effects of HYPERCAPNIA in COVID–19 ARDS

Writing committee

Anissa M Tsonas, A. Serpa Neto, F. Paulus, M.J. Schultz for the PRoVENT–COVID Collaborative group

STATISTICAL ANALYSIS PLAN, May 26, 2020

Inclusion criteria

Because of this analysis’ focus on early PaCO₂ management, we will exclude patients who started invasive ventilation in a non-participating hospital prior to inclusion. Also, patients who are missing ventilation data due to transfer to a non–participating hospital before calendar day 4, will be excluded from this analysis.

Patient categorization

For patient categorization, eight–hourly PaCO₂ measurements collected over the first four calendar days of ventilation will be used to classify patients as ‘hypercapnic’ if the majority of available PaCO₂measurements is above 45 mmHg, or ‘not hypercapnic’ if the majority is equal to, or below 45 mmHg. We will ignore the first PaCO₂ measurement in each patient. To determine heterogeneity of hypercapnia for each patient, this categorization will also be determined for each day of ventilation and presented in a Sankey Diagram. Day 1 will be defined by the measurements collected on calendar day 0 and calendar day 1 for all analyses.

Reporting

The following ventilatory variables and parameters will be shown in cumulative distribution plots and line graphs: tidal volume (V_T), respiratory rate (RR), ventilator ratio (VR), minute ventilation (MV), driving pressure (ΔP) and mechanical power (MP). The trend over time of ventilatory variables will be assessed with mixed–effect linear models with center and patients treated as random effect to account for clustering and repeated measurements and with PaCO₂ categorization, time (as a continuous variable), and an interaction of PaCO₂and time as fixed effect. Overall P values from this analysis represent the overall difference among groups over time and P values from interaction represent a statistical assessment of whether the trend over time differed among the groups. All daily measurements of ventilatory variables (three times a day) were aggregated as the mean per day.

Continuous variables will be presented as medians (quartile 25% – quartile 75%) and categorical variables as number and percentages. The two groups based on PaCO2 measurements will be compared using Student t-test, or Wilcoxon rank–sum test and Chi-square test, or Fisher exact test for continuous and discrete variables, respectively. ICU and hospital length of stay, and 28-, and 90-day mortality will be shown in Kaplan-Meier curves and groups are compared through a Log-Rank test.

Impact of PaCO₂ category on outcome

The impact of PaCO₂ category on 28–day mortality, the primary outcome, will be further assessed in a generalized mixed-effect model with center of admission as random effect to account for clustering. The following variables with a known association with ICU mortality with be included in the model: 1) ventilatory variables in the first day aggregated as the median from a maximum of four assessment (PEEP, PaO2/FiO2, V_T adjusted by predicted body weight, MP, VR, and RR); 2) use of assisted ventilation; 3) laboratory tests and vital signs in the first day aggregated as the median from a maximum of four assessment (arterial pH, creatinine, heart rate and mean arterial pressure); 4) organ support at the first day (use of vasopressor and cumulative fluid balance); and 5) demographic characteristics (age, gender, body mass index, hypertension, heart failure, diabetes, chronic kidney disease, chronic obstructive pulmonary disease, use of angiotensin converting enzyme inhibitor and use of angiotensin II receptor blocker). Peak pressure and driving pressure are not considered due to collinearity with respiratory compliance (which was judged to be more clinically relevant in the model), and FiO2 is excluded due to its association with PaO2/FiO2, used to define the outcome.

Analysis

All analyses were conducted in R v.3.6.3 (R Foundation, Vienna, Austria) and significance level was set at 0.05.