Estimation of Baseline Serum Creatinine with Machine Learning

<b><i>Introduction:</i></b> Comparing current to baseline serum creatinine is important in detecting acute kidney injury. In this study, we report a regression-based machine learning model to predict baseline serum creatinine. <b><i>Methods:</i></b> We developed and internally validated a gradient boosting model on patients admitted in Mayo Clinic intensive care units from 2005 to 2017 to predict baseline creatinine. The model was externally validated on the Medical Information Mart for Intensive Care III (MIMIC III) cohort in all ICU admissions from 2001 to 2012. The predicted baseline creatinine from the model was compared with measured serum creatinine levels. We compared the performance of our model with that of the backcalculated estimated serum creatinine from the Modification of Diet in Renal Disease (MDRD) equation. <b><i>Results:</i></b> Following ascertainment of eligibility criteria, 44,370 patients from the Mayo Clinic and 6,112 individuals from the MIMIC III cohort were enrolled. Our model used 6 features from the Mayo Clinic and MIMIC III datasets, including the presence of chronic kidney disease, weight, height, and age. Our model had significantly lower error than the MDRD backcalculation (mean absolute error [MAE] of 0.248 vs. 0.374 in the Mayo Clinic test data; MAE of 0.387 vs. 0.465 in the MIMIC III cohort) and higher correlation (intraclass correlation coefficient [ICC] of 0.559 vs. 0.050 in the Mayo Clinic test data; ICC of 0.357 vs. 0.030 in the MIMIC III cohort). <b><i>Discussion/Conclusion:</i></b> Using machine learning models, baseline serum creatinine could be estimated with higher accuracy than the backcalculated estimated serum creatinine level.

Angiotensin II infusion and markers of organ function in invasively ventilated COVID-19 patients

OBJECTIVE: The use of angiotensin II in invasively ventilated patients with coronavirus disease 2019 (COVID-19) is controversial. Its effect on organ function is unknown. DESIGN: Prospective observational study. SETTING: Intensive care unit (ICU) of a tertiary academic hospital in Milan, Italy. PARTICIPANTS: Adult patients receiving mechanical ventilation due to COVID-19. INTERVENTIONS: Use angiotensin II either as rescue vasopressor agent or as low dose vasopressor support. MAIN OUTCOME MEASURES: Patients treated before angiotensin II was available or treated in an adjacent COVID-19 ICU served as controls. For data analysis, we applied Bayesian modelling as appropriate. We assessed the effects of angiotensin II on organ function. RESULTS: We compared 46 patients receiving angiotensin II therapy with 53 controls. Compared with controls, angiotensin II increased the mean arterial pressure (median difference, 9.05 mmHg; 95% CI, 1.87–16.22; P = 0.013) and the Pao2/Fio2 ratio (median difference, 23.17; 95% CI, 3.46–42.88; P = 0.021), and decreased the odds ratio (OR) of liver dysfunction (OR, 0.32; 95% CI, 0.09–0.94). However, angiotensin II had no effect on lactate, urinary output, serum creatinine, C-reactive protein, platelet count, or thromboembolic complications. In patients with abnormal baseline serum creatinine, Bayesian modelling showed that angiotensin II carried a 95.7% probability of reducing the use of renal replacement therapy (RRT). CONCLUSIONS: In ventilated patients with COVID-19, angiotensin II therapy increased blood pressure and Pao2/Fio2 ratios, decreased the OR of liver dysfunction, and appeared to decrease the risk of RRT use in patients with abnormal baseline serum creatinine. However, all of these findings are hypothesis-generating only. TRIAL REGISTRATION: ClinicalTrials.gov NCT04318366.

MO128RETINOL BINDING PROTEIN (RBP) - NEW BIOMARKER OF KIDNEY INJURY IN MULTIPLE MYELOMA PATIENTS*

Background and Aims The aim of the study was to analyse the utility of retinol binding protein (RBP) in case of renal impairment in MM patients and investigate its relationship with acclaimed parameters of renal failure and markers of MM stages. Method We recruited 73 patients (35 women, 38 men, in age range of 29-90 years, mean 70 ± 10 years) with multiple myeloma (MM), including 6 (8%) with smoldering MM, 40 (55%) with International Staging System (ISS) stage I, 15 (21%) with ISS II and 12 (16%) with ISS III. The majority of patients (65, 89%) received at least one treatment scheme. Thirty patients (41%) received maintenance treatment at recruitment. Median eGFR based on serum creatinine (CKD-EPICr) equaled 67 (range 9 – 117) ml/min/1.73 m2. Results Significant correlation was observed between RBP and the ordered variable describing MM stage from smoldering myeloma to ISS III (R=0.36; p=0.002). There were no differences between patients in CR, PR, SD and PD at the time of samples’ collection. Patients who were on maintenance treatment at recruitment tended to have higher serum RBP (median 42.6 versus 37.7 mg/l), however, the difference was not statistically significant (p=0.068). The patients who received steroid treatment had significantly higher RBP concentrations. There were no such association with other medications. There was no association between RBP and the number of previous treatment lines (p=0.8). Serum RBP did not differ between men and women (p=0.7) and did not correlate with age (p=0.6). Significant correlations were found between RBP and serum creatinine, cystatin C and eGFR values calculated based on creatinine and/or cystatin C (Table 1). In multiple regression, serum creatinine or cystatin C and the treatment with steroids were associated with RBP independently of ISS stage (Table 2). Moreover, RBP correlated with β2-microglobulin, LDH, leukocyte count, α-klotho, FGF-23, GDF-15, uNGAL and uIGFBP-7, however, only the associations with β2-microglobulin and sTfR were independent of serum creatinine in multiple regression (Table 1). Baseline serum RBP concentration was significantly correlated with eGFR after a median of 19 months follow-up (range 1-24 months) (R=-0.35; p=0.003), however, the correlation was not independent of baseline serum creatinine ((beta ± SE: 0.06 ± 0.10; p=0.5). To the contrary, baseline serum cystatin C (beta ± SE: -0.36 ± 0.13; p=0.009) predicted final eGFR independently of baseline serum creatinine. Conclusion RBP may be useful marker in renal damage in patients with chronic kidney injury among patients with MM. This can lead to noninvasive biomarker-targeted diagnostic interventions and contribute to early beginning of treatment that may improve life expectancy quality of life in MM.

Incidence and clinical predictors of infections in patients treated with severe systemic ANCA-associated vasculitis

BackgroundImmunosuppressive therapy has improved the outcome of ANCA-associated vasculitis (AAV), but infectious morbidity and mortality remained high. Recognizing its risk factors seems crucial for prevention, aiming to increase survival of these patients.MethodsWe investigated the incidence and types of infections and assessed predictive factors in 132 patients with severe systemic AAV.ResultsPatients with lower than median incidence of total infections/patient-year during induction had lower baseline serum creatinine, dialysis requirement and Charlson comorbidity index (CCI), compared to those with higher than median incidence (P = 0.037; P = 0.024; P = 0.001; respectively). In subgroups with below and above than median number of severe infections/patient-year during induction, differences were found in baseline creatinine (P = 0.002) and dialysis requirement (P = 0.001); comparing the same cohorts during maintenance immunosuppression, baseline dialysis requirement, diabetes, CCI, and dose of cyclophosphamide (CYC) administered as induction therapy differed significantly (P = 0.019; P = 0.015; P = 0.001; P = 0.015, respectively). Severe infections were predicted by baseline serum creatinine (OR 1.002 [CI 1.001–1.003]) and pulmonary manifestation (OR 2.153 [CI 1.017–4.560]) during induction immunosuppression. In multivariable Cox regression model all-cause mortality was independently predicted by severe infection (HR 1.998 [CI 1.214–3.287]). Among the 168 positive cultures Gram-negative bacteria were responsible for blood stream infections in 33%, and respiratory tract infections in 72%.ConclusionsAdvanced renal failure, pulmonary involvement and high degree of comorbidities increase the risk of infection in AAV. Those who suffer infection during induction immunosuppression have worse long-term survival. Our findings indicate the need for high vigilance for infections and close follow-up of comorbidities when treating AAV.

Incidence of acute kidney injury and its association with mortality in patients with COVID-19: a meta-analysis

Acute kidney injury (AKI) is a complication of COVID-19. However, the incidence of AKI in COVID-19 varies among studies. Thus, we aimed to evaluate the pooled incidence of AKI and its association with mortality in patients with COVID-19 using a meta-analysis. We search Ovid MEDLINE, EMBASE, and the Cochrane Library for eligible publications reporting the clinical characteristics of patients with COVID-19 without language restriction. Incidence of AKI and mortality were reported. Meta-regression was used to describe the association between outcomes. From 26 studies (n=5497), the pooled incidence of AKI in patients with COVID-19 was 8.4% (95% CI 6.0% to 11.7%) with a pooled incidence of renal replacement therapy of 3.6% (95% CI 1.8% to 7.1%). The incidence of AKI was higher in critically ill patients (19.9%) compared with hospitalized patients (7.3%). The pooled estimated odds ratio for mortality from AKI was 13.33 (95% CI 4.05 to 43.91). No potential publication bias was detected. By using meta-regression analyses, the incidence of AKI was positively associated with mortality after adjusted for age and sex (Q=26.18; p=0.02). Moreover, age (p<0.01), diabetes (p=0.02), hypertension (p<0.01) and baseline serum creatinine levels (p=0.04) were positively associated with AKI incidence in adjusted models. In conclusion, AKI is present in 8.3% of overall patients with COVID-19 and in 19.9% of critically ill patients with COVID-19. Presence of AKI is associated with 13-fold increased risk of mortality. Age, diabetes, hypertension, and baseline serum creatinine levels are associated with increased AKI incidence.

ANCA-associated vasculitis with dominant renal involvement: clinical and morphological presentation and outcomes

THE AIM: The analysis of clinical and morphological presentations and outcomes of primary systemic vasculitis associated with anti-neutrophil cytoplasmic antibodies (ANCA-V) with dominant kidney involvement; the determination of clinical and morphological parameters associated with prognosis.PATIENTS AND METHODS. Eighty nine patients with morphologically confirmed ANCA-associated kidney vasculitis on standard immunosuppressive therapy (IST) were included in this retrospective study. Clinical, immunological, and histological indices were analyzed at the time of the kidney biopsy, and early in the short-term (3-6 months) and in the long-term follow-up. The following outcomes were evaluated: the achievement of clinical and immunological remission of the disease; eGFR at the end of follow-up, the progression of renal disease (by the composite point: initiation of renal replacement therapy (RRT) or the estimated glomerular filtration rate (eGFR) <15 ml/min/1.73m2 or decrease in eGFR >50 %); all-cause mortality. The prognostic significance of clinical and morphological parameters was evaluated in multivariable regression models.RESULTS. Most of cases (78 %) were represented by rapidly progressive or acute nephritic syndrome. Mean eGFR was 23 ml/min/1.73 m2. Fifteen percent of patients required acute dialysis. Dominant morphological phenotypes of glomerular lesions were sclerotic (34 %) and mixed (36 %) according to the International Pathology Classification (IPC). Median follow-up was 24 (8; 55) months. Cumulative 5-year and 10-year patient’s survivals and were 92 % and 86 %, respectively. Cumulative 5-year and 10-year renal survivals were 86 % and 68 %, respectively. The cumulative 5-year and 10-year proportions of cases without progression of kidney disease were 80 % and 55 %, respectively. Within 3-6 months of the induction IST 81 % of patients achieved clinical remission (complete (59 %) or partial (22 %)) (CR3-6), while 84 % of patients had immunological remission. Serum creatinine (Pcr) at the time of kidney biopsy was only the factor associated with the risk of renal progression (Expβ=1.73 (95 %CI 1.40-2.14) per 0.1 mmol/l increase). IPC classes and ANCA Renal Risk Score (ARRS) groups as well as other morphological indices of kidney injury had no independent associations with the renal outcomes in Cox models adjusted for Pcr. The independent factors associated with eGFR at the end of follow-up were: CR3-6 (β=0.36±0.08, p<0.001); age (β=-0.34±0.09, p<0.001), Pcr (β=-0.35±0.09, p<0.001) and the global glomerulosclerosis (β=0.28±0.08, p<0.001). CR3-6 (β=0.57±0.10, p<0.001), and the proportion of cellular crescents (β=0.26±0.12, p=0.023) and interstitial inflammation (β=0.27±0.11, p=0.026) were also independently associated with the change of eGFR by the end of follow-up.CONCLUSION. An unfavorable renal prognosis for ANCA-V determined by severe renal dysfunction due to inflammatory and fibrotic alterations of the organ can be significantly improved by adequate therapy with the achievement of higher patient’s and kidney’s survival. The baseline serum creatinine is only the factor associated with the long-term risks of dialysis and kidney disease progression. In addition to baseline serum creatinine and the development of early clinical remission, the separate assessment of global glomerular sclerosis, cellular crescents, and interstitial inflammation may be more useful for the individual prediction of long-term eGFR changes than IPC classes or ARRS.