Lung Function and Organ Dysfunctions in 178 Patients Requiring Mechanical Ventilation During The 2009 Influenza A (H1N1) Pandemic

Introduction Most cases of the 2009 influenza A (H1N1) infection are self-limited, but occasionally the disease evolves to a severe condition needing hospitalization. Here we describe the evolution of the respiratory compromise, ventilatory management and laboratory variables of patients with diffuse viral pneumonitis caused by pandemic 2009 influenza A (H1N1) admitted to the ICU. Method This was a multicenter, prospective inception cohort study including adult patients with acute respiratory failure requiring mechanical ventilation (MV) admitted to 20 ICUs in Argentina between June and September of 2009 during the influenza A (H1N1) pandemic. In a standard case-report form, we collected epidemiological characteristics, results of real-time reverse-transcriptase--polymerase-chain-reaction viral diagnostic tests, oxygenation variables, acid-base status, respiratory mechanics, ventilation management and laboratory tests. Variables were recorded on ICU admission and at days 3, 7 and 10. Results During the study period 178 patients with diffuse viral pneumonitis requiring MV were admitted. They were 44 ± 15 years of age, with Acute Physiology And Chronic Health Evaluation II (APACHE II) scores of 18 ± 7, and most frequent comorbidities were obesity (26%), previous respiratory disease (24%) and immunosuppression (16%). Non-invasive ventilation (NIV) was applied in 49 (28%) patients on admission, but 94% were later intubated. Acute respiratory distress syndrome (ARDS) was present throughout the entire ICU stay in the whole group (mean PaO2/FIO2 170 ± 25). Tidal-volumes used were 7.8 to 8.1 ml/kg (ideal body weight), plateau pressures always remained < 30 cmH2O, without differences between survivors and non-survivors; and mean positive end-expiratory pressure (PEEP) levels used were between 8 to 12 cm H2O. Rescue therapies, like recruitment maneuvers (8 to 35%), prone positioning (12 to 24%) and tracheal gas insufflation (3%) were frequently applied. At all time points, pH, platelet count, lactate dehydrogenase assay (LDH) and Sequential Organ Failure Assessment (SOFA) differed significantly between survivors and non-survivors. Lack of recovery of platelet count and persistence of leukocytosis were characteristic of non-survivors. Mortality was high (46%); and length of MV was 10 (6 to 17) days. Conclusions These patients had severe, hypoxemic respiratory failure compatible with ARDS that persisted over time, frequently requiring rescue therapies to support oxygenation. NIV use is not warranted, given its high failure rate. Death and evolution to prolonged mechanical ventilation were common outcomes. Persistence of thrombocytopenia, acidosis and leukocytosis, and high LDH levels found in non-survivors during the course of the disease might be novel prognostic findings.


Introduction
On April 2009, a novel influenza A (H1N1) virus emerged in Mexico and spread rapidly across the world [1,2]. As of 17 June 2010, more than 214 countries had reported confirmed cases of infection with pandemic 2009 influenza A (H1N1) virus, including at least 18,156 deaths [3]. Unlike seasonal influenza, in which hospitalizations occur among patients younger than 2 and older than 65 years, or in those with underlying diseases [4], this novel virus affected otherwise healthy young and middle-aged adults and obese individuals [2,5]. Patients with previous respiratory disease, immunocompromised hosts and pregnant women were affected as frequently as with seasonal influenza [6][7][8][9][10][11][12][13][14][15]. Although a mild form of the disease was prevalent, it soon became evident that the 2009 influenza A (H1N1) virus could also provoke severe, acute respiratory failure requiring admission to the intensive care unit (ICU) for mechanical ventilation [16], which was reflected in the severe pathological injury found at autopsy [17].
The Argentinian population was greatly affected during the pandemic, with a total of 1,390,566 cases of influenza-like illness requiring 14,034 hospitalizations. Of the 11,746 confirmed cases of patients infected with the new strain, 617 died [18]. This represents a death rate per infection of 4.3% in hospitalized cases; an intermediate figure compared to 3.6% in Brazil, 1.2% in Chile, and approximately 6% in Uruguay, Colombia and Venezuela [19]. It should be noted that these numbers reflect great uncertainty, particularly with regard to case diagnosis. Lack of testing of mild disease and difficulties due to laboratory overload have also been well described [15,20]. These general problems have been acknowledged by experts [21].
The severity of disease was rapidly perceived by health authorities and scientific societies. Hence, a committee of experts of the Argentinian Society of Intensive Care Medicine decided to focus on the most acutely ill patients: those presenting with diffuse viral pneumonitis requiring mechanical ventilation. They designed an epidemiological study, recently-published, to determine risk factors and outcomes [15]; this is one of many series up to the present that have described epidemiological and clinical aspects of the 2009 influenza A (H1N1) pandemic [6][7][8][9][10][11][12][13][14][15].
There remains, however, a paucity of data published on physiological evolution during ICU stay [22]. This present study, concurrently planned with the first by the same committee of experts, thus aims to provide such information. Our objectives were: first, to characterize alterations of oxygenation, respiratory mechanics and the use of mechanical ventilation; second, to explore compliance with protective lung ventilation; and, finally, to assess the evolution of laboratory findings and organ dysfunctions throughout the course of the disease.

Materials and methods
This was a multicenter, inception cohort study that included patients aged > 15 years admitted to the ICU with a previous history of influenza-like illness, evolving to acute respiratory failure that required mechanical ventilation during the 2009 winter in the Southern Hemisphere. These patients had confirmed or probable disease caused by the 2009 influenza A (H1N1) virus and were included in the Registry of Cases of the Argentinian Society of Intensive Care Medicine (SATI), created to characterize local aspects of the pandemic. On 27 June 2009, a form to collect online epidemiological data was posted on the official SATI website. A detailed description and analysis of this information was recently published [14].
There was also an optional, more comprehensive casereport form to complete, developed by experts of the SATI's Respiratory Committee for recording certain prespecified variables throughout ICU stay, which included mechanical ventilation (MV), respiratory mechanics, oxygenation, blood chemistry and organ failure variables. This information was collected over 10 days and is analyzed in the present study.
Patients were characterized as confirmed, probable or possible cases of 2009 influenza A (H1N1) [20] according to the findings in the respiratory samples collected on admission. Some specimens, however, were not analyzed because laboratories soon became overloaded, especially at the beginning of the pandemic. As of 25 September 2009, the weekly update of the Ministry of Health reported that in patients ≥5 years with influenzalike illness, the 2009 influenza A (H1N1) virus had displaced other respiratory viruses in 93.4% of the samples processed [23,24]. As a result of this, probable and suspected cases were considered as caused by the novel virus and were so included in the study.
We collected dates of hospital and ICU admission, and of MV onset; demographics; risk factors for influenza A; actual weight; height; severity of illness (Acute Physiology And Chronic Health Evaluation II, APACHE II), organ failures (Sequential Organ Failure Assessment, SOFA); type of MV used, as noninvasive (NIV) and invasive; and date of intubation. Ideal body weight (IBW, ml/kg) and body mass index (BMI) were calculated; obesity was defined as a BMI > 30.
In case of missing observations, local study coordinators were contacted to provide the corresponding values. Proportions were calculated as percentages of existing data.
No assumptions for missing data were made.

Statistical analysis
Statistical analysis was performed with SPSS 17.0 (SPSS Inc., Chicago, IL, USA). Data were analyzed for the entire population; for the subgroups of survivors vs. non-survivors; and for patients receiving NIV on admission vs. those who did not. Descriptive statistics used were: mean ± standard deviations (SD) and median and 25-75% interquartile ranges (IQR) for continuous data of normal and non-normal distribution, respectively; and percentages for categorical data. Differences between subgroups were analyzed with unpaired t test, Mann-Whitney U test, and Chi-square tests, as appropriate. A P-value of <.05 was considered statistically significant. A Kaplan-Meier curve was constructed to evaluate survival over the follow-up period. Over time, normally distributed data were analyzed with two-way repeated measures of ANOVA. At the pre-specified time points, differences within the entire group and subgroups, and between subgroups, were tested using paired and unpaired t tests, respectively.
In non-normally distributed data, differences over time within the entire group and the subgroups were analyzed with Friedman's and Wilcoxon tests. Comparisons between subgroups at the pre-specified time points were tested with Mann-Whitney U test. The Bonferroni correction was used to adjustments for multiple comparisons.
The local Institutional Review Boards waived the need for informed consent, given the general lack of knowledge on the clinical and outcome characteristics of the ongoing pandemic and to the non-interventional study design.

Results
General characteristics (Table 1) Between 6 June and 28 August 2009, the SATI's online Registry included 337 patients admitted to 35 ICUs with confirmed/probable/possible diffuse viral pneumonitis caused by influenza A (H1N1), with acute respiratory failure requiring MV (14). Of these, 178 consecutive patients admitted to 20 ICUs were followed over time, and are presented in this study. To address any potential concern that unconfirmed cases could belong to a different population of patients, we performed a sensitivity analysis of clinical and outcome characteristics data after exclusion of these patients. The results of this analysis did not differ from those of the primary assessments, so the 178 patients are considered for evaluation.
Briefly, patients were middle-aged, with no gender preponderance; they had a history of symptoms of nearly one-week duration and were ventilated at 1 [0 to -2] day after hospital admission. Pre-existent respiratory diseases, obesity, and diseases causing immunosuppression were the most frequent comorbid conditions; and prevalence of pregnancy was higher than in the general population, as expected [25]. Non-survivors were sicker on admission; duration of previous symptoms was longer; and organ failures were more severe. Obesity and immunosuppression were significantly more frequent as predisposing conditions. Ninety-three patients survived (52%) (See Figure 1). (Table 2) During the study period, the entire group had Vt values between 7.8 to 8.1 ml/kg of IBW, with plateau pressures remaining always < 30 cmH 2 O. Non-survivors displayed a trend towards lower Vt and higher plateau pressures, which differed significantly from survivors only at Day 7. Intermediate PEEP levels were used, and decreased in survivors from Day 3 onwards. Driving pressures were similar over time in all patients; only at admission did non-survivors exhibit higher values. PaO 2 /FIO 2 increased significantly over time in all patients and in survivors. It remained, however, < 200 in the whole group throughout the entire ICU stay due to non-survivor values. Non-survivors displayed significantly lower PaO 2 /FIO 2 at all time points.
The main causes of death were refractory hypoxemia (64%); followed by multiorgan dysfunction syndrome (15%) and shock (10%). Prolonged mechanical ventilation and long ICU and hospital stays were frequent (Table 1). Tracheostomy was performed in 29 patients (16%) at Day 14 [11][12][13][14][15][16][17][18][19][20][21]. Acid-base variables and fluid balance (Table 3) Arterial pH increased over time in the whole cohort and in both subgroups, perhaps secondary to general resuscitation measures. Despite this, non-survivors displayed significantly lower pH at all time points, owing to changes in base excess on Days 0 and 3, and to pCO 2 elevations thereafter. Respiratory rates remained unchanged, only increasing at Day 10 in non-survivors; nevertheless, this corresponded to the highest pCO 2 values, indicating the more severe respiratory compromise. Bicarbonate paralleled pH behavior. Changes in fluid balance did not show clear trends: only at Day 10 they decreased significantly, expressing survivors' behavior.

Use of noninvasive ventilation (Tables 4 and 5)
Forty-nine patients (28%) underwent a trial of NIV on admission; they were significantly less ill and had a lower incidence of immunosuppression. Oxygenation and outcome variables were similar to those of patients not receiving NIV.
Sixty-one percent of patients (n = 30) receiving NIV survived; duration of NIV was of 8 (2 to 18) hours. There were no differences between survivors and nonsurvivors in the duration of the procedure, or in the type of interface or respirator used. Of note, most patients on NIV (46 out of 49; 94%) had to be intubated and ventilated invasively for hypoxemic failure. Characteristics associated to NIV success/failure are shown in Table 5.

Blood chemistry and organ failures
The most consistent changes over time were found in platelet count, which increased significantly in the whole cohort (P < 0.000 for Days 3, 7 and 10 vs. Day 0), secondary to elevations in survivors. At all time points, platelets differed between survivors and non-survivors. Conversely, white blood cell count showed a progressive Creatine-kinase and markers of liver injury (alanine/ aspartate aminotransferases, serum bilirubin; not shown) were mildly elevated and displayed no substantial changes. On the contrary, lactate-dehydrogenase levels were significantly higher in non-survivors throughout the study. Creatinine levels were stable over the period, but were significantly higher in non-survivors on Days 0 and 3. Finally, SOFA score diminished over time in all patients (P < 0.000 for Days 7 and 10 vs. Day 0), as a result of the decrease in survivors. SOFA was significantly lower in survivors throughout the study.
In Figure 3, the differences between survivors and non-survivors are displayed.

Discussion
We report on a large, prospective cohort of 2009 influenza A (H1N1) patients that were mechanically ventilated for acute respiratory failure due to diffuse pneumonitis during the pandemic in Argentina. Though most were middle-aged, previously healthy adults, patients with preexistent lung disease, immunosuppression, obesity and pregnancy were also affected. Mortality was high and evolution to chronic critical illness was common, as shown by prolonged mechanical ventilation, high needs of tracheostomy, and lengthened ICU and hospital stays.
Patients had characteristically a history of protracted symptoms and displayed severe compromise of oxygenation compatible with ARDS throughout the study period, which only improved in survivors. At all time points, PaO 2 /FIO 2 differed significantly between survivors and non-survivors, requiring higher FIO 2 and PEEP in this last subgroup. Yet the levels of applied PEEP were only in the intermediate range, similar to mean values of 8.7 cmH 2 O of PEEP in an international study on mechanical ventilation [26], which may explain the relatively high FIO 2 used in our study. Driving pressures were similar in both subgroups most of the time, suggesting an intention to limit alveolar excursion as part of a protective strategy.
It is striking that, as has been described in similar studies on mechanical ventilation performed during the 2009 influenza A (H1N1) pandemic [6,7], tidal volumes used were between 7.5 and 8.3 ml/kg IBW, certainly higher than the 6 ml/kg demonstrated as being lungprotective [27]. Indeed, barriers to implementing lowtidal volume have been identified and might explain physician behavior [28]. Despite this, plateau pressures did remain below 30 cmH 2 O [29], indicating that lung compliance might have been preserved. Perhaps  clinicians focused on plateau pressures rather than on tidal volumes [30] since it still remains unclear which should be limited to avoid ventilator-induced lung injury [31]. We, like others [6,7,32,33], could not find differences in utilized tidal volumes between survivors and non-survivors. Even so, non-survivors tended to display lower values, probably reflecting physician efforts to intensify protective ventilation strategies in the most severely compromised. Some researchers [34,35] have suggested that allowing higher tidal volumes in a population of young and previously healthy patients with strong ventilatory drive might reveal an attempt to restrain heavy sedation and neuromuscular blocker use. Notwithstanding this, we believe that these findings may also represent clinicians' inadequate prescription, as described in other scenarios [36].
Not unexpectedly, VCV was the most common ventilator mode used. PCV use increased throughout the study period, peaking at Day 10. This is in contrast with the recently identified trend towards decreased PCV utilization. Transition to PCV mode was associated with preceding physiological worsening, so clinicians might have perceived PCV utilization as part of a global lungprotective strategy [37].
Refractory hypoxemia was the main cause of death. As in other studies [6,7,11], rescue therapies were frequently applied, with utilization highest 72 hours after admission. Recruitment maneuvers and prone positioning were the primary adjuvants utilized; ECMO and HFOV are currently not available in Argentina. A Table 3 Oxygenation and acid-base variables, and fluid balance in all patients, and in survivors and non-survivors.   prolonged mechanical ventilation course was frequent as reported elsewhere [6]. NIV was the first ventilation approach in 28% of cases, with 94% later requiring invasive ventilation, as has been documented in other studies [6,7,11]. These common experiences should caution against delaying proper ventilatory support in this group, given that rapid deterioration is common. A recent meta-analysis suggests that NIV does not decrease the need for intubation, so evidence to support its use in severe ARDS is questionable [38]. In our study, improved outcomes with NIV could be due to milder disease, evidenced by APACHE II. The small number of patients that were not intubated precludes a statistical analysis; however, they were younger, with less severe disease and better oxygenation.
Significant changes in fluid balance were late and reflected changes in survivors. Negative fluid balances could never be obtained, perhaps suggesting a continuing need for hemodynamic support: 72% of patients presented with shock [14]. On the whole, fluid balances remained between those achieved by "liberal" and "conservative" strategies of the fluids and catheters treatment trial, depending on the day evaluated [39]. Thus far, it is not clear whether the negative fluid balance has a causal role in improving outcome in ALI/ARDS, or if it simply expresses the global recovery of patients.
Another important finding was that arterial pH consistently and significantly differed between survivors and non-survivors, as described elsewhere [40,41]. During the first 72 hours acidosis had a major metabolic component, likely as a sign of hemodynamic impairment. After the first week, respiratory acidosis ensued, indicating either the effects of protective ventilation, or merely deterioration due to progressive shunt, profound ventilation/perfusion mismatch and increased deadspace.
With respect to blood chemistry, the usual findings of thrombocytopenia, leukocytosis and mildly elevated creatine-kinase blood levels were present [21,42]. Regrettably, the lymphocyte count was not recorded. In viral infections, thrombocytopenia occurred frequently. Although the mechanisms by which the 2009 influenza A (H1N1) virus causes thrombocytopenia are unknown, its lack of resolution is a marker of poor prognosis. Both leukocytosis and leucopenia have been found in hospitalized patients with 2009 influenza A (H1N1) [2,43]; in our study, persistent leukocytosis was associated with increased mortality. LDH elevations have been previously described in fatal cases [2], which corresponded to our finding of higher LDH levels in non-survivors at all time points. Such elevations have also been reported in seasonal influenza [44]. In experimental studies, increased LDH is a marker of human fetal membrane cell apoptosis induced by influenza virus [45]. Finally, multiorgan failure was frequent, and predictably more severe in non-survivors.
This study has several strengths: first, the clinical characteristics and time course of pandemic 2009 influenza A (H1N1) are thoroughly described and analyzed. Second, data were collected prospectively in consecutive patients and with a standardized casereporting form, representing a large, nationwide cohort. Third, temporal patterns of mechanical ventilation use, acid-base and blood chemistry variables, as well as fluid balance and organ failures, are carefully analyzed. Prognostic implications are highlighted. Finally, we present the largest experience with NIV use during the pandemic.
Study limitations include the focus on mechanically ventilated patients, excluding less severe cases also admitted to the ICU. Many cases could not be confirmed because laboratories were overwhelmed with clinical samples, which is also described elsewhere [7,14]. Data about transmission to healthcare workers were not recorded, especially regarding NIV. Currently, most information about its use during an epidemic relies upon expert opinion [46].

Conclusions
In 178 patients with diffuse viral pneumonitis caused by the 2009 influenza A (H1N1) virus admitted to the ICU and followed over time, ARDS was the rule, requiring high ventilation support and frequent use of rescue therapies. Death, organ failures, and evolution to prolonged mechanical ventilation were common. In most cases, noninvasive ventilation failed to prevent endotracheal intubation. Finally, elevated LDH levels, lack of recovery of platelet count and persistent acidosis and leukocytosis in non-survivors behaved as prognostic findings.

Key messages
• In 2009 influenza A (H1N1) patients, hospital admission with prompt indication of mechanical ventilation -a marker of severe disease -was associated with a history of symptoms of nearly one-week duration.
• An initial NIV trial was not effective to avoid intubation in most patients; thus, this ventilation approach should likely be discarded in this setting.
• Mortality and morbidity were frequent: death was common and was mainly caused by persistent, refractory hypoxemia. Prolonged mechanical ventilation and ICU and hospital stays were typical.
• pH, platelet count, LDH and SOFA differed significantly between survivors and non-survivors over time. Lack of recovery of platelet count and persistence of leukocytosis might be markers of poor prognosis.
• Every effort should be done to increase adherence to protective ventilation in the real world.