Estimated Risk of Post-Concussion Syndrome One Year after a Mild Head Injury may be Severely Biased, Bordeaux, France

INTRODUCTION: Reports estimating the frequency of Post-Concussion Syndrome (PCS) following Mild Head Injuries (MHI) are inconsistent, casting some doubt on the specificity of this syndrome. When sampling for the follow-up of a cohort, we experienced unbalanced losses to follow-up. The objective of this analysis was to assess how the prevalence of PCS could be biased by the likelihood of remaining in the sample.

METHODS: We included patients who attended the emergency department of the University Hospital in Bordeaux, France, between December 2007 and February 2009. Patients included were MHI with mild brain injury (MHI/B), MHI without brain injury (MHI/NB), and Non-Cranial Injuries (NCI). PCS was defined at M3, according to DSM-IV; all patients with PCS and 20% of those without PCS were proposed further follow-up. Estimates of the relative risk (RR) of developing PCS at M12 in patients with MHI compared to patients with NCI were estimated under various scenarios on the risk of PCS in losses to follow-up (LFU).

RESULTS:  Of 953 MHI/B, 1043 MHI/NB and 1180 NCI, 2124 (67.2%) were seen at M3. MHI/B and MHI/NB were 1.51 (95%CI: 1.33-1.71) and 1.38 (1.22-1.57) more often LFU before M3 than NCI (p<0.001). Among 611 patients selected for follow-up, LFUs at M12 were fewer in patients with (54.3%) than without PCS at M3 (63.8%; p<0.001). M3 prevalence was 11.3% in NCI, 13.9% in MHI/B and 14.3% in MHI/NB; M12 prevalence of PCS varied from 1.3% (MHI/NB without PCS at M3) to 58.8% (MHI/B with PCS at M3). Bias in the M12 relative risk of developing PCS could vary from -99.7% (PCS in all LFUs in MHI/B, but none in LFUs in NCI) to +23% (maximum underestimation of PCS frequency in all groups).

CONCLUSIONS:  Estimates of the frequency and associations with determinants of PCS are probably strongly biased.