Source: Good CD, Wade AM, Hayward RD, et al. Surveillance neuroimaging in childhood intracranial ependymoma: how effective, how often, and for how long? J Neurosurg. 2001;94:27–32.Following biopsy, resection, and adjuvant treatment (when indicated) for a pediatric brain tumor, patients are scheduled for regular neuroimaging studies until the tumor recurs or the patient is “cured.” The primary rationale for surveillance imaging is to detect a recurrence at the earliest possible time. On the other hand, surveillance imaging should withstand objective analysis with respect to cost, risk, unintended consequences, and benefit. Furthermore, as imaging technology and therapeutic options improve, such analysis requires continual updating. A 1994 analysis of surveillance imaging in the management of medulloblastoma,1 for example, was recently revisited because of advances in salvage therapy.2,3Unfortunately, there have been no dramatic advances in the treatment of intracranial ependymoma during the era of magnetic resonance imaging, the period of time covered by this retrospective report from the Great Ormond Street Hospital for Sick Children in London. Fifty-two children with intracranial ependymomas were treated from 1987 through 2000. Recurrences developed in 28 patients. In 16 patients (57%) the first recurrence was symptomatic; in 12 patients (43%) surveillance imaging disclosed the first recurrence before symptoms developed. The timing of surveillance imaging was fairly erratic because of the international referral population served at this institution, but the median interval between postoperative imaging studies was 5 months. The median interval between a symptomatic recurrence and the preceding (negative) surveillance study was 5.5 months. The longest sequence of negative surveillance imaging studies prior to a recurrence lasted 4.3 years. Survival from diagnosis was shorter for patients with symptomatic recurrence than for patients with recurrence detected by surveillance (median survivals not stated; P=.016 by Cox proportional hazards regression). Median survival from first recurrence was 10 months in the symptomatic group and 32 months in the surveillance group (P=.056). The authors recommend surveillance imaging at 6-month intervals for at least 5 years after treatment because after that time they consider the patient probably “cured.”Surveillance imaging must justify itself in terms of benefit because the costs, risks, and unintended consequences are substantial. In this study the cost for the detection of 18 asymptomatic recurrences (including 6 recurrences after retreatment) was 386 surveillance imaging studies (>21 studies per recurrence). The risk of sedation is low in a well-run imaging unit, but the periodic surges of anxiety that accompany surveillance imaging are an underappreciated detriment to quality of life. So what is the benefit?Symptomatic patients had shorter survivals from first recurrence than patients with recurrences detected by surveillance, but there are two important sources of bias that could be responsible for this observation. The first is lead time bias: detection of a disease earlier in its course makes survival from diagnosis look longer even if the treatment is not beneficial. Because the median interval between surveillance images was 5 months, the apparent improvement of 22 months in median survival from first recurrence is probably not entirely due to lead-time bias. The second is length bias: aggressive tumors are more likely to grow from undetectable to symptomatic in the interval between periodic surveillance images than are less aggressive tumors. Thus, patients who suffer a symptomatic recurrence despite surveillance imaging may have an intrinsically more malignant disease than do those patients whose recurrence is detected by surveillance. The authors do not discuss this kind of bias and, in fact, length bias is impossible to eliminate without analysis of volumetric tumor growth rates or some other measure of disease virulence. Postoperative surveillance imaging for intracranial ependymoma remains a matter of custom rather than science.Surveillance of Dr. Piatt’s commentary highlights two important principles of study interpretation— lead-time and length bias—both of which can adversely affect a study as it does in this article by Good et al. Keep these biases in mind the next time you need to evaluate a paper on how often to monitor for disease recurrence.