The Luca ventilator is a rapidly developed COVID-19 specific mechanical ventilator, designed and manufactured in a closed source process by Femtonics Kft. (Budapest, Hungary) in collaboration with the Department of Anaesthesiology and Intensive Therapy at Semmelweis University (Budapest, Hungary).

The electronically controlled pneumatic ventilator is equipped with an electronic display and manual interface (Fig. 1) and is able to operate with commonly used disposable, double limb ventilator circuits. Its software permits controlled, assist/controlled, and assisted pressure-controlled (e.g. pressure support) modes allowing spontaneously triggered breaths, a crucial factor in achieving patient-ventilator synchrony.15 During the design of the device, the main goal of development was to incorporate materials and parts that were readily available locally in order to ensure a continuous supply chain for possible increased production need. Main performance features corresponding to currently accepted requirements are presented in Supplementary Table 1. As can be seen, the design and features of the LUCA mechanical ventilator permits synchronization of mandatory breaths with the patient's respiratory effort as well as advanced patient tailored adjustments of ventilator parameters.

Fig. 1.

Luca mechanical ventilator with close-up of the user interface (online version: colour; print version: black and white).

(0.26MB).

Three prototypes of the Luca ventilator were used for the testing process with Intersurgical smoothbore breathing system 1.6m (Intersurgical, Wakingham, Berkshire, UK) circuits, Virobac II viral/bacterial filters (King Systems Noblesville, Indiana, USA) on both the expiratory and inspiratory limbs additionally to a Humid-Vent Compact S (Teleflex Medical, Athlone, Ireland) heat and moisture exchange filter (HMEF) at the distal end.

The testing process was designed with sequential stages, including an in vitro preclinical part, followed by a safety testing stage, a clinician usability test and concluding with in vivo, clinical testing on respiratory failure patients. All subsequent steps were dependent on the previous successful completion of the previous testing stage. Testing was performed in May 2020. All preclinical tests were performed during a 2-day period, except for endurance testing. Clinical tests were performed over a course of one week.

Accuracy and reliability were tested according to the RMVS performance testing specifications12 with additional endurance testing. Actual values of pressure and volume supplied by the Luca ventilators were measured with an ASL 5000 breathing simulator (IngMar Medical, Pittsburgh, PA) and compared to set/displayed parameters. The following settings were tested in a total of 72 combinations: compliance (10, 20 and 50mL/cmH2O), resistance (10, 20 and 50 cmH2O), inspiratory pressure (15 and 30cmH2O), respiratory frequency (12 and 20/min), inspiratory and expiratory ratio (I:E) (1:2), FiO2 (55 and 95%), and PEEP (5, 10 and 15cmH2O). As per the RMVS, accuracy was deemed adequate if the measured airway pressure was within±(2+(4% of set parameter))cmH2O; interface readings of expired volumes greater than 50mL were within±(4.0+(15% of the actual measured volume expired through the patient-connection port)) mL; and measured oxygen concentrations were ±5% of the set value. Inspiratory time (time from the onset of the flow towards the ASL until flow reversal was detected), expiratory time (time from the onset of the flow away from the ASL until flow reversal was detected), respiratory rate and successful triggering rate were measured by the ASL and compared to displayed values and deemed adequate within a 5% deviation. FiO2 readings were verified with a custom oxygen sensor.

Endurance was tested by continuous operation of a Luca prototype for 30 days in a hospital environment connected to an EasyLung (IMTMedical, Malaysia) test lung with a compliance of 25mL/mbar and resistance of 20mbar/L/s under continuous supervision.

The safety testing stage involved a list of simulated alarm scenarios (high/low pressure, high/low respiratory volume, low/high minute volume, high respiratory rate, FiO2 inaccuracy, gas supply inadequacy, circuit disconnect), based on previously published consensus on requirements and alarm profiles of commonly used ICU ventilators.16,17 Emergency situations were simulated using the ASL 5000 breathing simulator in order to test the following alarm scenarios: high/low pressure, high/low respiratory volume, low/high minute volume, high respiratory rate, FiO2 inaccuracy, gas supply inadequacy, circuit disconnect. Activation and timing of alarms were recorded. Closed Suctioning Test was performed according to RMVS guidance.12

Usability was tested with the participation of 10 critical care doctors who each performed a total of 13 tasks (see Supplementary Table 2) on a ventilation model consisting of a Luca ventilator prototype connected to the ASL after a 5-min introduction to the ventilator. Ventilation model settings were as follows: 5cmH2O resistance, 70mL/cH2O compliance within an initially passive system [0/min respiratory rate], then one with a spontaneous breathing with 18/min respiratory rate and 8cmH2O breathing effort. Baseline ventilator settings for the LUCA ventilator prototype were 5cmH2O PEEP, 15cmH2O inspiratory pressure, 15/min respiratory rate, 1.0s inspiratory time with the pressure trigger inactivated.

System Usability Scale (SUS), an evaluation comprising of 10 questions, was used to assess subjective usability.18 The SUS score has a range of 0–100, where higher scores indicate better usability,18 and scores correlate well with a standard adjective scale (worst imaginable, awful, poor, OK, good, excellent, best imaginable).19 SUS score for the LUCA mechanical ventilator was calculated based on the evaluation of the participating critical care specialists after the usability testing scenario.

The clinical testing was a comparative, observational study, during which respiratory failure patients were ventilated by standard commercially available ventilators (Hamilton G5, Hamilton Medical, Bonaduz, Switzerland) for 60min, followed by 60min of ventilation by a Luca ventilator prototype with identical parameter settings. Assist/control pressure control ventilation mode as per the Chatburn classification (patient pressure or ventilator time triggered, pressure controlled, time cycled) mode was used in both the Luca and standard ICU ventilators. Vital parameters (heart rate, blood pressure, oxygen saturation) and ventilation parameters (respiratory rate, tidal volume, minute ventilation, pressure, FiO2) were recorded every 10min additionally to two blood gas analyses during both 60-min periods. Signs of dyspnoea (increased work of breathing) were assessed by the study physicians or if in case the patient was able to cooperate, by the Borg dyspnoea scale.20 Patients were asked to rate the Luca ventilator as “similar or more comfortable compared to the standard ICU ventilator” or “less comfortable compared to the standard ICU ventilator”. Need for physician and nursing intervention as well as type and frequency of ventilator alarms were recorded. Rescue intervention (switching to the original ventilator with adequate clinical management of adverse events) was allowed in case testing physicians deemed it necessary.

Patients treated for respiratory failure were recruited for the observational study. Inclusion criteria were: >18 years of age, ongoing mechanical ventilation with pressure controlled, time-cycled mode. Exclusion criteria involved: pregnancy, unstable condition, lack of informed written consent and instability associated with rapidly changing respiratory patterns during the planed observation period. Primary endpoint was blood gas parameter values, secondary endpoints included clinical parameters recorded during the sessions.

Gas flowrate, volume, and leakage are expressed as standard temperature pressure dry (STPD) values for the preclinical stages and as body temperature pressure saturated (BTPS) values for the clinical stage of the study.

Data are presented as mean (±SD) and percentage (N). Groups were compared with Mann–Whitney U test, correlation was calculated with Spearman correlation. Generalized Linear Mixed Modelling (GLMM) with log normal distribution and a random effect by patient was used to assess the ventilator type controlling for ventilation time and the interaction between ventilator type and time on blood gas values and clinical parameters during clinical testing. Clinical relevance was evaluated by effect size according to Cohen.21 Cohen's D (the difference between two groups’ means divided by the pooled standard deviation) is considered small if the value is less than 0.2, medium if less than 0.5, and large if more than 0.8. A small Cohen's D values suggests that the difference is negligible, even if it is statistically significant. As Cohen's D might be effected when analysing variables of larger values, clinical relevance was confirmed with Cohen's f2 for each variable. Cohen's f2 (the proportion of variance accounted for by the variable of interest given the variance accounted for by all other variables in the model) is considered small if the value is less than 0.02, medium if less than 0.15, and large if more than 0.35. A small Cohen's f2 value suggests that the difference is negligible, even if it is statistically significant. Given that the data contains repeated effects, Cohen's D is often considered to be an inappropriate effect size due to the correlation within the patient. However, Cohen's D is often easier to interpret. Therefore, both effect sizes are included.

The number of patients required to prove clinical equivalence in the clinical testing stage was evaluated with sample size calculation using the GLMM approach with RMVS criteria (4+15% and 2+4% respectively), 10% and 5% difference for inspiratory pressure and tidal volume and 10% and 5% difference for pO2 and pCO2 values respectively, based on previously published data focused on clinical variability of blood gas values in stable ICU patients.22

Sample size calculation was conducted for difference in ventilator type with a Generalized Linear Mixed Model approach. The following assumptions were utilized. (1) There will be no clustering effect as all patients were treated at the same hospital. (2) Power was assumed to be 0.8 and Type 1 error was set to 0.05. (3) Unstructured correlation structure, for both time and ventilator Type. (4) Standard deviation was considered unchanged by time and ventilator type as found to be the case in the data. (5) Inspiratory pressure and tidal volume were used as the outcome variables of interest for the analysis.

Significance level was set at p<0.05.

Statistical analysis was performed by IBM SPSS Statistics version 25.0 (IBM, Armonk, NY)) and SAS version 9.4 (SAS Institute, Cary, NC).

This study was approved by the Hungarian Ethical Committee and the National Institute of Pharmacy and Nutrition (OGYÉI/25837/2020) and conducted in compliance with the Helsinki Declaration. All participating physicians and patients provided written consent before participating in the study. If a patient was unable to give written consent, consent was obtained from the next of kin and was reinforced once the patient was able to consent. All data were analyzed anonymously.

The 72 predefined settings were evaluated on all three prototypes of the Luca ventilator. As can be seen from Fig. 2, all measured parameter values were within the accepted range for all prototypes.

Fig. 2.

Set and measured pressure values and displayed and measured volume values of the Luca ventilator tested against lung models with different compliance and resistance specified in RMVS. Scatter plot of measured pressure readings for two set pressure values and measured and displayed volume readings charted using different lung models (as described in Methods). Different grey scale points represent the three different ventilator prototypes. Shaded zone represents accepted range as per RMVS specifications.

(0.15MB).

Measured and displayed inspiratory times differed by more than 5% when driving pressure (the difference between inspiratory pressure and PEEP levels) was within the range of 5cmH2O (data not shown). This issue was remedied by a more careful control of the rise time (the time during which inspiratory pressure is built up), which could be ensured with a minor software adjustment (using altered mathematical algorithm in the control system). Optimal trigger sensitivity was achieved by setting the manual inspiratory trigger to 1.5–2.5cmH2O below PEEP level.

All prototypes displayed alarm activation in response to all tested alarm simulation scenarios within a 1 second timeframe, except for FiO2 out of range alarms, which were activated within 1min (which is the commercially accepted range for chemical sensors).

In one prototype, the circuit leakage alarm was activated at different times based on the location of the simulated disconnection of the circuit. Significant delay or activation of lower level alarms was observed in instances when the disconnection occurred at the junction of the Y-piece and the expiratory limb, a phenomenon not uncommon even in more advanced commercially available mechanical ventilators. A minor software adjustment rectified this discrepancy.

Closed Suctioning Test was passed according to RMVS guidance but it was noted that pressure readings were highly variable based on suction flow and timing of the procedure (data not shown).

10 physicians (5 female and 5 male) participated in the study, with a mean critical care experience of 9.70 (SD=8.11) years. All 10 physicians performed all 13 tasks correctly. SUS was 90.50 (SD=8.40) for the cohort. SUS did not depend on physician sex (p=0.17) nor correlated with years of experience (rs=0.179, p=0.62).

Sample size calculation concluded that a minimum of 9 patients with 7 time points per patient would provide sufficient data to determine the difference between ventilators concerning volume, pressure and blood gas values. 15 patients, receiving invasive ventilation for a critical illness through an endotracheal tube on day 3–20 after admittance were recruited for the clinical testing stage. Characteristics of patients are listed in Table 1. Respiratory patterns were versatile (see Fig. 3).

Fig. 3.

Mean respirator parameters during ventilator sessions for all patients. Mean respirator parameters calculated from all checkpoints during the ventilator sessions.

(0.21MB).

All 15 patients completed the stage. No adverse events or need for rescue intervention was recorded. No sudden inexplicable changes or variability greater than 20% were observed for any of the vital parameters monitored in the 15 patients during the use of the Luca ventilator (see Supplementary Fig. 1). All blood gas values were within the acceptable range for all patients during both the standard ICU ventilator and Luca ventilator session (see Supplementary Fig. 2). Ventilator parameters recorded during the ventilation sessions for all 15 patients are shown in Fig. 3.

We found no significant, clinically relevant differences between vital parameters, blood gas values or ventilator parameters recorded during ventilation sessions of the Luca ventilator and the commercially available ICU ventilator (see Table 2), except for inspiratory peak flow results, which showed a mathematically significant, but clinically irrelevant difference between the two ventilators.

There were no differences between the number of physician or nurse interventions (data not shown), and although there was a significant difference in alarm frequency, most of these alarms were self-limited and required no intervention.

None of the patients showed signs of dyspnoea as per Borg score. Only 3 of the 15 patients were able to supply a subjective assessment about the Luca ventilator, all of which were “similar or more comfortable compared to the standard ICU ventilator”. Borg scale values for these patients were all 0.