⁶⁸Ga-DOTA PET for Diagnosis of Spinal Cerebrospinal Fluid Leaks

DISCUSSION

We propose CSF PET with ⁶⁸Ga-DOTA as a novel, state-of-the-art approach to RC that provides several possible advantages over conventional RC. By using a very rigorous diagnostic reference standard in an exceptionally large cohort of patients, we were able not only to validate the diagnostic accuracy of CSF PET for verifying—although not localizing—spinal CSF leakage but also to unravel several misconceptions and controversies concerning RC.

PET provides better spatial resolution (2–3 times higher) and sensitivity (2–3 magnitudes higher) than scintigraphy and SPECT (19), allowing the detection and quantification of the slow cephalad ascent of CSF even with short-lived radionuclides. The entire examination is completed within 5–6 h; in contrast, conventional RC often takes up to 48 h. In addition, individual PET scans are acquired within 20–30 min—thus comparing favorably with whole-body SPECT/CT examinations, which may take up to 1 h. The estimated dose for CSF PET with ⁶⁸Ga-DOTA is approximately 1.24 mGy/MBq for the spinal cord (highest organ dose), and the effective dose equivalent is 0.16 mSv/MBq; these doses are slightly higher than those for ¹¹¹In-DTPA (0.95 mGy/MBq and 0.14 mSv/MBq, respectively, assuming the same biokinetic model for both tracers [normal t_1/2,biol] and accounting for differences in energy transfer and physical half-lives) (20). Thus, for a typical injected dose of 37 MBq of ⁶⁸Ga-DOTA, the estimated effective dose is 5.9 mSv. CT scans add about 3–4 mSv, which may be further reduced with dose-sparing techniques. The fast decay of ⁶⁸Ga-DOTA (as opposed to ¹¹¹In-DTPA) also implies that there is no relevant radiation exposure to medical personnel if patients undergo spinal surgery shortly after diagnostics (often on the following day at our institution). The synthesis and quality control of ⁶⁸Ga-DOTA are also very simple and readily available at most larger centers (at least in Europe), rendering it a convenient alternative to ¹¹¹In-DTPA. Finally, the high image quality of PET/CT scans renders CSF PET with ⁶⁸Ga-DOTA much more appealing and convincing than conventional RC—factors that, in our experience, strongly improve the acceptance of RC in clinical routines.

The average t_1/2,biol for patients without verified leakage (9.5 ± 7.0 h) is somewhat lower than but still in line with that in the literature (13.5 ± 4.5 h) (10), considering that this group possibly includes patients with missed or iatrogenic leaks. We propose R5/3 as a simpler estimate of tracer clearance, circumventing the need for a 1-h scan and exponential fitting, with slightly higher diagnostic performance. However, quantitative analyses provided no additional benefit over visual readings, in line with findings for conventional RC (21). Still, we routinely perform quantification as a simple adjunct to visual reading, an approach that may be particularly valuable for follow-up assessments (10).

A particular strength of the present study is the use of a rigorous neuroradiologic reference standard that was additionally validated by surgical findings and outcomes. To the best of our knowledge, no previous study on RC included more patients with appropriate verification (e.g., in the seminal work by Schievink et al. (22), correlation of RC with CT myelography and surgery was available for only 8/11 and 4/11 patients, respectively), with earlier studies often using response to EBP as a reference (10,16,23,24). However, given the uncertainty about the efficacy and mechanism of action of EBP (e.g., success sometimes only after several attempts, targeted vs. nontargeted EBP, often only short-term efficacy, possible placebo effects) (25) as well as the current, improved understanding of SIH pathophysiology (see later text) (3,26–28), we are convinced that the reference standard in current use is the most appropriate. We were able to verify a spinal CSF leak in 18 of 31 patients (58%) with SIH but without previous surgery or alternative causes. Although a few spinal leaks may have been missed (see later text), this fraction is in line with those in the literature (3) and underscores the uncertainty associated with the clinical diagnosis of SIH, which does not require verification of a leak (4).

We made several important observations. First, the site of extrathecal tracer accumulation on CSF PET was not related to the actual location of leakage—something that has not been systematically shown before. We do not believe that this finding is related to the tracer used (e.g., later detection of actual sites of leakage with longer-lived isotopes) but assume that it represents a general shortcoming of RC. In fact, patterns observed in the present study and described in the literature are very similar. However, unlike investigators in earlier studies, we used a rigorous diagnostic reference standard that was a prerequisite for detecting this mismatch. Second, ⁶⁸Ga-DOTA PET, like conventional RC (10,12,13,17,24), may show several sites of extrathecal tracer accumulation—which, however, do not indicate that there are multiple leaks but rather that the tracer leaves the spinal epidural space at multiple locations. This notion is in line with the current concept that spontaneous spinal CSF leaks most commonly occur in the lower cervical spine and thoracic spine (22,26), frequently being caused by well-defined singular lesions (e.g., dural tears due to diskogenic microspurs) (3,26,27). Thus assumption is also reflected by current excellent surgical outcomes. Third, lumbar/lumbosacral extrathecal tracer accumulations are very rarely (if ever) indicative of leaks in that region but are usually of iatrogenic origin instead, although frequently being reported in the RC literature (10,23). Fourth, early bladder activity as an indirect sign possesses little (if any) diagnostic value, in contrast to early RC reports (11) but in agreement with more recent studies (17,21). Fifth, the term direct sign for extrathecal tracer accumulation in the paraspinal space in RC is most likely misleading. In fact, the only direct signs associated with verified leaks in the present study were probably RC equivalents of cervicothoracic extrathecal fluid accumulation on x-ray myelography, commonly referred to as false localizing signs (29). We assume that this finding is simply due to anatomic reasons (e.g., more flexible spinal alignment). To the best of our knowledge, no other systematic studies have provided a detailed (segmental) comparison of neuroradiologic or surgical and RC findings. Even in the aforementioned seminal work (22), anatomic description of CT myelography or RC findings did not pinpoint locations of actual leaks but only provided approximate segmental heights or ranges of extrathecal findings. Given rapid epidural spread and the uncertain mechanism of action of EBP (see earlier text), a therapeutic response to EBP can hardly be taken as a confirmation of an apparent leak localization given by RC, as has frequently been done in the literature (22–24).

CSF PET shows high diagnostic performance in a particularly challenging clinical situation. Extrathecal tracer accumulation at the cervicothoracic junction provides a high diagnostic specificity (90%) but only a moderate sensitivity (67%) for the verification of spinal CSF leaks. The latter fits the observation that direct signs of CSF leakage on conventional RC might be absent in a highly variable fraction of patients with SIH (10%–70%) (13,22–24). In turn, indirect signs of CSF leakage on conventional RC are found in the vast majority (>90%) of patients (11,13,22,24). A lack of tracer appearance over the cerebral convexities on a 5-h scan provides a very high sensitivity (94%) at a moderate specificity (67%). A very high sensitivity of this sign was also suggested by conventional RC studies (13,17,21). The summary rating of both raters was motived by clinical practice with conventional RC and comprised all direct and indirect signs (including those of no or little value), with weighting left to the discretion and experience of the rater. Thus, it is not surprising that this consensus reading was not superior to a combination of both aforementioned features by logistic regression.

Given the considerable rate of negative findings of CT and MR myelography in SIH (45%–74%) (3), it is interesting that 4 of 14 patients (29%) with negative neuroradiologic imaging results and treated with EBP showed a clinically relevant response to EBP. All 4 patients fell into the groups of patients who were judged to show “questionable/possible” and “probable” (2 patients each) signs of a spinal CSF leak but in whom no leak could be verified (Suppl. Figs. 1 and 2). Aside from direct and indirect signs suggestive of spinal CSF leaks (in 3 patients and 1 patient, respectively), all 4 patients also showed below-threshold t_1/2,biol and R5/3 values. Thus, it is tempting to speculate that these patients represented false-negative cases on neuroradiologic imaging, leading to an underestimation of the true sensitivity of CSF PET.

In light of these observations, the clinical role of CSF PET is clearly not to localize the site of a spinal CSF leak. CSF PET may, however, play an essential role in verifying spinal CSF leakage with high sensitivity (to rule out) or specificity (to rule in), depending on clinical need. For instance, extrathecal tracer accumulation at the cervicothoracic junction is most likely associated with a spinal CSF leak (90% specificity) and may prompt additional examinations. In turn, sufficient tracer accumulation over the cerebral convexities strongly argues against a spinal CSF leak (>90% sensitivity), leading to a defensive approach (see also Mokri (21)). One motivation would clearly be to minimize the exposure to ionizing radiation in a predominately young and female population. Patients have a long disease history of months to years and have already undergone a plethora of diagnostic and therapeutic actions. In the present study, on average, 2 combined digital subtraction myelography/CT myelography procedures were performed per patient at our institution only. The resulting median radiation exposure in the present study can be estimated to be about 52.6 mSv per patient (6) or, in other words, enough to cause 1 additional radiation-induced cancer death in 250–500 patients (30). This risk may be substantially reduced by including a gatekeeper examination.

The limitations of the present study include its retrospective nature, which warrants prospective validation. In particular, a direct comparison of CSF PET with conventional RC would be desirable (e.g., by coinjection of tracers). Furthermore, the limited number of cases did not allow for detailed statistical contemplation of interesting subgroups (e.g., SIH vs. non-SIH, type of leakage). Finally, the present study summarized the data for the first patients examined with this novel methodology that was developed in parallel. By retrospectively analyzing the data for this set of patients at once in a strictly masked fashion, we minimized possible effects of the raters’ learning curves and bias. However, optimal technical aspects, such as measures to avoid iatrogenic extrathecal tracer accumulation, still need to be defined. Moreover, although an earlier study showed that upright positioning or exercise does not affect CSF circulation under normal conditions (7), a preliminary report of 2 cases suggested that a sitting position may improve the rate of detection of spinal CSF leaks (31). Further studies are needed to explore whether such efforts or other technical refinements can increase the rate of detection of CSF leakage and possibly improve the localization of spinal CSF leaks by PET.