|Home | About | Journals | Submit | Contact Us | Français|
Background. SCreening for Occult REnal Disease (SCORED) is a novel screening guideline recently developed to identify individuals with a high likelihood of having prevalent chronic kidney disease (CKD). This simple scoring system, developed from general US representative samples and independently validated, was shown to outperform current clinical practice guidelines. Recently, CKD screening in individuals with cardiovascular disease (CVD) has been emphasized. We therefore evaluated the SCORED model in CVD patients in order to better understand the implications of CKD screening in this population.
Methods. Two clinical trials that enrolled patients with heart attack (N = 2481) or stroke (N = 3680) were combined to create our sample. The performance of the SCORED guideline was evaluated by standard diagnostic measures. Correlations among various risk scores and their predictive abilities for recurrent CVD were ascertained.
Results. For heart attack and stroke patients, respectively, the SCORED guideline yielded sensitivity of 94 and 97%, specificity of 27 and 11%, positive predictive value of 32 and 30%, negative predictive value of 93 and 89%, with AUC of 0.75 and 0.68. SCORED was strongly correlated with other risk scores and exhibited a similar performance in the prediction of recurrent CVD.
Conclusions. The higher risk of CKD in CVD patients with high SCORED values is demonstrated. This simple education and screening tool may help promote awareness of CKD in CVD patients, in addition to general populations, and assess the CKD risk and its relationship with recurrent CVD.
Cardiovascular disease (CVD) and chronic kidney disease (CKD) are each major public health problems that share risk factors and present in similar populations [1,2]. Yet, awareness for potential CKD by healthcare professionals and lay persons remains low among at-risk populations, including those with CVD. Indeed, a joint science advisory board from the American Heart Association (AHA) and the National Kidney Foundation (NKF) recently recommended screening for decreased kidney function among patients with known CVD . Identifying individuals with CKD in this population may provide an opportunity to adopt measures that slow the progression of kidney disease and prevent subsequent CVD.
We recently developed a user-friendly tool to systematically identify individuals with a high likelihood of having CKD, SCreening for Occult REnal Disease (SCORED) . This algorithm identified nine demographic and medical variables and provided a simple scoring system. SCORED has been validated in the general population, community samples and clinical settings; the screening tool also performed favourably compared to the NKF's Kidney Early Evaluation Program (KEEP) guidelines [4–6].
We now seek to test SCORED in a population with known underlying CVD. Our goals are 2-fold: first, we intend to evaluate test characteristics of SCORED in diverse CVD patients. Second, we would like to assess correlation of SCORED with other algorithms suitable for CVD and to compare their abilities for secondary CVD prediction.
We used the data collected from two recent multi-centre, cardiovascular clinical trials, Enhancing Recovery in Coronary Heart Disease (ENRICHD) and Vitamin Intervention for Stroke Prevention (VISP), in our investigation.
ENRICHD is a randomized controlled trial (RCT), sponsored by the National Heart, Lung, and Blood Institute, that tested a hypothesis that a new psychosocial intervention, devised to decrease depression and to increase social support, further improves a composite endpoint of ‘death and nonfatal reinfarction’ after acute myocardial infarction (MI) . Over 3000 participants were screened at 73 hospitals affiliated to eight academic sites in the United States, and 2481 were randomized into cognitive behaviour therapy or usual cardiology care. The trial was conducted from 1996 to 2001 with an average follow-up of 29 months.
VISP, sponsored by the National Institute of Neurological Disorders and Stroke, is an RCT undertaken to study the effectiveness of homocysteine-lowering therapy for recurrent vascular events in patients with non-disabling stroke . VISP aimed to determine whether high doses of folic acid, vitamin B6 and vitamin B12, given to lower total homocysteine levels, further reduce the risk of recurrent stroke over a 2-year period compared with low doses of these vitamins. A total of 3680 adults with non-disabling cerebral infarction participated in this study at 56 hospitals or medical centres across the USA, Canada and Scotland in 1996–2003.
The SCORED risk factors, along with additional demographic characteristics, personal health conditions and clinical information at randomization were retrieved from the databases. Unfortunately, not all variables included in SCORED were available in these studies. Specifically, peripheral vascular disease information was absent in VISP, anaemia information was absent in ENRICHD and urinalysis was not conducted in both studies.
For kidney function, we used estimated glomerular filtration rate (eGFR) from the Modification of Diet in Renal Disease study equation :
CKD was defined as an eGFR < 60 mL/min/1.73 m2, which corresponds to stage 3 or higher kidney disease . (Throughout the paper, we will designate eGFR < 60 mL/min/1.73 m2 as CKD rather than CKD stage 3–5.)
Baseline characteristics of the ENRICHD and VISP participants were summarized by descriptive statistics.
The potential value of SCORED was evaluated by estimating the percentage of subjects classified into ‘high risk’ (defined by score ≥4) as well as standard diagnostic criteria such as sensitivity, specificity, positive predictive value (PPV), negative predictive value (NPV) and area under the receiver-operating-characteristic curve (AUC) . In addition, we tested a new classification rule for defining a high-risk group as having at least one CVD risk factor (i.e. hypertension, diabetes, heart failure or peripheral vascular disease) included in SCORED.
We also fitted the SCORED model to each study. The association between each risk factor and CKD was summarized in terms of odds ratio, confidence interval and statistical significance. CKD prevalence by total score was estimated for general healthy individuals using the National Health and Nutrition Examination Survey (NHANES), Atherosclerosis Risk in Communities (ARIC) and Cardiovascular Heart Study (CHS) (that we used in our previous investigations) and for CVD patients from ENRICHD and VISP.
Next, Pearson correlation coefficients were calculated among SCORED and other well known risk scores in ENRICHD because many of the necessary variables were not available in VISP. Other risk scores include Charlson comorbidity index (for 10-year survival) , Framingham score (for incident coronary heart disease) , Jaffe scores (one for recurrent MI or mortality and the other for all-cause mortality)  and Stroke Prognosis Instrument (SPI) II (for recurrent stroke or death) . For prediction, we used ‘recurrent MI or death’ that was the primary endpoint in ENRICHD and ‘recurrent stroke’ that was the primary endpoint in VISP. In addition to discrimination, model-fit statistics such as Akaike and Bayesian information criteria (AIC and BIC) were computed. It is important to evaluate prediction models based on more than one criterion, especially beyond AUC, as each method addresses different aspects of prediction [16–19].
All analyses were performed using the SAS version 9.1 statistical software. Two-sided hypothesis tests with a 5% type I error were adopted for all statistical inferences.
A total of 2145 participants from ENRICHD and 3640 participants from VISP were included for analyses after excluding participants with missing data in key variables. Participants from VISP were, on average, 5 years older than participants from ENRICHD. ENRICHD had more female and minority participants. Notably, the mean eGFR (75 mL/min/1.73 m2) and the prevalence of CKD (27–28%) were similar between the two studies (see Table Table11).
In ENRICHD, sensitivity, specificity, PPV and NPV of the SCORED guideline were estimated to be 94%, 27%, 32% and 93%, respectively. By this rule, 78% of ENRICHD participants were defined to be at high risk for CKD. The application to VISP provided sensitivity of 97%, specificity of 11%, PPV of 30% and NPV of 89%, while 91% of VISP participants were defined to be at high risk. The rule based on the absence versus the presence of CVD risk factors generally yielded reduced test characteristics, although SCORED yielded low specificity. As an additional comparison, we used NHANES data from generally healthy individuals to evaluate three screening rules: SCORED, ‘CVD risk factors only’ and KEEP (Table (Table2).2). Table Table33 presents logistic regression analyses with the SCORED risk factors as predictors of CKD. Most of the risk factors were statistically significant, and the age effect was monotone.
Figure Figure11 presents a risk assessment chart for general healthy adults as well as CVD patients that would be useful for clinicians and lay persons for education and counselling purposes. As the score rises, a quadratic increase in the risk of CKD is observed in both populations, and the higher risk for CKD among CVD patients is clearly demonstrated.
We found that SCORED and previously validated CVD risk scores were highly correlated with one another: correlation coefficients between SCORED and these other risk scores ranged from 0.58 to 0.75 in ENRICHD (see Table Table4).4). This is not surprising given that many risk factors included in these scores are shared or correlated with one another.
Prediction characteristics of SCORED and traditional CVD risk scores for recurrent CVD events are presented in Table Table5.5. A total of 505 and 298 participants had the primary clinical events in ENRICHD and VISP, respectively. Multiple regression with SCORED risk factors yielded AUC of 0.69 for recurrent MI or mortality, while simple regression with total score from SCORED as a single predictor yielded AUC of 0.66. Other risk scores showed AUC of 0.67–0.72. As expected, Jaffe scores demonstrated the best model fits (with dramatically lower AIC/BIC) as those were developed from ENRICHD. Notably, impaired renal function (i.e. creatinine ≥1.3) had the largest risk ratio in Jaffe models . For recurrent stroke, SCORED and SPI II showed similar AUC and AIC/BIC. Overall, SCORED, which is a CKD screening/prediction model, appeared to perform well in secondary CVD prediction.
In this paper, we evaluated the SCORED algorithm in RCT participants that represent a diverse, multi-ethnic CVD patient population. Compared to healthy individuals that we tested in previous publications, greater numbers of individuals were identified with elevated risk by SCORED. Specificity was decreased and PPV was increased, while high sensitivity and NPV and moderate to high AUC values were maintained.
In the general population, the SCORED model was designed to identify individuals with undiagnosed CKD who could be referred for further laboratory evaluation and follow-up tests. We expect that the vast majority of CVD patients will have serum chemistry values performed by their primary care physicians and/or cardiologists/ neurologists. We therefore envision the SCORED screening tool serving purposes other than merely identifying CKD.
First, SCORED was designed as a simple checklist for patients to learn about their risk for kidney disease. While a cardiologist may be aware that his/her patient has underlying renal disease, that information may not be routinely communicated to a patient during a typical clinic visit or phone conversation; alternatively, the cardiologist may inappropriately assume that the primary care physician has already discussed the concomitant renal disease with the patient. Continued late referral has been reported in not only primary care settings but also in high-risk patients [20–25]. The SCORED tool is patient-friendly and easy to use; many high-risk patients may not know what their creatinine is (or what this measurement means) yet will be able to complete and understand their SCORED testing. SCORED can empower CVD patients to introduce the subject of CKD with their care providers and motivate themselves to be screened (e.g. based on their self-assessment using Figure Figure11 in this paper).
Second, SCORED can re-emphasize the importance of early kidney disease to cardiologists, neurologists and primary care physicians who may not consider small elevations in creatinine to be significant markers of kidney disease. As eGFRs are appearing more routinely in laboratory reports, this phenomenon of under-appreciating creatinine values should dissipate , but SCORED would provide another way to highlight to care providers that, in certain populations, even early deteriorations in renal function require heightened surveillance and, often, concomitant care by a nephrologist. In general, these duties are managed best by nephrologists.
There are some limitations to our investigation. Firstly, the diagnosis of CKD was based on only one determination of eGFR. More than one measurement over time is recommended for accurate clinical diagnosis capturing the ‘chronicity’ definition [10,26]. This is, unfortunately, a common problem in many epidemiologic studies, RCTs, and even some clinical settings. Also, eGFR derived from the MDRD formula might not be optimal, although its utility is regarded as, realistically, the current best definition for CKD [3,27–30]. Secondly, some variables included in SCORED were not available in these datasets, including proteinuria. However, SCORED has been shown to be robust to a few missing variables [4,5], and we intentionally did not impute missing variables in order to reflect real scenarios encountered in practice. For example, people may not know the status of some risk factors, or SCORED can be implemented using administrative database or retrospective chart review possibly with limited information . Particularly, low awareness of proteinuria and less availability of the associated data have been pointed out and improvement has been called for [2,5,31–33].
The importance of kidney disease screening, education and risk factor awareness continue to be emphasized globally [33–39]. Despite all of these efforts, CKD is still severely under-recognized and screening is not routinely conducted either in community settings or among high-risk individuals such as CVD patients or family members of end-stage renal disease patients [3,30]. SCORED would serve as a simple but useful screening and educational tool. It would be reasonable to encourage the use of SCORED in community screening settings and to require eGFR and urinalyses in CVD settings, in keeping with the AHA/NKF statement, recognizing that almost 80–90% of CVD patients were classified at high risk for CKD by SCORED. Moreover, common risk factors may support the development of unified prevention and management strategies for CVD and CKD.
The ARIC, CHS and ENRICHD studies were supported by the NHLBI, and the VISP study was supported by the NINDS. This manuscript was prepared using limited access datasets and does not necessarily reflect the opinions or views of these studies or agencies. Drs HB and MM were partially supported by CTSC grant at Weill Cornell Medical College (UL1-RR024996). We thank Ms Ya-lin Chiu for her help in programming the Framingham risk score. We also thank reviewers and editors for providing valuable comments.
Conflict of interest statement.None declared.