Central venous pressure and acute kidney injury in critically ill patients with multiple comorbidities: a large retrospective cohort study

We conducted a large-scale, single-center, retrospective cohort study using data collected from the MIMIC-III open source clinical database (version 1.4), which was developed and maintained by the Massachusetts Institute of Technology, Philips Healthcare, and Beth Israel Deaconess Medical Center [12]. One author (Qi Guo) obtained access to the database and was responsible for data extraction (certification number: 25233333). Information derived from the 61,532 electronic medical records of critically ill patients admitted to intensive care units (ICUs) between 2001 and 2012 was included in this free, accessible database. The database was approved for research use by the Institutional Review Boards of the Massachusetts Institute of Technology and Beth Israel Deaconess Medical Center, and studies using the database were granted a waiver of informed consent.

All patients in the database were screened according to the following inclusion criteria for this study: (1) adults (≥18 years of age at ICU admission); (2) ICU stay ≥1 day; and (3) for patients with multiple ICU stays, only the data for the first stay were considered. Patients with censored age, no CVP records, or no creatinine records were excluded.

CVP, creatinine, and urine output records during the ICU stay were extracted. Other day 1 ICU measurement records were also extracted, including age, sex, weight, blood pressure and admission illness scores (the Simplified Acute Physiology Score (SAPS) [13] and the Sequential Organ Failure Assessment (SOFA) score) [14]. Moreover, data on the use of vasopressors, inotropes, sedatives, diuretics, invasive mechanical ventilation, and comorbidities, including sepsis, congestive heart failure (CHF), arrhythmias, hypertension, diabetes, chronic renal failure and cancer, were extracted from the database. In this study, vasopressors included norepinephrine, epinephrine, phenylephrine, vasopressin, and dopamine. Inotropes included dobutamine and milrinone. Chronic renal failure was defined as abnormalities of kidney structure or function (estimated glomerular filtration rate (eGFR) < 60 ml/min/1.73m²), present for more than 3 months, with implications for health [15]. Cardiac surgery recovery unit (CSRU) was used as variable to indicate the ICU type one patient stayed was CSRU. The comorbidities were determined from the International Classification of Disease, 9th Edition, Clinical Modification codes.

The primary exposure was the mean CVP during the first 7 days after ICU admission. We divided the mean CVP into four levels according to interquartile range as follows: quartile 1, CVP ≤ 8.29 mmHg; quartile 2, 8.29 < CVP ≤ 10.64 mmHg; quartile 3, 10.64 < CVP ≤ 13.20 mmHg; and quartile 4, CVP > 13.20 mmHg.

The primary outcome was the odds of 2-day and 7-day AKI after ICU admission. We defined AKI by serum creatinine based on the KDIGO criteria [16]. AKI was categorized as Stage 1 if there was a 1.5–1.9 times serum creatinine increase from baseline, a 0.3 mg/dL serum creatinine increase or a urine output < 0.5 ml/kg/h for 6–12 h. Stage 2 was when there was a 2.0–2.9 times serum creatinine increase from baseline or a urine output < 0.5 ml/kg/h for ≥12 h, and Stage 3 was when there was a ≥ 3 times serum creatinine increase from baseline or a ≥ 4.0 mg/dL serum creatinine increase or urine output < 0.3 ml/kg/h for ≥24 h. The first serum creatinine record measured on ICU Day 1 was considered the “baseline”.

We calculated the CVP fluctuation within the first 2 days as follows: (mean CVP on the second day – mean CVP on the first day)/mean CVP on the first day. Patients were divided into 3 CVP trend groups: decreasing trend (fluctuation ≤ − 10%), increasing trend (fluctuation ≥10%), and stable trend (− 10% < fluctuation< 10%). Among the study population, 7397 patients suffered continuous CVP monitoring within the first 2 days. The association between this CVP trend group and AKI outcome was then evaluated for these patients.

Adjusted variables included age, male sex, weight, CSRU, ventilation use, vasopressor use, inotropes use, sedative use, diuretic use, SAPS score, SOFA score, sepsis, CHF, arrhythmias, hypertension, diabetes, chronic renal failure, cancer, systolic blood pressure (SBP), and diastolic blood pressure.

The association between CVP and AKI was further evaluated in patients using mechanical ventilation. To evaluate whether ventilation parameter influence our results, models were adjusted for age, male sex, weight, CSRU, ventilation use, vasopressor use, inotropes use, sedative use, diuretic use, SAPS score, SOFA score, sepsis, CHF, arrhythmias, hypertension, diabetes, chronic renal failure, cancer, SBP, diastolic blood pressure, and positive end-expiratory pressure.

Normally distributed continuous variables are presented as the mean ± standard deviation, whereas nonnormally distributed data are presented as the median (interquartile range). Categorical variables are presented as numbers (percentages). Baseline characteristics were stratified by quartiles of mean CVP during the first 7 days after ICU admission. Baseline data were compared using the analysis of variance test or rank-sum test, as appropriate, for continuous variables, and the chi-square test was used for categorical variables. We performed linear and logistic regression to compute odds ratios (ORs) for the association of mean CVP with the odds of AKI.

Among the 61,532 ICU admissions in the MIMIC-III v1.4 database, 11,135 patients were enrolled in our study based on the inclusion and exclusion criteria in the Methods section (Fig. 1). The number of patients in each CVP quartile was approximately 2780. The mean CVP in each quartile was 6.7 ± 1.3 mmHg, 9.5 ± 0.6 mmHg, 11.9 ± 0.7 mmHg, and 16.0 ± 3.2 mmHg in the lowest to highest quartiles, respectively. Interestingly, patients in higher quartile of mean CVP had lower urine output, higher creatinine and lower eGFR. Also, patients in the highest quartile of mean CVP presented the greatest weight and highest SAPS and SOFA scores. Furthermore, patients with the highest quartile of mean CVP were likely to be treated with vasopressors and diuretics and more likely to have comorbidities of sepsis, CHF, arrhythmia, hypertension and chronic renal failure (Table 1).

During the first 2 days and 7 days, 8544 and 9289 patients had AKI, respectively. The incidence of AKI at 2 days was greater among patients with higher CVP, ranging from 63.6% in patients with a mean CVP ≤ 8.39 mmHg (Quartile 1) to 88.4% in patients with a mean CVP > 13.20 mmHg (Quartile 4). A similar trend was detected between the incidence of AKI at 7 days and the mean CVP. Moreover, patients in the higher quartiles of mean CVP presented greater KDIGO AKI severity stages at 2 days and 7 days (Table 2). Additionally, linear regression showed that the CVP quartile was positively correlated with the incidence of AKI at 2 days (R² = 0.991, P = 0.004) and 7 days (R² = 0.990, P = 0.005) (Fig. 2).

We further performed a logistic regression analysis to determine the association between quartiles of mean CVP and AKI outcomes in 7 days. For the crude model, patients in higher quartiles of mean CVP had a greater incidence of AKI at 7 days than those in the lowest quartile of mean CVP, ranging from OR = 1.63 (95% CI: 1.44–1.85) to OR = 5.18 (95% CI: 4.37–6.14). After adjustment, the mean CVP quartile remained a significant predictor of AKI at 7 days (Fig. 3). Furthermore, the odds of AKI were 1.18-fold (95% CI: 1.16–1.20) higher per 1 mmHg increase in mean CVP. After adjusting for demographics, treatments and comorbidities, the odds of AKI was 1.10 (95% CI: 1.08–1.12).

Next, we determined the association between mean CVP and AKI in subgroups of patients with an older age, low SBP and a history of cardiac surgery, patients treated with vasopressors, diuretics and ventilation, and patients with CHF and sepsis as comorbidities. Our data showed that CVP in all 11,135 subjects was positively correlated with AKI (Fig. 4a). Additionally, among 6079 elderly patients (age ≥ 65 years), patients with a mean CVP of greater than 11 mmHg had higher odds of AKI than those with a mean CVP of 5 mmHg (Fig. 4b). Moreover, 5008 subjects with SBP ≤ 110 mmHg and higher mean CVPs had higher odds of AKI (Fig. 4c). Additionally, a higher mean CVP (approximately CVP > 10 mmHg for diuretics and CVP > 8 mmHg for vasopressors) suggested an increased incidence of AKI in patients with use of diuretics, sepsis, use of ventilation, use of vasopressors (Fig. 4d & e & f & g). However, in subjects with use of inotropes, with no use of vasopressors or inotropes, with CHF, the mean CVP from 5 mmHg to 10 mmHg did not show significant correlation with AKI (Fig. 4h & i & j). Nevertheless, an elevated mean CVP was positively correlated with the incidence of AKI for patients in the CSRU (Fig. 4k).

To address the effect of the ventilation properties on the results, subgroup analyses in patients using invasive mechanical ventilation were performed. After adjustment for a series of variables, the CVP quartile 4 group showed a significantly higher risk of AKI than the CVP quartile 1 group (OR: 3.02, 95% CI: 2.41–3.79). The odds of AKI were 1.11 (95% CI: 1.09–1.14) times higher per 1 mmHg increase in mean CVP in this ventilation subgroup (Fig. 5).

Given that CVP was a dynamic parameter, we also investigated the association between CVP trend and AKI. In the crude model, both the decreasing trend group and increasing trend group showed significantly lower AKI risk than the stable trend group (P < 0.001). These associations were diminished after further adjustment (P > 0.05) (Table 3).