The aim of this study was 1) to identify challenges and considerations for prototyping and producing medical devices via 3D printing (stereolithography) in limited resources such as during the COVID-19 pandemic and 2) to produce a ventilator splitting system to rapidly increase ventilator capacity. The design of the study was descriptive research article with proof-of-concept experimental lab and large mammal trials. The setting was multi-institutional academia in two universities with engineering schools, and affiliated large teaching hospitals.
1.0 Design Committee
The design committee composed of clinicians (critical care physicians, anaesthesiologists, plastic surgeons, burn surgeons), engineers (mechanical, civil, materials engineers), clinical research scientists, graphic designers, 3D Printing artist, and FDA consultant. The translational team (clinicians, clinical research scientists, FDA consultant, lead engineers) met to identify challenges of 3DP in limited resource settings, design needs for a ventilator splitter, and rapid pre-clinical testing. The technical team (engineers, graphic designers, 3D Printing artist) met to optimize designs and production conditions for limited resource settings. The rapid prototyping cycle included production of the devices by the technical team, testing by the translational team with feedback to designs, followed by redesign and production by the technical team for improved generation of devices with each cycle.
1.1 3D printing procedure
3D printing of the Vent-Lock splitters, flow regulators, and manometer adaptors were produced via stereolithography (Form 2, Form 3, or Form 3B, Formlabs) at 50 um layer resolutions, using surgical guide resin (Surgical Guide, Formlabs) and standard protocol per Formlabs. Print files were generated by CAD drawings (SolidWorks, Dassault Systèms) and converted into G-code using the printer’s accompanying software package (PreForm, Formlabs). Support structures were minimized through design and generated using PreForm where needed. Components were oriented in such a way that crucial surfaces such as threads or O-ring ledges were not impacted by support structures. Prints were post-processed by washes (2 cycle with 15 min per cycle) in >99.5% isopropyl alcohol (CAS Number: 67-63-0, Sigma Aldrich), followed by air-drying at 22 oC for 30 minutes, and post-cured for 30 minutes with heat 60 oC for the Form 2 printer and 70 oC for the Form 3B printer at 405 nm of light (Form Cure, Formlabs). O-rings (E1000-212/AS568-212, O-Rings EPDM, FDA EPDM, Marco Rubber & Plastics, Seabrook, New Hampshire, USA) were added for improved sealing. Production via fused deposition modeling (FDM) (e3d, BigBox3D Ltd, Oxfordshire, UK; Little Monster, Tevo 3D Electronic Technology Co. Ltd, Zhanjiang, China) used PETG Filament (PETG 3D Printer Filament, FilaMatrix, Virginia, USA). Print settings were a 0.2mm layer height with 30% infill, nozzle temperature of 250 oC, and bed temperature of 70 oC; supports were generated from the build platform, with no interior supports. If the previously mentioned machines are unavailable, we recommend five key requirements (resolution, surface quality, biocompatibility, sterilizability, and tolerancing) to successfully utilize other methods and machines (Fig. S8).
1.2 Sterilization Testing
3D-printed parts produced from surgical guide resin were sterilized by dry vacuum autoclave (Sr 24C Adv-PlusTM, Consolidated Sterilizer Systems, Boston, Massachusetts, USA), 3 cycles at 120.0 oC, 20 minutes sterilization time and 20 minutes dry time. Then, they were soaked in >99.5% isopropyl alcohol (CAS Number: 67-63-0, Sigma Aldrich) for 30 minutes, air-dried at 22 oC for 30 minutes, and placed in an oven at 40 oC in humidified air for 48 hours (VO1824HPC, Lindberg/Blue M Vacuum Oven 127.4L, Thermo Scientific, Waltham, MA, USA). Particle count analyses were conducted using a particle counter (SOLAIR 3100, Lighthouse Worldwide Solutions), detecting sizes 0.3 to 10 microns, for 1-minute cycles, and performed for parts pre-autoclave, post-autoclave and post IPA wash, and humidified warm air exposure at 40 oC.
1.3 Vent-Lock 1+n(1) circuit and components
Vent-Lock circuits were assembled as depicted in Fig. 1. Vent-Lock 3DP splitters, flow regulator, and manometer adaptors were used. Commercial components include manometer (Ambu Disposable Pressure Manometer, Ambu, Copenhagen, Denmark), one-way valves (22F x 22M, REF 50245, Mallinckrodt Pharmaceuticals), disposable bacteria filters (BSF104, Vincent Medical), and ventilator tubing (SKU: 999027588, Hudson Rci).
1.4 Simulation Center Testing
Vent-Lock 1+n(1) circuits were tested at the Johns Hopkins Medicine Simulation Center (JHMSC). The ventilator (Puritan Bennett 840 Ventilator System, Avante Health Solutions) was set to pressure control mode of ventilation (Volume Ventilation Plus™, Avante Health Solutions) with additional settings detailed in fig. S4. Vent-Lock 1+1 circuit was tested using test lungs simulating healthy lungs with variable compliances (Standard patient: Rp = 2 cmH2O/L/s, RespiTrainer Advance, QuickLung, IngMar Medical; Variable patient: Rp= 50 cmH2O/L/s, ASL 5000, IngMar Medical). Intrapulmonary data for both patients were collected; data included peak inspiratory pressures, tidal volumes, and peak end expiratory pressures. Five total values of tidal volume per data set were collected and averaged. Corresponding ventilator reported data was also recorded, including total expiratory volumes, peak inspiratory pressures, mean inspiratory pressures, and peak end expiratory pressures. Flow restrictors (#P20034 PVC SCH 40 ½-in FNPT Ball Valve; G300 Lead Free Brass Gate Valve; #P60SCPVC12 Stop and Waste Valve, American Valve, Greensboro, North Carolina, USA; Vent-Lock 3DP FloRest) were used to restrict the variable patient’s inspiratory flow rate per the 1+n(1) circuit (Fig. 1). Valve handles were turned at smallest increments permissible to close the valve and documented as % closure. Corresponding intrapulmonary data and ventilator reported values were collected per handle closure and standardized to values (volumes and pressures) of a fully open valve (reported as proportion of maximum, %).
In vitro studies were also conducted at the Washington University Simulation Center using two Datex-Ohmeda Aestiva anesthesia machines. One machine was set to deliver pressure control ventilation in a manner similar to that performed at Johns Hopkins University. This machine was connected in parallel to a 2L anesthesia bag reservoir and a second Datex-Ohmeda Aestiva machine that was set to spontaneous ventilation. The second machine served as a flow and volume sensor for the Vent-Lock 1+n(1) circuit.
1.5 In vivo swine studies
Experiments were performed in accordance with the Guide for the Care and Use of Laboratory Animals and were approved by the Institutional Animal Care and Use Committee of Washington University School of Medicine (St. Louis, MO). Domestic swine (Sus scrofa domesticus) were purchased from Oak Hill Genetics (Ewing, IL). The swine were females, 72 kg each, 5 months old, and were Landrace-cross swine. Swine were sedated with a telazol, ketamine, xylazine cocktail and intubated with a 7.0 endotracheal tube. Anesthesia was maintained with isoflurane. Femoral venous and arterial catheterization was performed. Standard ASA monitoring was maintained throughout the experiment. Swine were ventilated using a single ventilator (Drager Narkomed 2A) with two circuits in parallel in an 1+n(1) configuration with cross-ventilation restricted by using one-way check valves. Ventilation was maintained with volume control. One swine was not flow-regulated and thus considered the standard patient, while the other had a Vent-Lock 3DP 4.0 connected in the inspiratory limb and thus considered the variable patient. Flow was measured at each expiatory limb with a SS11LB airflow transducer (Biopac; Goleta, CA). Flow data were collected at 2kHz using an MP36 data acquisition unit and BSL 4.1.3 software (Biopac; Goleta, CA). The spirometry data was then smoothed with a 0.25 sec wide moving median filter after removal of instrument noise below 0.08 L/sec (determined by histogram inspection). The smoothed data was then numerically integrated to estimate respiratory tidal volume, and a first order numeric derivative was used to calculate the instantaneous respiratory rate. The noise floor for the integrated volume was determined by histogram inspection resulting in a threshold of 90 mL. The anesthesia record and the spirometry results were then aligned using common timestamps. All breaths spontaneously initiated by the swine (identified by respiratory rates more than 30% away from the ventilator set point) were removed from analysis. The mean and standard deviation for each anesthesia record entry were calculated for respiratory rate, tidal volume, minute ventilation, and lung compliance. All of the described analysis was performed using a custom MATLAB script (MATLAB 2019b, The MathWorks, Inc., Natick, MA)]. Arterial and venous blood gas data were collected 15 minutes following any changes to the Vent-Lock 3DP device. Following the procedure, swine were euthanized with an overdose (~150mg/kg) of supersaturated potassium chloride IV while under anesthesia. Necropsy was performed to assess for any gross lung pathology.
1.7 Statistical Analyses
Raw and calculated data were exported from MATLAB script (MATLAB 2019b, The MathWorks, Inc., Natick, MA). All statistical analyses were completed using Stata v.13 (StataCorp, College Station, TX) and Microsoft Excel (2018, Redmond, Washington). Variables were analyzed using two-tailed Student’s t tests and chi square analyses. The threshold for statistical significance was set at an alpha value of 0.05.
1.6 Vent-Lock 1+n(1) circuit and components
We validated a 1+n(1) system which can split one ventilator between one standard patient and one or potentially more variable patients (Fig.1). The standard patient ideally has the lowest lung compliance and has minimal components in the circuit to establish low resistance allowing the ventilator to maintain standard function. The standard patient will be ventilated at pressure settings unaltered from that delivered by the ventilator. Additional patients (n) added to the circuit are considered variable patients and can have their tidal volumes and PEEP altered by circuit components. This paper demonstrates use of a ventilator splitter adjusting for 1 control and n=1 variable patients. The 1+n(1) split contains Vent-Lock 3DP parts and commercial parts (Fig. 1). We 3D printed the splitters and the flow restrictors (needle valves). The other parts including the one-way check valves, the filters and the PEEP valves were all commercial parts (fig. S1). The Vent-Lock circuit was designed to be closed-circuit and leak-free to minimize risk of aerosolizing viral particles into the surrounding environment.
1.7 Vent-Lock 3DP Flow Restrictor (FloRest)
The Vent-Lock 3DP flow restrictor (Vent-Lock FloRest) is a flow restrictor based on a needle control valve design optimized for low flow rates to offer clinicians robust control over a range appropriate for human ventilation. Vent-Lock FloRest was designed to address the following concerns regarding ventilator “splitting” (Oshitani 2020): 1. Volumes would distribute unevenly between patients, 2. PEEP would be difficult to manage per patient, 3. Tidal volumes would be difficult to manage per patient and 4. Adjustment or discontinuation of ventilation to one patient would alter breathing dynamics to other patients.
The goal of FloRest (Fig. 2A) was to allow the clinician to modify the airway resistance delivered to the patient, thus providing ranges of flow rates, clinical tidal volumes, and PEEPs with control sensitivity and a reliable relationship between closure and flow rate, tidal volume, and pressures. The design emphasizes the minimization of build time and volume by reducing support material use and complex structures for consistent and higher quality printing. These considerations allowed for an air-tight and leak-proof design (fig. S2) and utilization of biocompatible materials that can withstand extended exposure to warm humidified air and sterilizing autoclave environment. Using a particle counter, post and pre-autoclave tests demonstrate significant micron particle reduction after autoclaving (fig. S3).
The needle valve utilizes change in flow momentum, flow path geometry and orifice flow design concepts allowing easy control of flow rate vs pressure drop ratio (i.e. flow coefficient) compared to gate and ball valve concepts (more binary valve concepts). The threading allows for control over the flow rates and offers the clinician the ability to make fine adjustments to the flow within the range of control. A gasket featuring an O-ring (E1000-212/AS568-212, O-Rings EPDM, FDA EPDM, Marco Rubber & Plastics, Seabrook, New Hampshire, USA) seated between the needle and chamber to ensure airtightness, thus reducing the risk of aerosolizing the virus into the surrounding environment. Final features of FloRest (Fig. 2A) included sealing to the external environment using unthreaded upper needle shafts for smoother interfaces between cap O-ring and needle during operation of valve and an increase in needle length prior to engagement of needle threading to provide more precise control of range of flow, and to allow for safe operation of valve by preventing full occlusion of flow to patient by clinician.
The FloRest designed with the clinician in mind, has advantages compared with commercial valves in terms of controllability, biocompatibility, and sterilizability. The FloRest had similar range of control compared to commercial gate valves (#P20034 PVC SCH 40 ½-in FNPT Ball Valve; G300 Lead Free Brass Gate Valve, American Valve, Greensboro, North Carolina, USA) (Fig. 2B). However, the Vent-Lock FloRest was found to be easier to control than commercial valves, characterized by more points available for the clinician to choose from corresponding to different tidal volumes delivered. Furthermore, FloRest is produced with biocompatible, nontoxic materials that can be safely sterilized, as compared to commercial ball valves with untested biocompatibility and unknown sterilization protocols. Vent-Lock FloRest can be produced at an estimated $3.50 per device in 3-hour 40 min print and process time via fused deposition modeling (FDM) (e3d, BigBox3D Ltd, Oxfordshire, UK; Little Monster, Tevo 3D Electronic Technology Co. Ltd, Zhanjiang, China) using PETG (PETG 3D Printer Filament, FilaMatrix, Virginia, USA). With stereolithography (SLA; Form 2, Form 3, or Form 3B, Formlabs), it costs approximately $25, and 16 hours production time with a 50 micron build layer height resolution, using surgical guide resin (Surgical Guide, Formlabs). We demonstrate that the FloRest is leak proof through air volume testing (fig. S2).
1.8 Vent-Lock 3DP Flow Restrictor (FloRest) control of tidal volumes and PEEP
We tested the use of Vent-Lock FloRest in the Johns Hopkins Medicine Simulation Center (JHMSC) to confirm the following: 1) Allowing volumes to be distributed evenly between patients, 2) variable patient control of PEEP, 3) variable patient tidal volume control and 4) changes in the variable patient breathing settings does not alter breathing dynamics to the standard patient.
We tested the Vent-Lock multiplexing system using a 1+1 split patient circuit (Fig. 1). We used one ventilator (Puritan Bennett 840 Ventilator System, Avante Health Solutions) to ventilate two patients with different lung compliances of 20 mL/cmH2O and 50 mL/cmH2O, and monitored the intrapulmonary gas volumes, pressures, and compliances of the simulated patient lungs. We first tested using a pressure control mode, with inspiratory pressures set at 25 mL/cmH2O (additional ventilator settings available in Fig. S7). The Vent-Lock FloRest allowed adjustment of tidal volumes delivered to patients between 7- turns to 9- turns (fully closed); thus, the range of control for FloRest corresponds with 2-⅛ turns (Fig. 3A). Within these turns, the tidal volume of the variable patient can be decreased 85.7% (compared to initial variable patient tidal volume) with negligible change in tidal volume delivery to the standard patient (range: 99.86% and 103.2% initial standard patient tidal volume, mean: 102.1% ± 0.98%). We note that the total expiratory volume reported by the ventilator trends with tidal volume delivered to the variable patient (Fig. 3A) and the peak inspiratory pressures (PIP) of the variable patient and ventilator peak inspiratory volumes also correspondingly decrease with decreases in tidal volume (Fig. 3B,C), while peak inspiratory volumes remain stable for the standard patient and peak end expiratory volumes remain stable for both patients during these changes.
We repeated Vent-Lock 1+1 multiplexing patient circuit with the ventilator on volume control mode to deliver a total of 2L of volume, corresponding to approximately 600 mL of tidal volume per patient (additional ventilator settings available in Fig. S6). We note that turning of FloRest on the variable patient resulted in decrease of both tidal volume and PIP (Fig. 4A,B). However, this was accompanied with an increase in tidal volume delivery and PIP to the standard patient (Fig. 4A,B), with relatively stable ventilator reported average pressures (Vent Pavg) and PEEP (Fig. 4B). Thus, unlike in pressure control mode where control of delivery to the variable patient was independent of the standard patient, flow restriction in the volume control mode resulted in the modification of the ratio of tidal volumes delivered (Fig. 4C, standard/variable patient tidal volume ratios).
We replicated results using anesthesia gas machines (North American Drager Narkomed 2a, Ardus Medical; GE Aestiva 5 7900, Datex Ohmeda) at an alternate test site (Washington University in St. Louis, St. Louis, Missouri, USA) to demonstrate generalizability across locations and different resource settings. The 1+1 circuit was tested with the Vent-Lock FloRest on the variable patient. On both pressure control and volume control settings, the Vent-Lock FloRest demonstrated control of tidal volume delivered to the variable patient with stable tidal volumes delivered to the standard patient. Pressure control allowed slightly greater range of control (Fig. 4D, reduction of 43.9% tidal volume at close, compared to 27.5% reduction of tidal volume at close with volume control). These results indicate that both patients’ inspiratory and expiratory times, respiratory rates, and PEEP is determined by the ventilator or anesthesia gas machine’s settings. Changes in one patient’s condition does not alter these settings for either patients, as this is pre-determined by the ventilator settings. The variable that can be moderated during pressure-control setting is the tidal volume. The tidal volume is set a constant volume, and can be adjusted by Vent-Lock FloRest to optimize tidal volume to the variable patient, with no change to the constant patient. Therefore, this allows for constant, stable patient conditions, even while the clinician is optimizing tidal volumes to one of the patients.
1.9 Real-time pressure reporting with Vent-Lock manometer adaptors
To facilitate continuous monitoring of pressures we designed a manometer adaptor that allows clinicians to either spot-check pressures or continuously monitor with the use of standard, disposable manometers such as those found on bag-valve-masks. The manometer adaptor can be added in the circuit at any point and is designed to accurately reflect breathing pressures, such as PIP and PEEP. We demonstrate that the manometer accurately reflects real-time pressures when incorporated in the circuit (Fig. 5A). In a 1+1 circuit with the ventilator on pressure control, the PEEP setting was incrementally increased, and associated ventilator detected PEEPs and Vent-Lock manometer reported pressures were recorded. The Vent-Lock manometer reported pressures were equivalent to the PEEPs (Figure 5B). We also conducted blind tests, where one researcher set the PEEP on the ventilator and a second researcher (blinded to ventilator PEEP settings) reported PEEP as reported by the Vent-Lock manometer. The second researcher consistently and accurately reported all test values between 0 cmH2O to 50 cmH2O, in 5 cmH2O increments, with total ten trials with no error.
1.10 Vent-Lock 3DP Flow Restrictor (FloRest) in Swine
Two domestic swine were anesthetized and ventilated using the Vent-Lock system with constant volume delivery. Swine were successfully ventilated for approximately four hours using a single ventilator. Initial calculated dynamic lung compliances were 50.3 and 48.1 ml/cm H2O for the standard and variable swine, respectively. Throughout the experiment, minimum and maximum dynamic compliance ranged from 50.3 to 244.5 and 37.9 to 87.1 ml/cm H2O for the standard and variable swine, respectively, reflecting the differences in tidal volumes those swine received during the flow restriction trial. Serial ABGs were monitored (Fig. 6) and initial shared ventilator settings were determined to be too high as both swine developed a respiratory alkalosis. At approximately two hours this was corrected and pH and paCO2 were allowed to normalize for one hour. Vent settings were not changed following this equilibration. Over the next hour the Vent-Lock system was adjusted from fully open to fully closed, where air was still allowed to pass even when Vent-Lock is closed to prevent unintentional hypoventilation. Respiratory characteristics including tidal volume ratios, percent of total set tidal volume delivered, inspiratory pressure and tidal volume are presented in Fig. 7. While the tidal volume delivered to the variable swine decreased marginally, a substantial increase in tidal volume was noted to the standard swine (Fig. 7D), similar to what is seen in simulation center testing with ventilator on volume control mode. Arterial blood gas measurements demonstrated hyperoxia in both swine (Fig. 7C). A hypercarbic respiratory acidosis occurred in the variable swine (Figure 6A, B) as the Vent-Lock closure reached its final turn. Necropsy performed to assess for gross lung pathology showed no significant findings of all lung lobes.