The research went through multiple phases, with each phase requiring several iterations (see Supplementary Figure S1). First, from a literature study and expert interviews requirements were identified. The most promising design concepts were selected by expert consulting. Concepts were evaluated on safety, hygiene, comfort, user experience and transfer success. In this study a simulation-based development method was taken (11), therefore fabrication was followed by a first round of user tests with fetal and maternal manikins. A Failure Mode and Effects Analysis (FMEA) was performed to reveal risks and mitigation strategies (30). This FMEA analysis has multiple steps: a Hierarchical Task Analysis (see Supplementary Table S2) gave an overview of all tasks, and a Fault Tree Analysis (see Supplementary Table S3) revealed possible dangers like injury and impaired efficiency and ergonomics (31). Hazards were ranked based on severity and probability using a risk assessment matrix. Mitigations for the hazards were drawn up and procedural and design adjustments were made accordingly, after which final liquid-based simulations were performed (6).

MRI data of a fetus of 24 weeks gestational age were used to develop a mold to cast a silicone manikin (Ecoflex 00-30, Smooth-on, USA). For the maternal manikin, a PROMPT Flex Birthing Simulator (Limbs & Things, United Kingdom) was used. The PROMPT simulator was 3D scanned (Eva Scanner, Artec 3D, Luxembourg) to enable subsequent 3D modelling and 3D printing of a custom watertight abdominal insert (PLA filament, Ultimaker, the Netherlands). This insert, including tissue layers and a uterus, was developed to allow for increased simulation realism and validation of the prototypes. The skin layer, with a premade skin incision, consisted of a top layer of silicone Ecoflex 00-30 (Smooth-on, Macungie, USA), glued on top of 2 cm thick polyether foam. A bag, functioning as amniotic sac was placed in the uterus and could be interchanged and varied in size based on gestational age of the fetal manikin.

Different study stages included semi-structured interviews and simulations. Simulations increased in complexity and number of participants throughout the process, starting with dry simulations (no “amniotic fluid”) in the beginning, to liquid-based simulations at the final stages of the design process. Participants were selected for clinical expertise or preterm delivery/care experience. Animations or concept drawings of transfer procedures and equipment were displayed in interviews. Participants tested low- to high-fidelity prototypes on manikins in simulations.

Fabrication of transfer device prototypes took an iterative approach. The set of instruments for the CS include: two wound retractors, a retractor connector and a transferbag with integrated gloves. An inflatable wound retractor and transferbag with integrated gloves are included for VB. The transfer station adjusts for delivery type, keeps the transferbag near the placenta, with attachment for the oxygenator. The fabrication of all these prototypes is described below. A list of anatomical and (bio)mechanical parameters on which design decisions were based can be found in Supplementary Table S1.

The procedures and handling of the transfer device prototypes contain many steps, different instruments, and require teamwork. By performing risk assessments early in the development process, possible flaws and their associated hazards can be identified, making it much easier to minimize error-prone aspects of the design (41, 42). In high-stress environments, this novel procedure potentially presents future operators with a high (mental) workload and associated safety risks, highlighting the critical importance of device usability for ensuring safe and efficient procedures (43). A Failure Modes and Effect Analysis (FMEA) provides a systematic framework for the analysis of the potential failures, their causes, and their effects on the entire procedure: (1) task identification (2) hazard identification, (3) risk assessment, and (4) risk mitigation in case of unacceptable risk levels (42). We made use of the ISO 14971:2019 standard for the application of risk management.¹

The first set of simulations were dry simulations, hence performed without filling the intra-uterine amnion bag and transferbag with water. The simulation was conducted with a total of five participants (P29–P35), including three perinatologists, a technical physician, and a medical engineer also educated as a midwife. Sessions took place at the Dept. of Industrial Design (Eindhoven University of Technology) and at a medical simulation center (Medsim, Eindhoven, the Netherlands). The sessions involved the use of fetal and maternal manikins. Participants were instructed on procedural tasks, tool handling and their place within the entire clinical team. Only tasks related to non-pharmacological treatments were included. Throughout the protocol, discussions were stimulated, and idea generation was facilitated using the prototypes. Illustrations, animations, or live instructions showed participants procedural tasks, device handlings, and operational hierarchy. Task cards with checklists helped participants complete tasks in correct sequence throughout the simulation (adapted from Figures 3, 4). Liquid-based simulations sessions were held with seven participants (P43–P49). Also, within these simulations, the simulation of pharmacological treatments was omitted. Liquid-based transfer simulations included in this study were only for CS, not (yet) for VB which included only dry simulation.

During the first set of simulations, photos of prototype interaction and procedural tasks were taken of the participants. Minutes were drawn up after each session. Of the liquid-based simulations transcripts were made from the (audio/video) recordings, in certain cases recording was not possible and minutes were made directly after the session. The knowledge gained through interviews and simulations was continuously compared to previously collected information. Through data and source triangulation internal validity was assured. Data triangulation was assured using a variety of methods such as simulation, individual and group interviews. Source triangulation was assured by including specialists from different disciplines and sources of experience.

Testing the approach of filling only the lungs instead of submerging the entire perinate's body with liquid, is beyond the scope of this paper.

For the present study we focused on infants of 24 weeks gestational age, in line with current Dutch guidelines on care at the edge of perinatal viability. Past preclinical studies have shown successful extracorporeal support up to 4 weeks, and therefore perinates between 24- and 28-weeks of gestational age have been the target group for this study, in relation to development of the Perinatal Life Support system (9). Expert input led to various assumptions to guide procedure development within specified boundaries for the first patient group to be addressed. Twin pregnancy is not included, both mother and fetus are healthy and stable, the fetus is in cephalic presentation, the placenta allows VB and CS, the umbilical cord is not entangled and the perinate is cannulated while ex-utero. Future studies should include a wider range of scenarios and should include cannulation while the fetus is still in-utero, similar to EXIT-to-ECMO procedures to account for uterus involution and shearing of the placenta. The obstetric procedure for perinate placement in the APAW system requires an adaptation from current preterm care. It should prevent transition to neonatal physiology, facilitate rapid umbilical vascular cannulation to the extracorporeal circuit and shield the perinate from environmental factors such as temperature shifts, mechanical trauma, sound, and light (9).

What factors initiate neonatal transition is not entirely understood. Although the suppression of the breathing reflex could be maintained by administering certain drugs, the strategy behind this study was to envision a method to prevent transition through non-pharmacological means. In both modes of delivery these interventions can be ensured by creating (1) an air and watertight artificial passageway, either in the birth canal or via the abdominal incision to access the uterus. This passageway should shield the perinate from air, but also keep the environment sterile by preventing exposure to e.g., maternal pathogens (Group B strep), which should be of extra concern for transfer by VB. The environment in which the perinate is captured, the transferbag, should (2) allow for the supply of warmed artificial amniotic fluid (AAF). Lastly, the procedure should (3) allow the umbilical cord to be cannulated in proximity to the placenta due to its limited length, which is 40 cm on average at 24 weeks of GA (45).

Based on the expert interviews and the subsequent simulations, a list of general procedural and design requirements was drawn up, see Table 1. For a comprehensive list see Supplementary Table S1.

Wound protector-retractors provide circular retraction by separating the edges of the incision, hence moving tissue aside to improve access. In general, retractors exist of two flexible rings that are connected by a cylindrical plastic film, thereby forming a tunnel that can be placed in bodily cavities. Using the wound retractors, a watertight artificial passageway can be created. The outer end of the retractor forms an interface onto which other instruments can be attached, such as a liquid-filled transferbag. After delivery, by removing the transferbag from the birth canal, the umbilical cord is exposed and can be cannulated while placing the transferbag on a stable platform in proximity to the mother (see Figure 3).

In this study an inflatable retractor was created to form an airtight passage from the uterus to the transferbag. Accordingly, the retractor was designed to secure against the birth canal whilst reducing maternal stress and perineal damage. Pressure closure and thereby automatic adjustment and comfort to surrounding tissue and anatomic structures favored inflatable constructions. Inflatable constructions were preferred for their pressure closure mechanism, enabling automatic adjustment to and comfort in the surrounding tissue and anatomical structures. In addition, they minimize the chance of trauma by lacking sharp or rigid elements.

One of the first iterations for the retractor design allowed the retractor and transferbag to be connected and attached to a handle (see Supplementary Figure S2). The handle hosted the air and fluid supply/drain tubing. Upon entry in the birth canal, the retractor could be inflated, and the perinate could then slide into the transferbag. Having all components (retractor, transferbag, supply/drain, user interface) in a single device made it easier to operate, but it also had several drawbacks related to sterility and accessibility to measure cervical dilation. In the final design (Figure 2), integrated gloves in the transferbag replace the device handle to help direct the perinate into the bag. Separating the transferbag and retractor permits the retractor to be placed in the birth canal and upon sufficient cervical dilation, the retractor can be inflated and the transferbag attached. Additionally, sterility can be better maintained since the transferbag has less contact with the birth canal.

In the final prototype, manufacturing of the inflatable retractor is based on directed lamination of two plastic films, containing air pockets and bonded areas. The inflatable retractor's asymmetrical sleeve conforms to the birth canal's shape, therefore this emphasizes the necessity for proper insertion by the user into the manikin's birth canal.

One way to reduce degrees of freedom during insertion was by pre-assembling the retractor with an applicator. Three flexible arms are placed in the inner pockets of the sleeve which can be pulled out easily afterwards. The handle secures that the sleeve can only be inserted in the right way. The final prototype has an additional fixation ring that is embedded in one end of the retractor that, while pre-folded upon entry in the birth canal, will unfold around the cervix to allow passage of the perinate (see Figure 2A). During insertion, the retractor is pushed up towards the cervix, by the clinician's index and middle finger. The other end of the retractor, outside the vulva, is connected to a rigid ring onto which the transferbag can be attached. This transferbag is again made of TPU with two integrated gloves, an inlet for AAF supply and a valve to release trapped air.

The length of the umbilical leaves limited space to perform the cannulation if the perinate is already delivered in the transferbag. Hence, we designed a transfer station comprising an adjustable platform for supporting the transferbag, an integrated reservoir for heated AAF supply and a holder for the oxygenator during perinatal umbilical cord cannulation.

Figure 3 shows the sequential steps of the VB transfer procedure that were performed during the simulation. In case of anticipated preterm birth, the position of the fetus and placenta is assessed by ultrasound, and routine fetal and maternal monitoring is used. The perinatologist determines the required cervical dilation based on gestational age and perinate head size. After cleaning the perineum and vulva, the interior ring of the inflatable retractor is introduced in the birth canal, while the exterior ring remains external. The transferbag is pre-filled and placed on the transfer station while one perinatologist places one hand in the integrated glove. After membrane rupture, the transferbag is connected to the retractor. An air valve in the transferbag releases trapped air and sufficient AAF is supplied. Uterine contractions, maternal pushing, and guidance by the clinician allow the perinate to slide into the birth canal and subsequently into the transferbag. The transferbag is then detached from the retractor and placed in the holder to prevent fluid leaking and to allow visual examination of the perinate. To create an extracorporeal circuit with the AP, the umbilical vein and arteries must be readily cannulated. The perinate is transferred to the APAW system where monitoring, oxygenation, and nourishment of the perinate is taken over. The perinatologist guides placenta delivery and sutures perineal tears if required.

To broaden the applicability of this treatment, and because half of premature infants are born via CS (2, 8), we investigated the possibility of a transfer via CS. Making use of the same concept, a passageway is created to which a liquid-filled transferbag can be attached. Following skin incision, an existing wound retractor SurgiSleeve (Medtronic, USA) can be used to obtain a clear operating field. Subsequently, the retractor connector can be placed onto the outer facing ring of the retractor. The inner ring is located tight against the peritoneal layer. A second wound retractor is placed after the uterine incision is performed. This uterus wound retractor has the shape of a truncated cone, with the inner facing ring having a smaller diameter than the outer-facing ring. The outer facing ring can be placed in a notch in the retractor connector. A transferbag can connect to this set of tools, thereby creating a direct passageway from the uterus (see Figure 4).

Several design adaptations were made after the first rounds of simulations. To prevent the flow of AAF from the pre-filled transferbag into the abdominal cavity of the mother, the second wound retractor was added. With the additional retractor, potential inflow from blood from the incision wound edges into the uterine cavity is also prevented. This would otherwise mix with amniotic fluid and can then cause cloudiness and limit the view during the CS. Lastly, the wound retractors protect the incision wound edges against contamination. A second adaptation is related to the improved degree of freedom for the clinicians during the operation. During a standard CS, gynecologists need to make great use of their hand dexterity, however, the mounted transferbag onto the incision makes it harder to deliver the baby. Hence, gloves were integrated in the transferbag to reach for and guide the perinate, similar to the VB transferbag.

The retractor connector contains grooves for placement of the two outer-facing rings of the retractors (see Figure 3B). Rubber labyrinth and O-ring seals were placed at strategic places, on both the retractor connector and the transferbag to ensure watertightness (see Figure 2C). Three clips secure the transferbag onto the retractor connector, a reversible action when the perinate has been delivered. When the infant is delivered into the transferbag, the bag needs to be detached from the base by lifting it from the wound retractor, aided through the integration of notches (see Figure 2C).

The proposed transfer procedures require a specific sequence in tool usage. To enhance success and reduce cognitive load, this study underscored the importance of prioritizing usability and learnability. Design concepts like affordance (design elements that imply how to use an object) and signifiers (visual/auditory function indicators) provide the operator clues for usage and improve user experience (55, 56). We employed perceivable markers, such as labels and colors, to convey where and in what order an action should be performed. The design also includes affordances such as clips, gloves, but also geometric shapes in which the other elements can only be assembled in one manner.

Figure 4 shows the proposed sequential steps of the CS transfer procedure that were also performed in simulation. The mother is placed in 15 degrees left lateral tilt position (57) with body support components [e.g., an (inflatable) wedge]. Following the incision in the skin, the wound retractor and retractor connector are used. After the uterus incision, additional artificial amniotic fluid is supplied to the uterus, to keep the perinate's head submerged. The transferbag is filled with AAF and connected to the retractor connector. The mother is tilted to her left side and the perinate is guided into the transferbag. After placement into the holder of the transfer station, the umbilical vessels need to be rapidly cannulated to the oxygenator and the mother can be placed in a supine position. If needed, the transfer procedure can be halted at any moment during the procedure and a switch can be made immediately to standard preterm birth care.

Given the similarities between the designs, a joint FMEA and joint hazard mitigation were chosen. For each part it will be indicated whether it concerns CS, VB, or both.

All critical hazards (Table 2) were addressed in the last phase of the study, during (liquid-based) simulations and expert interviews (P36-P50). Table 3 lists the mitigations used to reduce or eliminate the risks with highest criticality scores. Adjusting procedure instructions and protocol guidelines decreased several hazards in addition to design changes.

Continued submersion of the manikin's upper airways appeared difficult. To improve this, participants kept the transferbag horizontal throughout a CS simulation (P43–P47). Liquid-based simulations included placing the maternal manikin in 15° left lateral tilt, yet this proved insufficient. Fluid transferred from the bag into the native uterus when the hand entered the glove to reach for the manikin. Participants suggested positioning the glove in the center of the bag and making a fluid pouch at the transfer bag's end (P43–P47) (see Table 3). This design change made it easier to maintain the fluid volume but difficult to reach the manikin. Procedural adjustments included supplying extra fluid to the open uterine incision site; this amnioinfusion-like supply tube has again to be removed before attaching the transferbag. This can be facilitated when the surgical assistant controls the fluid supply and ensures that the fetal head remains submerged. Simulations showed that filling half of the transferbag before connection to the mother, allowed the perinatologist to comfortably put on the integrated glove. After connecting the transferbag to the device base, trapped air was released, additional fluid supplied, and the manikin was delivered under continuous submersion. To minimize dehydration and temperature decline, participants recommended keeping the perinate's temperature at the native uterus' temperature and the UC moist and warmed (P39, P41, P42).

Allowing placement of the manikin's whole body in the warmed liquid helped maintain its warmth (see Table 3). Design mitigation proposals for the transfer station were a warming pad onto which the transferbag can be placed and a continuously warmed liquid reservoir. Another aspect mentioned was the increased risk for infection in case the fragile skin gets into contact with the exterior, which was also addressed by the mitigation of full-body immersion (P39, P42).

At a certain point in the procedure, the transfer bag carried multiple liters of fluid and a manikin, which posed a risk of operator fatigue and an increased likelihood of dropping it. To alleviate this, a platform onto which the bag could be placed was proposed, i.e., a transferstation, and proved essential during subsequent simulations for bag handling, to stably place the manikin after delivery for examination (see Figures 5, 6). In the simulations, the transferbag was filled while supported vertically whereas during the transfer, the station was tilted horizontal, and back to vertical mode during cannulation. A separate holder for the transferbag ring prevents spilling out of AAF. The limited length of the UC will make it necessary to keep the perinate close to the mother during cannulation, especially in case of a bypass cannulation. Using flexible and moveable arms the transfer station could be placed close to the mother (see Figures 5, 6).

The hazardous situation, wherein the perinate is unreachable while having the transferbag attached on the retractor connector, may arise due to various errors such as the perinate's inability to pass through the retractor, inadequate hand dexterity or grip. In a VB transfer, inflation of the retractor could form a hazard when overinflated, or when the retractor is wrongly positioned in the birth canal and the perinate cannot pass through. Making this retractor using TPU film instead of silicone and embedding specific laminations provided expandability within required ranges. An additional retractor ring at the interior side of the sleeve that fits around the cervix was suggested to prevent the retractor from moving (P50). This additional retractor ring could be bent upon entry, and thus mimics the insertion of a NuvaRing, a GelPOINT V-Path or the cervical retractor by Whalen et al. (35). In one simulation session, it was noted that the retractor, while moving slightly along with the manikin, remains positioned in the birth canal without outward movement due to the fixation ring. Aside from its benefits in positioning, the additional retractor ring demonstrated another positive effect. During another simulation session it was observed that although the retractor could slightly move along with the manikin, the retractor remained fixed in the birth canal due to the fixation ring, preventing outward displacement. Another suggestion for improved positioning that was fabricated, and tested was to embed a two-finger integrated glove in the retractor film to facilitate insertion or employ an applicator with arms that glide in and out (see Table 3). Lastly, markers or other labeling on the sleeve or ring were suggested to visually guide positioning.

During a regular delivery, either CS or VB, the minimum dexterity needed for perinatologists is to deliver with one hand and hold the fetal armpits with the ring and index fingers. With the aim of making a one-size fits all integrated glove, participants tried out gloves with 2 pockets (mitten), 3 pockets, and 4 pockets (middle and ring finger joined). The mitten and three-finger glove appeared too small to deliver the manikin (P29–P35), while 4-pockets functioned better (P43–P49). Simulation participants preferred a 5-finger glove (P43-P49), making it essential to develop transferbags with varied glove sizes to accommodate all medical staff. Yet, this creates an additional problem when two gynecologists want to switch roles after the start of the delivery.

To facilitate a secure grasp on the manikin and to speed up the delivery, an extra integrated glove designed for the right hand was introduced on the opposite side of the transfer bag, thereby also rendering the bag suitable for use by both left and right-handed gynecologists (P48, P49). The indirect contact with the manikin and the slippery environment caused by using fluids made it difficult to get a proper grasp on the manikin during liquid-based simulations. Participants proposed texturing the integrated glove on the material side that faces the clinician, which could be embedded in future prototypes (P45–P47). To attach the transferbag to the connector, clips must be locked, and this required some time and force during the simulations. Also, retractor film can become entrapped. Snap-fit, turn-and-lock and other sealing mechanisms were proposed that may reduce forces, allow easier attachment, and speed up the process (see Table 3).

During several simulation rounds participants addressed sterility management and proposed that the transfer station could be easily supplied with sterile wrapping. When the transferbag is left open, either before or after deliver, a hatch placed over the transferbag (with an opening for the UC) could improve to keep the AAF sterile (P43–P47).

Procedural improvement and human-related error mitigation regarding task division were also evaluated. In the first set of simulations the main surgeon, i.e., perinatologist 1, made incisions while positioned at the right side of the patient. Perinatologist 2, positioned at the left, inserted the retractors and delivered the manikin into the transferbag. Participants felt uncomfortable since the main surgeon usually stands on the patient's left side during a CS. Participants suggested, tested, and favored the left side execute incisions and deliveries, while the right-side perinatologist inserts wound retractors (P43, P44). It was evident that two scrub nurses, one on either side of the table, were required to handle the transfer device's parts, fill the bag with AAF, and adjust the transfer station (P45–P47).