN U

TE D

*

≥27

EP

CHF = congestive heart failure; BUN = blood urea nitrogen; MACE = Death,

AC C

non-fatal myocardial infarction, target vessel revascularization, renal replacement therapy, stroke, and re-hospitalization.

AC C

EP

TE D

M AN U

SC

RI PT

ACCEPTED MANUSCRIPT

AC C

EP

TE D

M AN U

SC

RI PT

ACCEPTED MANUSCRIPT

AC C

EP

TE D

M AN U

SC

RI PT

ACCEPTED MANUSCRIPT

AC C

EP

TE D

M AN U

SC

RI PT

ACCEPTED MANUSCRIPT

ACCEPTED MANUSCRIPT All authors have participated in the work and have reviewed and agree with the content of the article. None of the article contents are under consideration for publication in any other

RI PT

journal or have been published in any journal. No portion of the text has been copied from other material in the literature (unless in quotation marks, with citation).

SC

I am aware that it is the author's responsibility to obtain permission for any figures or

AC C

EP

TE D

M AN U

tables reproduced from any prior publications, and to cover fully any costs involved.