X-ray detection of ingested non-metallic foreign bodies

To determine the utility of X-ray in identifying non-metallic foreign body (FB) and assess inter-radiologist agreement in identifying non-metal FB.

Focus groups of nurses, fellows, and attending physicians were conducted to determine commonly ingested objects suitable for inclusion. Twelve potentially ingested objects (clay, plastic bead, crayon, plastic ring, plastic army figure, glass bead, paperclip, drywall anchor, eraser, Lego™, plastic triangle toy, and barrette) were embedded in a gelatin slab placed on top of a water-equivalent phantom to simulate density of a child's abdomen. The items were selected due to wide availability and appropriate size for accidental pediatric ingestion. Plain radiography of the embedded FBs was obtained. Five experienced radiologists blinded to number and types of objects were asked to identify the FBs. The radiologist was first asked to count the number of items that were visible then to identify the shape of each item and describe it to a study investigator who recorded all responses. Overall inter-rater reliability was analyzed using percent agreement and κ coefficient. We calculated P value to assess the probability of error involved in accepting the κ value.

Fourteen objects were radiographed including 12 original objects and 2 duplicates. The model's validity was supported by clear identification of a radiolucent paperclip as a positive control, and lack of identification of plastic beads (negative control) despite repeated inclusion. Each radiologist identified 7-9 of the 14 objects (mean 8, 67%). Six unique objects (50%) were identified by all radiologists and four unique objects (33%) were not identified by any radiologist (plastic bead, Lego™, plastic triangle toy, and barrette). Identification of objects that were not present, false-positives, occurred 1-2 times per radiologist (mean 1.4). An additional 17% of unique objects were identified by less than half of the radiologists. Agreement between radiologists was considered almost perfect (kappa 0.86 ± 0.08, P < 0.0001).

We demonstrate potential non-identification of commonly ingested non-metal FBs in children. A registry for radiographic visibility of ingested objects should be created to improve clinical decision-making.