Background: Pica is characterized as repeatedly eating or chewing of a non-nutritious substance including, but not limited to ice, clay and dirt, starch, raw pasta, chalk, coal, paint, or paper. Life-threatening conditions, such as gastrointestinal infections and lead poisoning, may occur depending on the substance consumed. The Diagnostic and Statistical Manual of Mental Disorders, Fifth Edition defines pica as an eating disorder, but with its strong link to iron deficiency, pica is also characterized as a hematological condition. Blood donation is associated with substantial iron loss. Blood donors often become iron deficient and are susceptible to pica, as well as restless legs syndrome (RLS), an often concurrent condition characterized by an irresistible urge to move legs due to uncomfortable sensations such as tingling, crawling, burn, and pain. The National Heart, Lung, and Blood Institute sponsored REDS-III program prospectively enrolled 13,403 racially diverse blood donors in its RBC Omics study to investigate demographic, clinical, hematological, and biochemical factors associated with pica.

Methods: Demographic data, survey questions of pica behaviors and RLS, complete blood count (CBC), and ferritin were collected and analyzed using univariable and multivariable logistic regression analysis. Exclusion criteria were ferritin >800 ng/mL, body mass index (BMI) <17.5, and those with 9 or more donations in the 2 years prior to enrollment who were purposely oversampled for in RBC Omics. Pica was rigorously defined by questionnaire responses: 1) "Yes" to eating ice and, 2) Consumes "8 oz - 24 oz" or "more than 24 oz" of ice and, 3) At frequency of "once a day" or "several times a day". Alternatively, "yes" to eating non-ice substances regardless of amount and frequency was also characterized as pica.

Results: The analysis included 11,418 donors with 7,141 Caucasians, 1,581 African Americans, 1,601 Asians, and 1,170 Hispanics. Pica was present in 1.4% (158 out of 11,418) of the study subjects. Asians were less likely to have pica than Caucasians (odds ratio [OR], 0.50 p=0.030), African Americans (OR, 0.41; p=0.014), or Hispanics (OR, 0.33; p=0.002). Females (n=6,044) were more likely to have pica than males (n=5,449; OR, 1.43; p=0.029). Donors ≤30 years were more likely to have pica than donors >30 years (p<0.001). Donors with BMI >30 were more likely to have pica than donors with BMI 17.5 - 25 (p=0.036) and BMI 25-30 (p<0.001). Ferritin was negatively associated with pica (p<0.001). For donors with ferritin <12 ng/mL 2.4% (52/2155), 12-50 ng/mL1.3% (67/5118), and >50 ng/mL 0.9% (39/4145) had pica. CBC values associated with low iron stores were also associated with pica: low hemoglobin (p=0.002), low hematocrit (p=0.014), low mean corpuscular volume (MCV) (p<0.001), and high red blood cell distribution width (RDW) (p<0.001). In multivariable analyses, RLS emerged as the most powerful predictor of pica in our forward stepwise logistic regression model (OR, 4.38; 95% confidence interval, 2.14 - 8.17), p<0.001). Other significant demographic predictors of pica included hormone supplement use (p=0.006) and nulligravida females (p=0.009).

Conclusions: Iron deficiency was confirmed as a strong predictor of pica. Consistently, demographic (female), clinical (RLS) and hematological (hemoglobin, MCV, RDW) parameters associated with iron deficiency also were associated with increased risk for pica. However, pica symptoms were also reported by donors with ferritin >50 ng/mL suggesting that non-iron related factors influence pica presentation as well. Pica has a significant prevalence in the blood donor population, and healthcare providers should consider non-iron related factors when diagnosing pica.


Mast:Novo Nordisk: Honoraria, Research Funding.

Author notes


Asterisk with author names denotes non-ASH members.