Open Development and Clinical Validation of Multiple 3D-Printed Nasopharyngeal Collection Swabs: Rapid Resolution of a Critical COVID-19 Testing Bottleneck

Abstract

The pandemic caused by severe acute respiratory syndrome coronavirus 2 (SARS-CoV-2) has caused a severe international shortage of the nasopharyngeal swabs that are required for collection of optimal specimens, creating a critical bottleneck blocking clinical laboratories' ability to perform high-sensitivity virological testing for SARS-CoV-2. To address this crisis, we designed and executed an innovative, cooperative, rapid-response translational-research program that brought together health care workers, manufacturers, and scientists to emergently develop and clinically validate new swabs for immediate mass production by 3D printing. We performed a multistep preclinical evaluation of 160 swab designs and 48 materials from 24 companies, laboratories, and individuals, and we shared results and other feedback via a public data repository (http://github.com/rarnaout/Covidswab/). We validated four prototypes through an institutional review board (IRB)-approved clinical trial that involved 276 outpatient volunteers who presented to our hospital's drive-through testing center with symptoms suspicious for COVID-19. Each participant was swabbed with a reference swab (the control) and a prototype, and SARS-CoV-2 reverse transcriptase PCR (RT-PCR) results were compared. All prototypes displayed excellent concordance with the control (κ = 0.85 to 0.89). Cycle threshold (C_T ) values were not significantly different between each prototype and the control, supporting the new swabs' noninferiority (Mann-Whitney U [MWU] test, P > 0.05). Study staff preferred one of the prototypes over the others and preferred the control swab overall. The total time elapsed between identification of the problem and validation of the first prototype was 22 days. Contact information for ordering can be found at http://printedswabs.org Our experience holds lessons for the rapid development, validation, and deployment of new technology for this pandemic and beyond.

Keywords: COVID-19; SARS-CoV-2; diagnostic testing; epidemiology; virological testing.

FIG 1

Control and prototype swabs. (a) From left to right, the control swab (C; Copan 501CS01), a repurposed urogenital cleaning swab approved for NP testing through our process (R), prototype 1 (Resolution Medical), prototype 2 (EnvisionTec), prototype 3 (Origin.io), and prototype 4 (HP, Inc.). (b) From top to bottom, close-ups of the heads of the swabs in panel a. Bars, 1 cm. (c) Examples of Gram stains of cheek swabs using control (top) and prototype (bottom) swabs. Bar, 10 μm. (d) Examples of materials testing. Clockwise from top left, head flexibility and robustness to fracture, neck flexibility and robustness to fracture, robustness to repeat insertion into and removal from a tortuous canal (diameter, 3 cm), and break point evaluation.

FIG 2

Categorical concordance versus control swab. (a) Two-by-two tables giving counts for each prototype versus the control swab and for control versus replicate control obtained within 24 h from the same individual. Discordant results are in gray, totals for each swab are below and to the right of each box, and the total number of pairs is in bold. K, Cohen’s kappa. (b) Scatterplot of C_T values for pairs of swabs for which at least one swab was SARS-CoV-2 positive. For discordant pairs, the negative swab was assigned a C_T value of 37 (the maximum number of cycles run).

FIG 3

Subjective feedback. (a) Round-robin A/B testing of net preferences among prototypes 1 to 3 (large bold numbers) and the control (C). Each arrow points from the less preferred to the more preferred swab. Arrow weight indicates strength of relative preference. Preferences were unanimous except where noted with numbers separated by a slash: the first number is the number of responses for the direction indicated by the arrowhead, while the second number is the number of responses that had the opposite preference. The weight of the arrow is proportional to the difference (e.g., 7 − 3 = a net preference of 4). Unless otherwise noted, each arrow represents 12 to 15 separate responses. (b) Numbers of positive and negative comments received from study staff who administered the swabs, tabulated by category. In each plot, negative feedback is to the left of the zero, while positive feedback is to the right. The presence of bars on both the positive and negative sides of zero reflects different opinions among study staff. n, total number of comments received about each prototype from study staff.