Purpose: The use of head circumference (HC) growth curves is recommended at all well child visits until two years of age to screen for pathologic conditions that may cause abnormal head size. However, little guidance is available for their use. We sought to evaluate the ability of HC growth curves to distinguish between children with and without pathology that may cause macrocephaly.
Methods: We performed a retrospective cohort study of children born before 1/31/2008 with at least one HC measurement in the electronic record of a large primary care network before 1/31/2009. The primary outcome was the new diagnosis of an intracranial expansive condition or metabolic/genetic condition that can cause macrocephaly. Subjects with the outcome were identified by reviewing the records of subjects with diagnostic or billing codes indicating potential pathology. Subjects diagnosed with an outcome-defining condition prior to their first recorded HC or a birth weight <1500g were excluded. Subjects’ HC percentiles were evaluated using the CDC and WHO growth curves as well as primary care network (PCN) growth curves derived from this population. The test characteristics of previously proposed cutoffs (95th, 97th and 99.6th percentile as well as HC percentile crossing two major percentile lines) were evaluated.
Results: Of 73,439 evaluable subjects, 85 (0.12%) with the outcome were identified; 15 (0.02%) underwent surgery for their condition. LRs, sensitivity, and PPV are presented in the table. All NPV were 99.9%.
| Cutoff | PCN | CDC | WHO | |
| 99.6th | LR Sens PPV | 18.5 17.7% 2.2% | 10.8 24.7% 1.3% | 8.8 31.8% 0.9% |
| 97th | LR Sens PPV | 8.1 36.5% 0.8% | 5.4 42.4% 0.5% | 4.2 49.4% 0.4% |
| 95th | LR Sens PPV | 5.7 41.2% 0.5% | 4.2 47.1% 0.4% | 3.7 56.5% 0.3% |
| Crossing | LR Sens PPV | 2.1 40.5% 0.2% | 1.9 63.1% 0.2% | 1.4 48.8% 0.2% |
Conclusions: The incident prevalence of pathology that may cause macrocephaly in this primary care population was low. Although the LRs for the HC percentile cutoffs are high, particularly for the PCN curves, all studied cutoffs have a low PPV and low sensitivity. Prospective research is needed to provide guidance to clinicians regarding how to combine the information provided by HC percentiles with other clinical information to guide clinical decision-making.
Candidate for the Lee B. Lusted Student Prize Competition
See more of: The 32nd Annual Meeting of the Society for Medical Decision Making