Septic shock is a major therapeutic problem in intensive care units (ICUs) and constitutes a severe hemodynamic and immune challenge for the patient. The high incidence and high mortality rate associated to septic shock resulted in an economic burden of nearly $17 billion per year during the 90’s in the United States 1 2 . More recently the average per-case cost for hospitalization due to severe sepsis in ICU was estimated between 29,990 and 36,218$ 3 .

Chronic kidney disease (CKD, defined by an estimated glomerular filtration rate (eGFR) < 60ml/min/1.73m² stable for more than three months) is estimated to affect over 8 million people in the United States and these patients are exposed to increased risks of death, cardiovascular events and hospitalization 4 5 . Moreover, CKD is also an independent risk factor for death due to lung or bloodstream infections 6 7 . Given that (i) CKD exposes the patient to a high risk of death due to cardiovascular and infective causes and (ii) septic shock is a major immune and hemodynamic challenge, we decided to study whether CKD prior to onset of septic shock is an independent risk factor for death.

In this study, non-dialysis CKD and baseline eGFR were found to be independent risk factors for death in septic shock patients. The leading cause of death in CKD patients is cardiovascular disease, but it accounts for only 50% of overall mortality. The causes of the remaining 50% mortality have been less intensively studied. The second leading cause of mortality in end-stage kidney disease patients, after cardiovascular disease, is infectious disease. Previous studies have clearly established CKD as a strong risk factor for septicemia and lung infection 11 12 13 . However, the prognosis of infections in non-dialysis CKD patients has been less well studied, despite the fact that the incidence of infectious disease and the associated mortality rate in non-dialysis CKD patients are known to be higher than in non-CKD populations 14 . Adult patients with bacteremia and serum creatinine on admission >266 μmol/l are at increased risk of death (although creatinine on admission does not exactly reflect baseline GFR) 15 . In our study, serum creatinine on admission was elevated but similar in both survivors and non-survivors. This result suggests that serum creatinine on admission is mainly the result of septic-shock-related acute (but possibly reversible) kidney damage and does not have the same prognostic impact as chronic changes in baseline eGFR. James et al. demonstrated that non-dialysis CKD patients are exposed to a higher risk of lung infection and a higher risk of death from lung infection 7 . They also observed an increased risk of bloodstream infections and death from community-acquired bloodstream infections in non-dialysis CKD patients 6 . However, the study by James et al. dealt with infections in general and did not specifically focus on septic shock.

The mortality rate in our population of septic shock patients (57% after 28 days) was similar to the usually reported values of 40% to 70% 16 17 . Moreover, the SAPSII-predicted hospital mortality of our population was about 70%, which is fairly similar to the observed rate of 65% 18 . In an observational cohort of more than 192,000 patients in the United States, Angus et al. reported an increased mortality rate from severe sepsis in CKD patients (36.7%), relative to the overall study population (28.6%) 1 . However, this study concerned sepsis rather than septic shock in particular and CKD was not clearly defined. Similarly, Alberti et al. reported an increased frequency of CKD in non-survivors from a cohort of 3,608 infected ICU patients 19 . Annane et al. looked for prognostic factors in a population of 8,250 septic shock patients in Europe and North America 16 . As in the study by Alberti et al., CKD was not reported as an independent risk factor for death in septic shock. However, Annane et al. did not clearly defined CKD and did not specify whether chronically dialyzed patients and kidney transplant patients were included in the analyses. The strengths of our study are that it focuses on septic shock (i.e. the most lethal infectious state) with CKD defined according to international guidelines (i.e. as a function of baseline GFR) 20 .

In 14 patients we could not determine the eGFR in the absence of stable serum creatinine before the septic shock and the absence or abnormal eGFR after the episode. A proportion of those patients could be unknown CKD patients or non-CKD patients suffering of chronic kidney failure after recovering from the septic shock. Among those 14 patients, 9 (64%) died (all during the first 28 days after the onset of septic shock). Even if we hypothesized that those 14 patients all belonged to the non-CKD group this would have only slightly modified the 28 day mortality rate (from 50% to 52%) and the difference with the CKD mortality rate would remained significant (p=0.033). It is also usually reported that RIFLE classification (defined as an increase of creatinine above the baseline creatinine value) is associated with mortality in septic shock patients 21 . In our study the degree of AKI (RIFLE classification) was not predictive of mortality. Although RIFLE is not a significant risk factor in the uni and multivariate analyses in our study, we still find a significant difference between survivors and non-survivors for the RIFLE Failure groups and close to significant for the global RIFLE classification (p=0.08) (Table 2). This partial result could be the consequence of an insufficient number of patients. However this result could also result from a reduced increase of creatinine during sepsis as suggested by a study in septic mice by Doi et al. 22 . The authors showed a decreased production of creatinine during sepsis in mice and in our study this could have reduced the elevation of creatinine, altering the RIFLE classification.

The introduction of SAPS II into the multivariate models modified the relation between the different variables and the mortality. Generally speaking, the multivariate backward analyze introduces all the variables into the model and than delete the variable that will improves the most the predictive model by being deleted. This process is repeated until no further improvement is possible. In our study, the age dropped below the significance threshold (Tables 4 and 5) when including SAPS II in the different models meaning that the elimination of the age improved the prediction of mortality. The reason is that SAPS II is a predictive hospital mortality score combining 17 different variables including the age. This explains the stronger predictive value of SAPS II than the age alone in our models.

Also the introduction of SAPS II raised the history of peripheral artery disease (Model 2+SAPS II Table 4) and coronary artery disease (Model 1 and 2+SAPS II Table 5) above the significance threshold. This signifies that those 2 histories bring additional values to predict the mortality that are not shared with SAPS II but were with the age.

The mechanisms linking CKD to increased mortality during sepsis have not been fully elucidated, but hemodynamic and immune causes may be involved. We can hypothesize a possible role of the cardiovascular alterations linked to CKD that could aggravate septic shock. The development of left ventricular hypertrophy, diastolic dysfunction and/or aortic stiffness related to even early stage of CKD is associated with higher cardiovascular morbidity and mortality 4 . In the context of septic hemodynamic stress, these abnormalities could be accentuated, resulting in increased mortality. However, in the present study, CKD appeared to be a risk factor independent of cardiovascular comorbidities. Kidney failure and uremia are associated with severe alterations in the immune system. The two mechanisms required for complete activation of T cells are compromised 23 24 . After activation, T lymphocytes differentiate into Th1 or Th2 lymphocytes that promote cell-mediated and humoral immunity respectively. This differentiation is depressed in uremic patients 25 . Moreover, in CKD, the neutrophil count is not altered but neutrophils are less able to kill microorganisms after phagocytosis 26 .

Our findings in humans are supported by data from two rodent models. Sepsis (induced by cecal ligature-puncture) in mice with pre-existing CKD (created by 5/6th nephrectomy or the injection of folic acid) was associated with a higher mortality rate than in septic non-CKD mice 27 28 . The presence of kidney dysfunction prior to sepsis was associated with increased vascular permeability, bacteremia, elevated vascular endothelial growth factor and serum IL-10 levels and splenocyte apoptosis. More recently, the increased release of high-mobility group box protein 1 (HMGB1) and splenic apoptosis in septic CKD mice has highlighted these pathways in the CKD/sepsis interaction 28 .

Our study is limited by its small sample size preventing analysis of the effect of an eGFR between 30 and 59 ml/min/1.73m² (stage 3 CKD) or between 15 and 29 ml/min/1.73m² (stage 4 CKD) on risk factors in this population, which could only be divided into two groups (eGFR ≥ or < 60ml/Kg/1.73m²). It is also important to note the high prevalence of CKD in our population of patients hospitalized for septic shock in our medical ICU (43% when taking into account the excluded transplanted and chronically hemodialysis patients). As a retrospective single-center study performed in a Nephrology Department’s medical ICU, patient recruitment was probably biased towards CKD.

Non-dialysis CKD appears to be an independent risk factor for death after septic shock in ICU patients. These results emphasize the importance of kidney function before onset of septic shock. Although the mechanisms underlying this elevated mortality rate have yet to be explored, severe alterations of the cardiovascular and/or immune systems may well play a major role.

CKD: Chronic kidney disease; COPD: Chronic obstructive pulmonary disease; eGFR: Estimated glomerular filtration rate; ICU: Intensive care unit; SAPS II: Simplified Acute Physiologic Score II; MV: Mechanical ventilation; Scr: Serum creatinine; Hb: Hemoglobin; ESKD: End-stage kidney disease; IHD: Intermittent hemodialysis; CVVHDF: Continuous venovenous hemodiafiltration; CVVHF: Continuous venovenous hemofiltration

The authors declared that they have no competing interests.

JM and RD ensured data acquisition and wrote the manuscript. JM, RD and ZM performed statistical analysis. BD, ES, EZ were involved in conception, design and data acquisition. JM, CT, ZM, GC and MS were involved in the conception and coordination of the study and corrected the manuscript. All authors have read and approved the final manuscript.