DOI: https://doi.org/10.21203/rs.3.rs-152669/v1
Background: Sepsis has high incidence and fatality rates in intensive care units, often leading to renal failure. The effectiveness of continuous renal replacement therapy (CRRT) in sepsis-associated acute kidney injury (S-AKI) patients is currently uncertain.
Aim: Joint mode was used to determine the effect of CRRT on the lactate levels and survival of S-AKI patients.
Methods: A retrospective study was applied to patients with sepsis and AKI, which were extracted from the MIMIC-III public database, with the endpoint being 28-day mortality. Every lactate level measurement within 28 days was observed and calculated using logarithms. Joint model combined the longitudinal analysis of the natural logarithm of the lactate level [log(lactate)] in longitudinal submodel and Cox regression by trajectory function, demonstrating the effects of CRRT on 28-day survival and log(lactate) changes, and its final relationship with the event status.
Results: Among the 717 S-AKI patients, 157 received CRRT. CRRT was not associated with 28-day mortality. After adjustments, the relationship between CRRT use and log(lactate) elevation was statistically significant. The parameter estimation of CRRT and log(lactate) indicated that using CRRT will increase log(lactate) by 0.041 in S-AKI patients. The joint model also instigated a fixed association between changes in the lactate level and the event result, revealing an exp value of (1.755) =5.78, indicating that an increase of one unit in log(lactate) will increase the risk of 28-day mortality 5.78-fold.
Conclusion: CRRT does not improve the prognosis of patients with sepsis and acute kidney injury in critical-care units and has a tendency of increasing lactate levels, which is a significant risk factor for the prognosis.
Sepsis is a common clinical critical illness1 that causes dysfunction of the immune and blood coagulation systems, and affected patients usually have insufficient tissue perfusion2. As a product of glycolysis, lactate capacity directly reflects the degree of anaerobic glycolysis in tissues and cells, and is therefore a commonly used clinical indicator that effectively reflects tissue hypoxia and hypoperfusion. It has also been shown to be closely related to the prognosis of sepsis patients3. Several studies have shown that sepsis patients can have increased lactate levels due to low clearance or liver and kidney dysfunction4.
Sepsis is also characterized by persistent refractory hypotension, hyperlactic acidemia, and organ dysfunction after aggressive fluid resuscitation2. Studies have indicated that sepsis can often lead to acute kidney injury (AKI). Up to 50% of sepsis patients develop AKI, leading to poor prognosis and a mortality rate (75%)5,6 that is significantly higher than that in sepsis patients without organ failure7. Continuous renal replacement therapy (CRRT) is a method of extracorporeal blood purification. Current clinical practices consider CRRT to able to maintain internal environment stability by removing toxic substances from the kidneys, and also supports organs function. CRRT has recently been widely used in sepsis treatment7,8. However, the efficacy of CRRT in S-AKI patients remains uncertain. Studies have indicated that lactate levels and the 24-hour lactate clearance rate after CRRT use are related to the 28-day mortality of S-AKI patients9, while CRRT with massive hemofiltration is ineffective for severe lactic acidosis10,11.
A joint model (JM) technique was used in this study to investigate the influence of CRRT on lactate levels and its efficacy on S-AKI patient survival. Applying a JM to longitudinal and event time data has become a valuable follow-up data analysis method that combines the linear mixed model for longitudinal data and the Cox proportional-hazards model for time-to-event data by trajectory function12. A JM provides a more effective method for predicting the impact of treatments on outcomes, a more effective estimation of treatment effects on longitudinal data, and reduces the bias of the overall prediction when compared to the single linear mixed model and the Cox proportional-hazards model12,13.
Data were extracted from the Medical Information Mart for Intensive Care-III (MIMIC-III) database, which is a public database containing data on over 50,000 patients in critical-care units14. SQL was used to extract data before further analysis. The data extracted were the natural logarithm of the lactate level [log(lactate)], CRRT use, ventilator use, vasopressor use, demographic data including age, sex, and race, vital signs including mean blood pressure, respiration rate (RR), heart rate, skin temperature, oxygen saturation, and partial pressure of carbon dioxide, laboratory test results including urine output (UO), pH, albumin, bicarbonate, creatinine (CR), glucose, magnesium (MG), phosphorus (PHOS), potassium, sodium (NA), hematocrit, hemoglobin, platelet, red blood cell width (RDW), white blood cells, neutrophil/lymphocyte ratio (NLR), comorbidities including heart failure, hypertension, liver diseases, fluid-electrolyte imbalance, and SOFA score.
This study selected 1295 S-AKI patients from the MIMIC-III database. The inclusion criteria included meeting Sepsis-3 criteria15, since they have severe infection and organ failure (SOFA score ≥ 2); AKI that occurred during hospitalization (increase in SCr by 0.3 mg/dl [6.5 µmol/l]) within 48 hours; a 1.5-fold increase in SCr from baseline, which is known or presumed to have occurred within the previous 7 days; urine volume < 0.5 ml/kg/hour for 6 hours16. Patients younger than 18 years and without longitudinal records of lactate levels within 28 days were excluded, leaving 717 patients as the study population.
Every lactate level measurement made within 28 days in this study was observed and assessed using logarithms. Linear mixed-effect models were used to analyze longitudinal data. The dependent variable was log(lactate). The independent variables were CRRT use, time of lactate level, and their two-way interaction, NLR, SOFA score group (dichotomized into < 6 and ≥ 6), and age.
Time-to-event data were first analyzed using the Cox proportional-hazards model to determine the relationship between CRRT and 28-day survival. Variables were screened using multivariable regression if they differed significantly between the CRRT and non-CRRT groups.
The JM combined the longitudinal analysis of log(lactate) and Cox regression using a trajectory function, revealing the effect of CRRT on 28-day survival as well as log(lactate) changes and their relationship with the event status (Fig. 1).
All statistical analyses were performed using R software, with the JM being constructed using the “JM” package. Continuous variables were presented as medians with normality-based quartiles, with p values calculated using Student’s t-tests. Categorical variables were presented as numbers and percentages, with p values calculated using chi-square tests.
Among the 717 S-AKI patients, 157 received CRRT. Table 1 lists the baseline characteristics of the two CRRT groups including demographics, vital signs, laboratory-result comorbidities, and SOFA scores. These characteristics demonstrate that CRRT was associated with significant differences in age, race, RDW, NA, MG, PHOS, CR, and UO, liver disease status, ventilator and vasopressor use, SOFA score, and 28-day mortality.
The longitudinal lactate data of 3661 observations were displayed using trajectory functions and plotted using interaction figures. Figure 2 indicates the linear trajectory record of lactate and log(lactate) for each patient over the 28-day analysis period. The figure shows that most observations were concentrated within the first 5 days after patient admission. Lactate levels of patients ranged from 0.5 to 20, with log(lactate) ranging from − 0.4 to 1.3. Cox regression was performed to analyze the relationship between CRRT and 28-day survival, with CRRT being adjusted for by age, race, RDW, NA, MG, PHOS, CR, and UO, liver disease status, ventilator and vasopressor use, and SOFA score. The Schoenfeld residuals test (Fig. 3) was used to determine the independence of residuals and the time to test for the proportional-hazards hypothesis in the Cox model. The results in Fig. 3 revealed that CRRT was not a time-dependent variable and therefore could be analyzed directly using Cox regression. CRRT was also found to not be related to 28-day survival in S-AKI patients after adjusting multiple variables.
Linear mixed-effect models, on the other hand, indicated the association between CRRT and log(lactate). Among the independent variables, CRRT and lactate observation time were statistically correlated with log(lactate) in longitudinal status. The result of the JM demonstrated that after combining the longitudinal submodel of lactate and the survival model of CRRT, CRRT was still not related to the 28-day survival of S-AKI patients. After adjusting the survival submodel, the correlation between CRRT use and elevation of log(lactate) was statistically significant. The parameter estimation of CRRT and log(lactate) indicated that using CRRT increased log(lactate) by 0.041 in S-AKI patients. The JM also instigated a fixed association between lactate level changes and the event result, producing an exp value of (1.755) = 5.78, indicating that an increase of one unit of log(lactate) will increase the risk of 28-day mortality 5.78-fold (Table 2).
Our survival submodel indicates that CRRT use has no effect on the 28-day mortality of S-AKI patients in critical care. S-AKI patient often have weaker glomerular filtration functions, which can increase the probability of adverse symptoms such as electrolyte/acidolysis imbalance and also impair metabolism17. Therefore, preventing further aggravation through eliminating metabolic toxins to prevent further aggravation is necessary. CRRT is currently one of the main treatment methods18. This intervention focuses on replacing kidney function, removing toxic substances such as CR, or improving the fluid-electrolyte imbalance that allows for future treatments since patients are in a stable condition18–20. Although CRRT is still the most common treatment for S-AKI, studies have indicated that renal replacement therapy cannot improve its survival8,21. Whether to apply high-intensity CRRT to S-AKI patients remains controversial. Some studies have indicated that the use of high-dose renal replacement therapy does not improve the overall survival of patients or the recovery of renal function21,22. The efficacy of CRRT on the prognosis of S-AKI can therefore not be confirmed.
The analysis of longitudinal submodel indicated that the use of CRRT is associated with increases in lactate levels within 28 days, while the joint modeling of longitudinal and survival data indicated that lactate level changes were associated with mortality and that log(lactate) is a risk factor for 28-day mortality in S-AKI patients. Many studies have confirmed lactate as a powerful biomarker for sepsis with renal damage and can accurately predict mortality23–25. Reducing lactate is therefore a vital procedure for improving the likelihood of patient survival. However, although CRRT is the most popular treatment in patients with sepsis and AKI, there is still a possibility that it cannot reduce acidosis, resulting in the continued elevation of lactate levels after CRRT is performed26. Potential explanations include how in clinical practice, since it is generally believed that if lactate fluctuates to a lower range, there will be no negative impact on the prognosis of patients. Therefore, these lower lactate levels may not receive adequate attention. Additionally, lactate levels may be affected by the buffer used in CRRT27. For example, when using lactate-based fluids, sepsis patients failed to completely metabolize lactate28, potentially leading to increased lactate levels, eventually developing into metabolic acidosis hyperlactatemia. Finally, when sepsis patients with AKI have more serious infections, vascular permeability is increased and the responses to vasopressors are poor. Although CRRT can temporarily remove toxins and other substances from the body, circulatory ischemia and hypoxia still cannot be improved, and lactate will also continuously increase29. Our research found that across all ranges of lactate levels, the risk of death increased 5.78-fold for every unit increase in log(lactate). When using CRRT to treat patients with sepsis and AKI, we should therefore pay more attention to changes in patient lactate levels.
This study had some limitations. Some factors during the use of CRRT will impact survival, such as filter coagulation, the conversion of patients to intermittent hemodialysis after hemodynamic stability is reached, and the death of patients during treatment. The limited database means we could not analyze these impacts retrospectively. Additionally, the single-center nature of the database reduces the generalizability of our results, which must therefore be tested in further research.
CRRT is not an effective treatment for S-AKI patients in the critical-care unit, and has a tendency to increase lactate levels, which is a significant risk factor for the prognosis. Other treatments that focus on controlling lactate levels should receive more attention in critical care.
Ethics approval and Consent to participate
The study was an analysis of a third-party anonymized publicly available database with pre-existing institutional review board (IRB) approval.
Consent for publication
Not applicable
Availability of supporting data
The data were available on the MIMIC-III website at https://mimic.physionet.org/
Competing interests
The authors declare that they have no competing interests.
Funding
This work was supported by the National Natural Science Foundation of China (No. 82072232; 81871585), the Natural Science Foundation of Guangdong Province (No. 2018A030313058), Technology and Innovation Commission of Guangzhou Science, China (No.201804010308).
Authors’ contributions
LZ created the study protocol, performed the statistical analyses and wrote the first manuscript draft. ZW conceived the study and critically revised the manuscript. FX assisted with the study design and performed data collection. YR assisted with data collection and manuscript editing. DH confirmed the data and assisted with the statistical analyses. CL assisted with study coordination and helped draft the manuscript. JL assisted with manuscript revision and data confirmation. HY contributed to data interpretation and manuscript revision. All authors read and approved the final manuscript.
Acknowledgment
None
Table.1 Baseline characteristics between CRRT Unreceived /Received Group
Characteristics |
CRRT Unreceived |
CRRT Received |
p value |
N |
560 |
157 |
|
Age |
71.0(61.0,81.0) |
67.0(58.0, 77.0) |
0.001* |
Sex n,(%) |
|
|
0.459 |
Male |
347(62.0) |
103(65.6) |
|
Female |
213(38.0) |
54(34.4) |
|
Ethnicity n,(%) |
|
|
0.013* |
White |
408 (72.9) |
109 (69.4) |
|
Black |
85 (15.2) |
16 (10.2) |
|
Others |
67 (12.0) |
32 (20.4) |
|
Vital signs |
|
|
|
Mean blood pressure (mmHg) |
70.3(65.1,75.6) |
68.4(64.1,74.6) |
0.057 |
Heart rate (min-1) |
84.0(73.9,98.2) |
86.9(76.8,97.3) |
0.271 |
Respiratory rate (min-1) |
19.7(17.4,22.9) |
19.9(17.4,22.8) |
0.633 |
Oxygen saturation (%) |
97.5(96.2,98.9) |
97.2(96.0,98.7) |
0.119 |
Skin temperature (℃) |
36.7 (36.2, 37.2) |
36.7 (36.2, 37.3) |
0.906 |
Partial pressure of carbon dioxide (mmHg) |
39.0(33.0,48.0) |
42.0(36.0,48.0) |
0.091 |
Laboratory result |
|
|
|
White Blood Cell (k/uL) |
11.80 (7.40, 17.12) |
11.70 (7.40, 16.80) |
0.850 |
Neutrophil/lymphocyte ratio |
11.86 (5.90, 21.80) |
11.11 (6.80, 21.50) |
0.563 |
Hemoglobin (g/dL) |
10.60 (9.40, 12.10) |
10.70 (9.30, 11.90) |
0.731 |
Red blood Cell Distribution Width (%) |
16.25 (14.80, 18.20) |
16.80 (15.50, 18.40) |
0.023* |
Hematocrit (%) |
32.60 (28.90, 37.00) |
32.60 (28.50, 37.40) |
0.999 |
Platelet (K/uL) |
194.00 (126.25, 272.25) |
196.00 (120.00, 271.00) |
0.972 |
Sodium (mEq/L) |
138.00 (134.00, 141.00) |
137.00 (133.00, 140.00) |
0.017* |
Potassium (mEq/L) |
4.50 (4.00, 5.20) |
4.60 (4.00, 5.30) |
0.320 |
Magnesium (mg/dL) |
2.00 (1.70, 2.30) |
2.00 (1.80, 2.40) |
0.034* |
Bicarbonate(mEq/L) |
22.0(19.0,26.3) |
22.0(18.0,26.0) |
0.526 |
Phosphate(mg/dL) |
3.90 (3.00, 5.00) |
4.60 (3.50, 6.20) |
<0.001* |
Albumin(mg/dL) |
2.9(2.5,3.3) |
2.8(2.4,3.3) |
0.629 |
Glucose(mg/dL) |
128.00 (102.00, 179.25) |
126.00 (98.00, 175.00) |
0.789 |
Creatinine(K/uL) |
2.70 (1.87, 4.20) |
3.70 (2.70, 5.20) |
<0.001* |
PH |
7.4(7.3,7.4) |
7.3(7.3, 7.4) |
0.061 |
Urine output (mL) |
804.0 (258.5, 1459.4) |
195.0(39.0, 602.0) |
<0.001* |
Comorbidities, n (%) |
|
|
|
Congestive heart failure |
310 (55.4) |
91 (58.0) |
0.624 |
Hypertension |
448 (80.0) |
121 (77.1) |
0.490 |
Liver disease |
123 (22.0) |
47 (29.9) |
0.049* |
Fluid electrolyte |
343 (61.3) |
96 (61.1) |
0.999 |
Ventilator use n,(%) |
255 |
119 |
<0.001* |
No |
305(45.5) |
38(75.8) |
|
Yes |
255(54.5) |
119(24.2) |
|
Table.1 Baseline characteristics between CRRT Unreceived /Received Group(continued)
Characteristics |
CRRT Unreceived |
CRRT Received |
p value |
Vasopressor use n,(%) |
374 |
155 |
<0.001* |
No |
186(33.2) |
2(1.3) |
|
Yes |
374(66.8) |
155(98.7) |
|
SOFA |
8.00 (6.00, 10.00) |
11.00 (9.00, 12.00) |
<0.001* |
28-Days Mortality n,(%) |
|
|
0.009 |
No |
307(54.8) |
67(42.7) |
|
Yes |
253(45.2) |
90(57.3) |
|
*P-value is less than 0.05
Table.2 Result of effects between two submodels and the Joint Model
Parameter |
COX |
LME |
JM |
||||||
|
Parameter Estimate |
SE |
P value |
Parameter Estimate |
SE |
P value |
Parameter Estimate |
SE |
P value |
CRRT Treatment Effect on 28-Days Survival |
0.176 |
0.141 |
0.213 |
/ |
/ |
/ |
0.148 |
0.145 |
0.307 |
CRRT Treatment Effect on Log(Lactate) |
/ |
/ |
/ |
0.049 |
0.018 |
0.008 |
0.041 |
0.013 |
0.002 |
Log(Lactate) Effect on 28-Days Survival |
/ |
/ |
/ |
/ |
/ |
/ |
1.755 |
0.625 |
0.005 |
Cox: Cox Proportional Hazards Model; LME: Linear Mixed Effect Model; JM: Joint Model
LME submodel independents includes: CRRT use, time of lactate level, CRRT*Time, NLR, SOFA Group and age.
Cox submodel was adjusted by: Age, ethnicity, RDW, NA, MG, PHOS, CR, UO, liver disease status, ventilator use, vasopressor use, and SOFA score.