KEY POINTS
Current knowledge
• Cardiovascular disorders represent the leading cause of death in dialysis patients.
• Alteration of FGF 23, fetuin A, PTH, calcium, phosphorus and vascular calcifications play a fundamental role in cardiovascular mortality in these patients.
Contribution of the article to current knowledge
• The vascular calcifications quantified by the Adragao score were the only independent predictor of cardiovas cular mortality at 10 years of follow-up in end-stage renal disease patients.
• This score represents a simple, useful and superior tool to the biomarkers of bone and mineral metabolism.
The first cause of death in patients with chronic kidney disease (CKD) is cardiovascular events1,2. The risk of their appearance increases progressively according to the stage of CKD, reaching up to 20 times higher than the general population3. Data from the Argentine Registry of Dialysis and Transplants reflects something similar, showing that the cardiovascular and cerebrovascular causes represent 46.8% of the total deaths on chronic dialysis in Argentina4.
Both bone and mineral metabolism (BMM) alterations, and vascular calcifications (VCs) have an important role in long-term mortality in subjects with CKD5-10.
The VCs detected by simple radiographs of the hands and pelvis have a good correlation with coronary calcifications and vascular territories in dialysis patients. They are also good predictors of hospitalization-free survival, cardiovascular mortality and long-term global mortality11. In patients with CKD vascular calcifications occur decades earlier than in the general population and their progression accelerates at the beginning of dialysis treatment12,13.
Recent studies found that BMM biomarkers, such as fibroblast growth factor 23 (FGF 23) and fetuin A, can also be predictors of mortality in patients at different stages of CKD14-16. In these studies, elevated levels of FGF 23 were associated with mortality, regardless of serum phosphorus values and other known risk factors14. Likewise, high levels of fetuin A were significantly associated with lower valve and coronary calcifications and reduced mortality from any cause17. However, these biomarkers are not yet available for use in clinical practice in our setting and are used only in the field of research.
It is very useful, for a correct diagnosis and manage ment of these patients, to analyze the risk factors associ ated with death and cardiovascular events and to know the usefulness of the new BMM biomarkers and the effect of the presence of vascular calcifications to predict events that affect long-term morbidity and mortality.
The objective of this study was to analyze the role of biomarkers of BMM and VCs as predictors of long-term cardiovascular cause mortality, and find the best cut-off point for these biomarkers, in patients with end stage CKD.
Material and methods
A prospective, observational and analytical cohort study was performed.
The inclusion criteria were: patients with end stage CKD in renal replacement therapy for at least three months of perma nence prior to inclusion in the study, in the hemodialysis (HD) or peritoneal dialysis (PD) modality, prevalent in September 2009, at the Hospital Privado Universitario de Córdoba.
Subjects under 18 years of age, pregnant women, patients requiring renal replacement therapy for less than three months of stay, or for acute renal failure, or for delayed graft function (in the case of kidney transplants) were excluded. In addition, the patients who were hospitalized at the time of the study, those who refused to participate in it and those who did not complete the total of the required complementary studies were excluded. Among the subjects that met the selection criteria, they were invited to participate voluntarily in the study, after signing the informed consent.
Baseline clinical variables were relieved: age (years), sex, body mass index (kg/m2), modality and time of renal replace ment therapy (months), history of arterial hypertension, diabe tes mellitus, coronary heart disease, stroke, disease peripheral vascular, etiology of chronic kidney disease and presence of hemodialysis catheter. Biochemical determinations were performed in peripheral blood: hemoglobin (g/dl), albumin (g/ dl), calcium (mg/dl), phosphorus (mg/dl), PTH (pg/ml), FGF 23 (pg/ml) and fetuin A (g/l). And different radiographic sets were evaluated to assess the degree of vascular calcifications (Adragao and Kauppila scores).
Each subject was followed up until the kidney transplant, dialysis center change (loss of follow-up), death, until the cut of the study on 1/1/2019 or whichever came first.
Adragao score18: It was determined from a simple radio graphic set of hands and pelvis, measuring the absence or presence of VCs. Both radiographs were divided into 4 quad rants, giving each quadrant 1 point when the presence of VCs is detected. This score analyzes the presence of VCs of iliac, femoral, radial and digital arteries and the final value ranges between 0 and 8 points (0-4 image of both hands; 0-4 image of hips; where 0 means absence of VCs and 8 presence in all quadrants).
Kauppila score19: was determined from a lateral abdominal x-ray that includes from T10 to the first two sacral vertebrae. The aorta is identified as a tubular structure in front of the spine, only the segments of the aorta that are in front of the first four lumbar vertebrae (L1-L4) are analyzed. The score allows to divide into mild calcification (corresponds to 1/3 of the length of the vertebral body), moderate (corresponds to 2/3 of the length of the vertebral body) and severe (more than 2/3 of the vertebral body), according with the length of each calcified plaque detected. The anterior and posterior wall of the aorta is taken into account and a score ranging from 0-24 points is obtained, where 0 means absence of VCs and 24 the presence of these in both walls of the aorta and in the entire length it covers the height of the named vertebrae. All images were evaluated by the same operator.
FGF 23: ALPCO ELISA (enzyme immunosorbent assay) was used, which measures the intact FGF 23 molecule. The assay is a “sandwich” technique that uses two polyclonal anti-bodies directed towards epitopes of the amino and carboxylends of the FGF23 molecule. Plasma with EDTA is used for the measurement, and samples in freezer were preserved at -70 degrees until measurement.
Ultrasensitive fetuin-A: Ultrasensitive ELISA of ALPCO was used. The assay is a “sandwich” technique that uses two polyclonal antibodies that bind different epitopes of human fetuin-A, for the measurement of fetuin-A serum was used without special preparation of the patient and like the mea surement of FGF 23 the samples were preserved in serum at -70 degrees until the time of measurement.
Calcium and phosphorus levels were measured with an autoanalyzer (Hitachi 917; Hitachi, Ltd., Tokyo, Japan) and those of intact PTH with electrochemiluminescence (Nichol`s Institute, San Juan Capistrano, Calif., USA).
To analyze the categorical variables, absolute (n) and relative (%) frequencies were used, and for continuous vari ables, mean (X) and standard deviation (SD), or median (M) and interquartile range (IQR), as appropriate. To compare the categorical variables, the X2 test or Fisher’s exact test was used, as appropriate.
As predictors of mortality, the values of calcium, phos phorus, PTH, FGF 23, Fetuin-A, Kauppila score and Adragao score were evaluated. The area under the curve (AUC) of the ROC (receiver operating characteristic) curve analysis and its 95% confidence interval (95% CI) was used. The calculation of the AUC was performed using the method of Hanley and Mac Neil. ROC curves were plotted. Of the predictive variables with statistical significance, the best cut-off point was established using the Youden Index J criterion20. In turn, sensitivity, speci ficity, positive likelihood ratio, negative likelihood ratio, positive predictive value and negative predictive value were evaluated. The risk of cardiovascular mortality from the best cut-off point was quantified using the relative risk (RR) with its 95% CI.
The association between all the variables included in the database and the best predictive tool for cardiovascular mortality were analyzed. To evaluate the association between continuous variables, Pearson’s correlation (r) was used and for the dichoto mous variables the odd ratio (OR) with its 95% CI. Both the best predictive tool for cardiovascular mortality, and its associated variables, were included in the multivariate logistic regression analysis. To this analysis the variables were added: months of permanence in dialysis at the beginning of the study and months of follow-up at the end of the study, at the discretion of the researcher, regardless of the previous association analysis.
All tests were two-tailed and a value of p less than 0.05 was considered statistically significant.
Statistical analysis was performed with the Stata 14 program (StataCorp. LP. College Station, TX) and Medcalc 15.11.4 (MedCalc Software, Ostend, Belgium).
The study was approved by the Research Committee of the Private University Hospital of Córdoba (number 2018_53).
Results
Of the 82 patients undergoing renal replacement therapy at our institution in September 2009, six met some exclu sion criteria (One had started renal replacement therapy less than 3 months before, two with acute renal failure, one with delayed function of the graft, one under 18 years old and one hospitalized patient). Of the 76 that met the eligibility criteria, 30 subjects complied with all necessary complementary studies. All of them agreed to participate in the study (Fig. 1). Baseline characteristics, pathological personal history, specific dialysis characteristics, blood tests in general and specific bone mineral metabolism, and calcification scores of the subjects are summarized in Table 1.
At the end of the follow-up, of the 30 patients, 11 (36.7%) had received a kidney transplant, 2 (6.7%) continued on chronic dialysis, 17 (56.7%) had died, and of these 7 (41.2%) were due to cardiovascular causes. The average follow-up time was 44.1 ± 30.4 months (range = 1.4-112 months).
In the ROC analysis, although calcium (AUC = 0.69; p = 0.082), phosphorus (AUC = 0.61; p = 0.465), PTH (AUC = 0.65; p = 0.175), FGF 23 (AUC = 0.68; p = 0.175), Fetuin-A (AUC = 0.61; p = 0.433) and the Kauppila score (AUC = 0.61; p = 0.461) had high AUC values, the same they did not reach statistical significance to predict long-term cardiovascular death. The only predic tive variable in the ROC analysis for cardiovascular death was the Adragao score (AUC = 0.82; 95% CI = 0.64-0.94; p < 0.001) (Table 2). Figure 2 shows the ROC curves of the different variables in predicting cardiovascular mortality.
According to the Youden Index J analysis, the best cut-off point for the Adragao score to predict cardiovascular mortality was 5 points or more, with a sensitivity of 85.7%, specificity of 78.3%, positive likelihood ratio of 3.94, nega tive likelihood ratio of 0.18, a positive predictive value of 54.5% and a negative predictive value of 94.7%.
The incidence of cardiovascular death in the group of patients with an Adragao score below 5 points was 5.3% (1/19) and in the group with a score of 5 points or more, it was 54.6% (6/11), there being a significant differ ence between both groups (p = 0.002). Subjects with an Adragao score of 5 points or more have a 10-fold higher risk of cardiovascular mortality than those with a score lower than 5 (RR = 10.36; 95% CI = 1.4-75.2; p = 0.002).
Of all the variables included in our database, the only ones associated with an increase in the Adragao score were: age (r = 0.43; p = 0.018), Kauppila score (r = 0.39; p = 0.034), the history of diabetes mellitus (OR = 10.2; p = 0.008) and coronary heart disease (OR = 7.1; p = 0.029). In multivariate analysis, Adragao score was a predictive tool for cardiovascular death regardless of the variables mentioned above, and also adjusted for the time spent on dialysis at the beginning of inclusion in the study and follow-up time during the study (Adjusted OR = 1.77; 95% CI = 1.06-2.96; p = 0.028) (Table 3).
Discussion
Cardiovascular disease is the leading cause of death in patients with CKD on dialysis1-4 and alterations in BMM and VCs play an important role in its development.
Over time, new molecules involved in MOM and in the physio pathogenesis of VCs such as FGF 23 and Fetuin A have been discovered. According to recent studies, both high FGF 23 values and decreased Fetuin A values increase the risk of cardiovascular events21-23 and death14-16 in patients with CKD. Currently, serum determination of these markers is expensive and there is little availability for use in clinical practice. In our study, neither of the two determinations was useful to predict cardiovascular mor tality over time in patients with terminal CKD in different dialysis modalities.
In Argentina, there is a high prevalence of VCs in sub jects with CKD on dialysis detected by plain radiography24. A recent meta-analysis shows that the Kauppila score is a significant predictor of all-cause mortality and cardio vascular events in dialysis patients25. However, this score does not appear to be higher than the Adragao score to predict mortality26, as our study also shows. Adragao et al.18 showed that patients with CKD in hemodialysis who have an Agragao score greater than 3 have almost four times greater risk of cardiovascular mortality than those with a lower score. This paper also evaluates the role of the phospho-calcium product and PTH in cardiovascular mortality, but not of FGF 23 and fetuin A, since at the time of that study, these molecules were still poorly understood. In our trial, the cut-off point of the Adragao score with the best predictive capacity for cardiovascular death was somewhat higher than 3 points (5 points or more), with a risk of cardiovascular death increased 10 times in these patients. This tool was superior in predictive capacity not only of the traditional MOM analytical determinations such as calcium, phosphorus and PTH, but also of the new biomarkers such as FGF 23 and Fetuin A, and that the Kauppila score.
It is important to keep in mind that while BMM biomark ers reflect the risk to which an individual is exposed at the time of measurement27, VCs images quantified from the scores represent the cumulative result of prolonged exposure to multiple factors of risk over time28.
The increase in the Adragao score was associated with traditional clinical variables such as age, history of diabetes mellitus and coronary heart disease, and also calcifications in other territories such as the aorta (Kaup pila score). Like the time spent on dialysis at the beginning of each patient’s study and the follow-up time at the end of the study, these variables mentioned above could also be risk factors for cardiovascular death and therefore could function as confounding factors for predictive ability of the Adragao score on cardiovascular death. To avoid this bias and evaluate the true role of the score, the adjusted multivariate analysis was performed taking into account all these variables and thus demonstrating their true utility.
Both the recommendations of the Spanish Society of Nephrology in 201129, the Argentine Society of Nephrology in 201730, the KDIGO guides in 201731, and other authors28,32 consider reasonable the use of VCs information to guide the management of BMM alterations.
The main strengths of our study are the evaluation of new biochemical markers of BMM such as FGF 23 and fetuin A, as there are few publications evaluating these measurements in dialysis patients as predictors of cardiovascular mortality, and prospective follow-up with prediction of long-term hard events.
In conclusion, vascular calcifications quantified from the Adragao score were the only independent predictor of long-term cardiovascular mortality. The best cut-off point for the Adragao score was 5, with a high predictive value. This score represents a simple, low cost, useful, accurate and superior tool to the new BMM biomarkers.