Don’t delay, but don’t despair: symptom duration, comorbidity and outcome after closure of spinal cerebrospinal fluid leaks

The duration of symptoms before surgery was determined by the personal statements of the patients at admission. Patients’ characteristics like sex, age, Body Mass Index (BMI), nationality, type of spinal CSF leak and the perioperative course (including postoperative neurological deficits and the need for revision surgery) were analyzed retrospectively after systematic collection.

All patients underwent the local diagnostic protocol [27]. Finally, the type and exact level of the spinal leak was confirmed by dynamic digital subtraction myelography, cone-beam CT, and/or dynamic CT myelography [27, 28]. A minimally invasive tubular unilateral approach was used for all spinal leaks (Volz et al., Operative Neurosurgery, accepted 11/2023, https://​doi.​org/​10.​1227/​ons.​0000000000001042​). Ventral leaks were sealed transdurally with an intradural/extradural sandwich patching technique [13]. Lateral leaks were sealed via surgical fibrin-patch augmentation, occasionally with additional clipping or suturing. CSF-venous fistulas were sealed with thermocoagulation of the draining vein after intraoperative visualization by intrathecal fluorescein injection [29] and additional clipping of the nerve root. Initial confirmation of successful sealing of the leak was based on the intraoperative assessment of the responsible surgeon and a subjective and objective clinical improvement of clinical symptoms in the first days after surgery. Final confirmation of a successful closure was achieved via clinical assessment and via routinely performed MRI of the brain and spine 3 months postoperatively.

For statistical analyses, SPSS Statistics^® (IBM Corporation, IBM SPSS Statistics for Macintosh, Version 29.0, Armonk, NY, USA) was used. Continuous data are presented as median with interquartile range (IQR), while categorical data are reported as frequencies. Patient-reported outcomes (PROMs) related to headache impact (HIT-6) and quality of life (EQ-VAS and EQ-Index) were compared between baseline and 6 months using the Wilcoxon signed-rank test. Subsequently, subgroup analyses were performed based on symptom duration (</≥ 90 days) and split by the median of the SCQ summary score (≤/> 1), with comparisons conducted using the Mann–Whitney U test. The association of symptom duration and comorbidities measured by SCQ summary score, age, sex, BMI, height, and leak type with patient-reported outcome measures at baseline and 6 months was assessed using multiple regression models, respectively. Bootstrapping (sample size 1000) was employed to correct for optimism. Symptom duration was log-transformed due to right-skewed data. In subsequent regression models, variables demonstrating potential effects (p < 0.2) were retained. Model fitting was evaluated using an F-statistic in an analysis of variance (ANOVA). We anticipated an indicative moderate goodness-of-fit (adjusted R² ≥ 0.13) for reporting [30]. Statistical significance was assumed at p < 0.05.

Between September 2020 and March 2023, 137 patients underwent surgical closure of a spinal CSF leak in our institution. Among them, 30 were either technically or linguistically unable to complete the questionnaires or declined to do so. Additionally, seven patients could not recall a distinct onset of their symptoms; these individuals presented with gait ataxia (3/137), superficial siderosis (2/137), bibrachial amyotrophy (1/137), and localized back pain (1/137), thereby necessitating their exclusion. Consequently, the analysis included 100 patients (Fig. 1).

Of these patients, 61% were female. The median age was 43.5 years (IQR 36–53), and the median BMI was 23.9 kg/m² (IQR 21.3–27.0). Leak classification revealed that 72% had ventral leaks, 21% had lateral leaks, and 7% had CSF-venous fistulas. The median duration of symptoms was 5 months (IQR 2–15), equivalent to 173 days (IQR 80–482), spanning a range from 9 days to 13 years. Regarding the SCQ summary score, the median was 1 (IQR 1–4), with scores ranging from 0 to 15. Notably, only 28 patients had a summary score of ≥ 4, and merely 2 had a score of ≥ 10 (Table 1).

The median HIT-6 scores significantly improved 6 months after sealing of the CSF leaks from 65.5 (IQR 61.8–69.3) to 51.5 (IQR 40.0–61.0, p < 0.001). Both measurements for Quality of life (QoL) significantly increased: EQ-VAS increased from 40 (IQR 30–60) to 79 (IQR 60–90), and EQ-Index increased from 0.669 (IQR 0.345–0.799) to 0.910 (IQR 0.796–0.943) (p < 0.001, respectively) (Table 2, Fig. 2).

Essentially, all patients, divided by symptom duration (</≥ 90 days) and divided by comorbidity (SCQ summary scores ≤/> 1), demonstrated a significant improvement in the impact of headaches and quality of life (p < 0.001, respectively, Table 2, Fig. 2).

Multiple regression analysis revealed a moderate correlation of longer symptom duration with reduced QoL 6 months after surgery (p = 0.020 for EQ-VAS and p = 0.035 for EQ-Index). Higher comorbidity scores moderately correlated with a more pronounced impact of headaches and reduced QoL 6 months after surgery (p = 0.011 for HIT-6, p = 0.001 for EQ-VAS and p = 0.006 for EQ-Index) (Table 3). However, no substantial effect of symptom duration was observed in the second component of the QoL measurement (EQ-VAS) or in the HIT-6 score, both at baseline and the 6-month follow-up. Male sex showed a moderate correlation with a worse perceived QoL at baseline (p = 0.004 for EQ-VAS and p = 0.003 for EQ-Index). Age, BMI and type of leak did not correlate with the outcome parameters.

In the subgroup analysis, patients with a symptom duration < 90 days reported higher baseline HIT-6 scores compared to those with longer lasting symptoms, with both groups experiencing severe impact of headaches (68.5 (IQR 65–71) vs. 64 (IQR 61–68), p = 0.007). However, while HIT-6 scores at the 6-month mark were not statistically different (46 (IQR 40–57.5) vs. 52 (IQR 44–61.8), p = 0.157), the improvement was higher in the patient subgroup with a shorter symptom duration. Patients with a symptom duration < 90 days exhibited a significantly better outcome in EQ-Index (0.923 (IQR 0.909–0.992) vs. 0.881 (IQR 0.737–0.943), p = 0.037).

Patients with a SCQ summary scores ≤ 1 had higher HIT-6 scores preoperatively (66 (IQR 63–70) vs. 64 (IQR 61–69), p = 0.08) and a more pronounced improvement after the surgery (48 (IQR 40–48.5) vs. 52 (IQR 44–62), p = 0.25), but without a statistically significant difference, respectively. Patients with SCQ summary scores > 1 had consistently lower EQ-VAS and EQ-Index scores preoperatively. Postoperatively, there was a significant overall improvement in both subgroups. However, patients with SCQ summary scores ≤ 1 reached significantly higher QoL scores after surgery (EQ-VAS 84 (IQR 67–93) vs. 71 (IQR 50–85), p = 0.003 and EQ-Index 0.918 (IQR 0.866–1.00) vs. 0.851 (IQR 0.734–0.934), p = 0.003).

Except for expectable unilateral hypoesthesia in the respective dermatome in cases with clipping of a thoracic nerve root (3/21 lateral leaks and 7/7 cases of CSF-venous fistulas), no permanent neurological deficit occurred. Ten patients (10%, 6 ventral leaks and 4 lateral leaks) needed revision surgery: three patients (all ventral leaks) due to an epidural hematoseroma. Two patients (one ventral and one lateral leak) had inconclusive intraoperative findings; the subsequent postoperative imaging revealed “wrong level” surgery. Five patients (two ventral and three lateral leaks) had no improvement or a recurrence of their original symptoms; the subsequent imaging revealed a recurrence of the original leak within 10 days up to 5 months.

Symptom duration before a targeted treatment of spinal CSF leaks shows a high variability. Our study’s median symptom duration was 5 months, ranging between 9 days and 13 years. Cheema et al. report a median time from symptom onset to the correct diagnosis of 2 months with a range of 0–180 months, and after that, a median period of 10 weeks until surgery [3]. Before surgical treatment of CSF-venous fistulas, Wang et al. report an average duration of symptoms of around 40 months [11] and Lohkamp et al. a symptom duration between 1.5 and 3 years [31]. Notably, neither study reports an impact of symptom duration on the outcome. Häni et al. report a mean symptom duration of around 12 months before surgical treatment [9] of different leak types. In studies of patients with CSF-venous fistulas, Brinjikji et al. report a mean symptom duration of 40.9 months before transvenous embolization [16] and Callen et al. a mean symptom duration of 34.6 months before CT-guided fibrin glue occlusion [18].

The latter three studies report a significant impact of symptom duration on the postoperative outcome. In Häni’s series, the duration is the only factor with a significant impact. They report, that a symptom duration below 87 days was correlated with a better outcome [9], promoting the recommendation to strive for a prompt diagnosis and therapy, ideally within 3 months if SIH is suspected [3, 4]. In Brinjikji’s series, several factors influence the outcome, with symptom duration as the only modifiable variable [16]. Callen et al. found, among other variables, a significant association between symptom duration and the clinical improvement [18]. However, in all three studies, the clinical outcome was analyzed using three-step ordinal scales without specific scores for the impact of headaches or QoL.

Our study correlated the symptom duration and the level of comorbidities among other clinical parameters using validated patient-reported outcome measures (PROMs) for the impact of headaches on daily life (via HIT-6) and quality of life (QoL via EQ-VAS and EQ-Index), allowing for a more nuanced description and interpretation. Prolonged symptom duration and increased comorbidities moderately correlated with reduced QoL. Therefore, prompt treatment to avoid prolonged disease duration is recommended, as it is the only modifiable variable. However, it is important to note that our results pertain to a natural logarithmic scale of disease duration. This implies that higher levels of change occur in the early phase of the disease, with diminishing effects in later stages. In simpler terms, our findings suggest that postponing treatment to 1 or 2 years from symptom onset, instead of intervening at 90 days, could potentially reduce EQ-VAS scores by about 4 to 6.5 points, respectively. Conversely, early intervention, such as at 10 days, could lead to an additional improvement of 7 points compared to surgery at 90 days (Table 3). Given the wide confidence intervals and the predominance of patients with a disease duration of more than 90 days, caution is necessary in interpreting these approximations, and further studies are necessary to confirm the results.

HIT-6 scores at baseline were significantly higher in patients with shorter symptom duration, representing a more severe impact of headaches. This observation aligns with the commonly reported phenomenon where orthostatic headaches, characteristic at the onset of the disease, may become less typical and recede into the background during a prolonged course [1, 20, 32‐34]. The subgroup with shorter symptom duration demonstrated a higher improvement in HIT-6 scores of 22.5 points 6 months after surgery (baseline 68.5 to 46 at 6 months). Notably, the postoperative score of 46 is clearly below the widely recognized threshold of 50 points, indicating no or minimal impact of headaches on daily life [11, 12, 23, 24]. Patients with a symptom duration exceeding 90 days, while still benefitting significantly from surgery, did not reach this decisive cut-off (median HIT-6 score: baseline 64 to 52 at 6 months). Possible explanations for these findings may include the chronicity of headaches, a potential association with anxiety or depression after a prolonged course, a potential overlap with different headache types, and extended readaptation processes of CSF and venous homeostasis. However, the observed improvement of 12 points in the HIT-6 Score is undeniably significant and clinically relevant [12, 35, 36]. These results reinforce the importance of prompt diagnosis and appropriate therapy, while longer symptom duration does not preclude a significant headache improvement.

This study marks the first systematic investigation into comorbidities among surgically treated SIH patients using a well-established scoring system. The Self-administered Comorbidity Questionnaire (SCQ) efficiently integrates patient-reported data into a systematic Patient-Reported Outcome Measure (PROM) tool, sidestepping potential issues with incomplete datasets seen in medical record-based assessments. This method provides reliable results without relying on medical professionals’ records. Various studies have highlighted the SCQ’s predictive power, likening it to the commonly used Charlson Comorbidity Index [21, 22, 37]. Obviously, pre-existing comorbidity stands as an inherent non-modifiable factor before any surgical treatment. Our study demonstrates that comorbidity is relatively low in a representative cohort of SIH patients. 50% exhibited no comorbidity, with only 2% exhibiting a SCQ score above 10. This reaffirms previous claims about the relevance of the disease, which suddenly severely affects otherwise completely healthy patients in their everyday lives [3, 38] (Volz et al., Neurology Clinical Practice, accepted, CPJ-2023-000465R2). However, this low comorbidity contributes decisively to the markedly positive benefit–risk profile and the swift, substantial postoperative improvement.

Corresponding to the findings regarding symptom duration, patients with lower comorbidity initially present with higher preoperative HIT-6 scores, yet they achieve better improvement, reaching a median score below 50 points, indicating minimal or no impact of headaches. However, these trends, while suggestive, do not reach statistical significance. Notably, the subgroup with higher comorbidity consistently displays lower QoL scores. While both subgroups significantly benefit from surgery, patients without comorbidity reach an even better outcome.