Persistent lymphopenia is a risk factor for ICU-acquired infections and for death in ICU patients with sustained hypotension at admission

This study was approved by our institutional review board (CECIC Clermont-Ferrand—IRB n°5891; Ref: 2007–2016), which waived the need for signed informed consent of the participants, in accordance with French legislation on non-interventional studies. However, the patients and their next of kin were asked whether they were willing to participate in the database, and none declined participation.

Data were prospectively collected daily by senior physicians in the participating ICUs. For each patient, the data were entered into electronic case report forms using VIGIREA^® and RHEA^® data capture software, and all case report forms were then entered into the OutcomeRea^® data warehouse. All codes and definitions were established prior to study initiation. For each patient, age, sex, and McCabe score were recorded. Severity of illness was evaluated on the first ICU day using the Simplified Acute Physiology Score (SAPS II), Sequential Organ Failure Assessment (SOFA) score, and Glasgow Coma Scale (GCS) score, and Knaus’ scale definitions were used to record preexisting chronic organ failures including respiratory, cardiac, hepatic, renal, and immune system failures. Admission category (medical, scheduled surgery, or unscheduled surgery), admission diagnosis (cardiac, respiratory, or neurological failure, infection, and other), invasive procedures (arterial or venous central catheter, Swan-Ganz catheter, or endotracheal intubation), and treatment of organ failures (inotropic support, hemodialysis, and mechanical ventilation) and the use of corticosteroids, gastro-protective drugs, and antibiotics were also recorded. Daily lymphocyte counts were retrospectively collected from four ICUs participating to OUTCOMEREA database between July 2006 and May 2012. All patients with a lymphocyte count in the first day of admission were included in the study. In order to avoid confusion bias, we excluded patients with chronic lymphocytic leukemia (CLL), infection with the human immunodeficiency virus (HIV) or aplasia at admission. We also excluded patients with limitation of life-sustaining therapy in the four first days after admission. Patients with shock or persistent low blood pressure below 90 mmHg of systolic blood pressure in the first day of admission were included. Study variables were the first lymphocyte count on the first day of admission and its evolution at day 3 after admission. The lymphocyte count at admission was categorized in four predefined classes: normal (>1.5 × 10³ cells/µL); subnormal (1 × 10³ cells/µL < lymphocytes ≤1.5 × 10³ cells/µL); low (0.5 × 10³ cells/µL < lymphocytes ≤1 × 10³ cells/µL); very low (≤0.5 × 10³ cells/µL).

The evolution of lymphocyte count at day 3 versus baseline was defined as a binary variable: normal count (≥1.5 × 10³ cells/µL) or relevant increase (more than 0.2 × 10³ cells/µL) and decrease or no relevant increase (≤0.2 × 10³ cells/µL). We handled missing values at day 3 (n = 166, 22.1%) by taking the value one day before or after.

Nosocomial infection was defined as bacteremia, pneumonia, or catheter-related infection occurring after 72 h from admission. Definition of nosocomial infection provided from the HELICS (Hospital in Europe Link for Infection Control through Surveillance) project [16]. Bacteraemia was defined as the presence of pathogenic bacteria in blood culture. Pneumonia was defined as a chest X-ray with suggestive image of pneumonia with clinical and biological signs of pulmonary infection associated with a positive quantitative bacteriological culture from a respiratory sample: a broncho-alveolar lavage [BAL ≥10⁴ colony-forming unit (CFU)/ml]; a protected specimen brush (≥10³ CFU/ml); a blind protected bronchial sampling (≥10³ CFU/ml); a tracheal aspiration (≥10⁵ CFU/ml). Catheter infection was defined as positive quantitative catheter culture (≥10³ CFU/ml) treated by physicians in charge. Only the first event was considered for analysis.

Characteristics of patients were described as count (percent) or median [interquartile range, IQR] for qualitative and quantitative variables, respectively, and were compared between groups using Chi-square or Mann–Whitney tests, as appropriate.

In order to decrease the risk of confusion bias between lymphopenia and acquired-ICU infection, we developed a propensity score aimed to predict the probability to have a nosocomial infection conditionally on variables recorded in the first 2 days of admission [17].

A logistic regression was used to construct the propensity score including variables on clinical relevance or statistic comparison on univariate analysis. Linearity of the logit of continuous covariates was checked. The following clinically relevant variables were entered in the model: age, gender, admission category, center, Knaus definitions, McCabe score, main reason for ICU admission (multi-organ failure, cardiogenic shock, septic shock, coma, acute respiratory deficiency), diabetes with complications (binary variable), severity illness related to specific organ assessed by the sequential SOFA score categorized in 2 classes, lower or equal to two or higher (cardiovascular, neurological, hepatic, renal, coagulation failures), acute respiratory distress syndrome, mechanical ventilation, central venous catheter, arterial catheter or arterial pulmonary catheter, temperature, use of gastro-protective drugs, antibiotics, or corticosteroids.

Then, an inverse probability of treatment weighted (IPTW) [18] based on the propensity score was computed to create a pseudo-population in which the probability to develop or not an ICU-acquired infection was equal. We performed a model with covariates using for the construction of the propensity score weighted by the IPTW including the explicative variables, baseline lymphocyte count, and evolution at the third day [19]. We took the 5–95th percentiles of IPTW to create a new pseudo-population to assess the robustness of the model.

Sub-distribution hazard ratios (sHRs) were developed to assess the independent effects of lymphocyte count at admission and the evolution at day 3 on subsequent risk of ICU-acquired infection. Discharge alive from ICU was treated as competing events. Data were censored at 28 days since the fourth day after admission.

For the secondary objective, risk of death related to initial lymphocyte count and its evolution at day 3, the same protocol was used. We developed a specific propensity score aiming to predict the probability to die in ICU within 28 days of inclusion conditionally on variables recorded within the first 2 days of admission. The following clinically relevant variables were entered in the model: age, gender, admission category, center, Knaus definitions, cardiogenic shock as symptom at admission, continuous monitoring as reason of admission, complicated diabetes, severity illness related to specific organ assessed by the SOFA score categorized in two classes, lower, or equal to 2 or higher (cardiovascular, neurological, hepatic, renal, coagulation failure), respiratory failure severity reflected by acute respiratory distress syndrome, requiring invasive mechanical ventilation, central venous catheter, arterial catheter or arterial pulmonary catheter, temperature, use of corticosteroids.

Sub-distribution hazard ratios (sHRs) were developed with covariates using for the construction of the propensity score weighted by the IPTW. Discharge alive from ICU was treated as competing events. For all analyses, p < .05 was considered to statistically significant. All analyses were performed using SAS, version 9.3 (SAS Institute, Cary, NC, USA).

Of the 2402 patients recorded within the 4 participating ICUs (Fig. 1), 753 patients were included. The mean age was 68 [56; 78] years, 467 patients (62%) were males, and the median SOFA score at admission was 8 [5, 11]. Medical admission represented the most frequent cases [596 patients (79%)], and septic shock was the first diagnosis at admission in 154 patients (21%). Mechanical ventilation was required for 559 patients (74%) and vasoactive agents at day 1 or 2 for 480 patients (63.8%). The median length of stay in ICU was 9 days [6‐18]. A total of 174 (23%) patients had ICU-acquired infection and 161 (21%) patients died in ICU during the study period (Table 1).

The median number of lymphocyte counts measurements was 6 [4‐13]. The percentage of day with a lymphocyte count by patient during ICU stay was 75%, and the median range between two blood samples with lymphocyte count was 1 day. The median of the lymphocyte count at admission was 0.80 [0.51–1.29] × 10³ cells/µL. The distribution in 4 classes was as follows: 149 patients (20%) had a normal lymphocyte count with a median of 1.97 [1.70–2.80] × 10³ cells/µL; 141 patients (19%) had a lymphocyte count ranging between 1 and 1.5 × 10³ cells/µL with a median of 1.19 [1.10–1.30] × 10³ cells/µL; 278 patients (37%) had a lymphocyte count ranging between 0.5 and 1 × 10³ cells/µL with a median of 0.72 [0.61–0.84] × 10³ cells/µL; 185 patients (24%) had a lymphocyte count lower than 0.5 × 10³ cells/µL with a median of 0.34 [0.24–0.43] × 10³ cells/µL.

Among the total of 174 (24%) ICU-acquired infections, pneumonia was diagnosed in 113 (64.9%) patients, bacteremia in 37 (21.3%) and catheter-associated infection in 36 (20.7%). In 13 patients, 2 sites of infection were diagnosed the same day. Enterobacteriaceae bacteria were the most frequent pathogens isolated, followed by Pseudomonas spp. and Staphylococcus aureus (Tables 2, 3).

There were no relationships between the lymphocyte count and the SOFA score, and between the delta of the SOFA score and the variations in the lymphocyte counts. This result is consistent with our results about the independent role of immune paralysis and organ failures.

Comparisons between patients with ICU-acquired infection and the others are shown in Table 1. The final logistic model used to calculate propensity score is given in Additional file 1: Table E1. The cumulative incidence curve of ICU-acquired infection is shown in Fig. 2a.

Sub-distribution hazard ratios (sHRs) of ICU-acquired infection were significant for abnormal values at admission (Table 4), with no difference between subnormal and very low lymphocyte counts. The absence of relevant increase in the lymphocyte count at day 3 was associated with an increased risk of developing an infection (sHR of 1.37 [1.12–1.67], p = 0.002) (Fig. 2b). The interaction term between baseline lymphocyte count and lymphocyte increase at day 3 was not significant. Importantly, the onset of ICU-acquired infection was associated with an increased day-28 mortality (p < 0.001).

Comparisons between patients’ dead in ICU and others are shown in Table 1, using the final logistic model used to calculate propensity score (Additional file 1: Table E2). The incidences of 28-day mortality according to baseline lymphocyte count and its evolution at day 3 are shown in Table 5. The baseline count of lymphocyte had no impact on the 28-day mortality in ICU. However, the decrease or the non-significant increase on day 3 was significantly associated with the death in ICU [sHR of 1.67 [1.37–2.03], p < 0.0001 (Table 5)]. The cumulative incidence curve of death according to the evolution of lymphocyte count is represented in Fig. 2c.

A large cohort of ICU patients with shock at admission, we demonstrated the independent impact of a low baseline lymphocyte count and its non-relevant increase at day 3 with the risk of ICU-acquired infection and, for persistent lymphopenia, its impact on 28-day mortality. Total lymphocyte count appears as a simple and routine marker of immune dysfunction, and might be useful for selecting patients that could benefit of potential immune-adjuvant therapies [27].