Global incidence of spinal perineural Tarlov’s cysts and their morphological characteristics: a meta-analysis of 13,266 subjects

Background Spinal perineural Tarlov’s cysts (TCs) are considered incidental findings that occasionally might exert pressure upon nerve roots and correspond with patients’ signs and symptoms. Purpose of this meta-analysis is to deliver global incidence and characteristics (location, size, and shape) of TCs. Methods Following PRISMA checklist, all major databases were searched by two authors for radiologic studies reporting incidence and morphologic features (location, size, and shape) of TCs. Anatomical Quality Assessment tool was applied for risk of bias evaluation. Meta-analysis of random-effects model was employed. Subgroup analysis for regional distribution, gender, sacral levels, age, correspondence with symptoms, and persistent genital arousal disorder (PGAD) were planned ahead. Results 22 radiologic studies of level 3 evidence involving 13,266 subjects were included. Global pooled prevalence of TCs was 4.18% (95% CI 2.47–6.30). Mean pooled sagittal diameter was 11.86 mm (95% CI 10.78–12.93). Sacral cysts strongly prevailed over the other segments. Of the sacral, S2 level was the most common (46.7% [95% CI 29.4–60.5]). Geographically, the highest incidence was found in Europe (6.07% [95% CI 1.49–13.00]), followed by North America (3.82% [95% CI 0.49–9.44]), and Asia (3.33% [95% CI 1.52–5.75]). TCs were more common in women than in men (5.84% vs 3.03%, p < 0.001, test of homogeneity, χ2). Subjects with PGAD had incidence of 37.87% (95% CI 2.45–81.75). TCs in pediatric population are rare—0.53% (95% CI 0.02–1.51). 15.59% of TCs corresponded with symptoms. Conclusions Spinal perineural (Tarlov) cysts are found in a minority of population. S2 level of the sacral bone is affected most frequently. There is female predominance. Correspondence with symptoms is seen in less than one-fifth of TCs. Studies with stronger evidence level are needed to corroborate the results. The purported high incidence in PGAD requires confirmation in case–control studies for the risk-ratio calculation. Electronic supplementary material The online version of this article (10.1007/s00276-020-02644-y) contains supplementary material, which is available to authorized users.


Introduction
Spinal perineural Tarlov's cysts are thought to be common radiological findings, often incidental and asymptomatic. They are a dilation of the nerve-root sheath and tend to communicate with subarachnoid space through a valve-like mechanism, which thus contain cerebrospinal fluid (CSF) as well as neural tissue [3,6] (Fig. 1). Traditionally, according to Nabor's classification (see Table 1), they are type 2 spinal meningeal cysts, albeit differentiation of these types might often be done only on histological inspection [16]. Incidence varies widely in the literature with reports of as high as 17.7% and as low as 0.38% [4,14,24]. Also, it has been a subject of controversy which spinal level and nerve root they are most often associated with. Although they are usually linked to the sacral segment of the spine, rarely, they may be seen at other levels, including the cervical and thoracic spine. Despite being mostly asymptomatic, they have been discussed as a potential cause of radiculopathy, persistent genital arousal disorder, as well as a source of CSF leak being a primary culprit of idiopathic intracranial 1 3 hypotension [22]. Authors present the first meta-analysis aiming to discover global incidence and characteristics of Tarlov's cysts.

Search strategy
Major electronic databases (

Eligibility
Analysis was limited to those studies that met our inclusion criteria: (1) clear incidence of Tarlov's cysts among a specified group of subjects; (2) the incidence regarded the group of at least 10 subjects. Papers were excluded if any criterion of the following was found: (1) conference abstracts, (2) commentaries, (3) letters to the Editors, (4) insufficient data, (5) incidence of Tarlov's cysts was aggregated with other lesions and could not be disaggregated, (6) case reports, and (7) series with less than 10 subjects among whom presence of the cysts was evaluated. TK and WO are fluent in English and Polish. In case of other language, professional translators were inquired.

Data extraction
Data were extracted by two of us (TK and WO). The following characteristics were of interest: (1) total number of subjects evaluated for presence of Tarlov's cysts, (2) number of individuals with at least one TC, (3) number of individuals with multiple TCs, (4) total number of TCs in the group, (5) number of patients with only one, only two, only three (and so on) TCs, respectively, (6) number of TCs in sacral, lumbar, thoracic, or cervical region, respectively, (7) number of TCs corresponding with symptoms (excluding back pain) or signs, (8) number of TCs not corresponding with symptoms or signs, (9) number of TCs in patients with persistent genital arousal disorder, (10) number of TCs in patients

Statistics
Formal statistical estimation of pooled prevalence was executed using software MetaXL 5.3, EpiGear International Pty Ltd. (Brisbane, Australia). Estimation of mean sagittal diameter was executed with Comprehensive Meta-Analysis V3 (Englewood, New Jersey, USA). For single-categorical variables, pooled prevalence was estimated with corresponding confidence intervals of 95%. I 2 and χ 2 were incorporated into evaluation of heterogeneity. I 2 value was interpreted with classic approach: 0-40%: not be important, 30-60% moderate heterogeneity, 50-90%: substantial heterogeneity, and 75-100%: considerable heterogeneity. Arbitrary p value of < 10% was set for Cochrane Q significance (< 0.10). Conversely, significance of comparative tests was agreed to standard < 5% (< 0.05). Random-effects model for metaanalysis of pooled prevalence was applied. Subgroup differences in prevalence were compared using test of homogeneity based on χ 2 (Statistica 13.3.0, TIBCO Software Inc, Palo Alto, USA).

Subgroup analysis
Subgroup analysis was planned with regard to geographical regions, specific entities such as persistent genital arousal disorder, sex, and age (children and adults). Distribution within the sacral spine was also analyzed to determine which sacral level is affected most commonly.

Selection process
Initial search yielded 4762 records including 2 from references of the pertinent articles. Following the de-duplication, 3518 titles and abstracts were checked for relevance to the topic. 41 full-text articles were assessed for eligibility ( Fig. 2). In total, 22 papers were included in meta-analysis [2, 4, 5, 9-15, 17-21, 23-29]: 19 for estimation of the general incidence as well as for subgroup analysis, whereas three articles were included solely in subgroup analysis without involving them in the general pool, because they pertained specifically to small groups of patients with symptoms of PGAD, two of them presented unusually high incidence of TC, and hence were deemed to pose a risk of bias for the general pool.

Risk of bias and quality
Tabular display of AQUA tool summary is presented in Table 2. The highest risk was related with methodology, whereas the lowest risk was attributed to study design and reporting results. The funnel plot is presented in Fig. 3.

Characteristics of the included studies
Characteristics are presented in Table 3. In total, 13,266 subjects were included into the meta-analysis. 21 studies sought Tarlov's cysts in MRI scans, whereas one study evaluated only myelograms. No study with evidence stronger than level 3 was identified. Time frame of the studies spanned from 1980 to 2020. Age was known in 14 studies (n = 6717). Thus, mean age of those in the main cohort was 48.9 years. Geographical distribution was as follows: 8 studies from North America (n = 1298), 6 from Europe (n = 5903), 7 from Asia (n = 5808), 1 from Africa (n = 217), and neither from Australia nor South America.  Table 4).

Discussion
Although in literature a varyingly wide range of incidence of TCs is reported, the present meta-analysis shows that mean global pooled prevalence is 4.27% and the confidence interval is relatively narrow (95% CI 2.56-6.38). The highest general incidence-17.65%-was suggested by Larsen et al. in 1980 [14]. As contributory and impactful their research might have been, it was the oldest of the included studies and also the only one that took advantage of myelography. All the other papers were from the new millennium and evaluated cysts using magnetic resonance imaging. This crucial methodological difference might partially explain the discrepancy. 8 studies addressed gender-specific prevalence. In women, TCs occurred at 7.01% (95% CI 2.57-13.21) whereas men had prevalence of 4.05% (95% CI 1.07-8.57). Apparently, females are more often affected by TCs (p = 0.0003, test of homogeneity, χ 2 ). Literature has not explored the causes of this sex-related predominance, yet. Our hypothesis is that formation of TCs in adult females might be associated with the level of primary female sex hormones, since in pediatrics, when these hormones are usually low, TCs are an exceptional finding [12]. This hypothesis might be corroborated in future by conducting case-control research studies comparing a group of hypogonadism with healthy controls.

Correspondence with symptoms
This meta-analysis shows that only about 15.59% of TCs are symptomatic. This is somewhat close to what has been suggested in the past by Paulsen et al. who indicated that approximately only 1 in 5 TCs might produce clinically significant symptoms [19]. Clinical picture that may truly be related to the presence of TCs stems from neural compromise and most commonly consists of radicular pain in the relevant dermatomal distribution or motor weakness, rarely myelopathy if a TC is located above the conus medullaris [7,13,19]. Persistent genital arousal disorder is under debate as an entity that could be linked with TCs [11]. Nonspecific back pain was not attributed to TCs in the present meta-analysis.

Sacral segments
Sacral perineural cysts were the ones to be documented most consistently. It was a subject of controversy which sacral level is usually involved. Incidence of S2 TCs was declared highest in three studies [2,9,15]. S3 was the most prevalent in two papers, whereas S1 in one [12,14,25]. Pooling the data revealed that, in fact, S2 level is the one to be disturbed most frequently (46.7% [95% CI 29.4-60.5], Q = 49.12, I 2 = 85.7%, p < 0.001).

Tarlov cysts and persistent genital arousal disorder
Subgroup analysis reveals that patients with PGAD, mostly women, might be more prone to developing TCs. Two of three studies involving those patients gave an account of high TC incidence (50-66.67%) [1,17]. However, one study stated to the contrary (6.67%), indicating that the previously mentioned studies might have been biased [21]. Pooled prevalence for the time being is 37.87% Currently, more studies of larger samples, optimally of case-control design, would be appreciated to dispel doubts.

Implications, limitations, and future directions
Despite being comprehensive, this meta-analysis is not free from limitations. High methodological bias found in many papers is highlighted, primarily due to the fact that in many of them, measures were not taken to reduce intra-/interobserver variability. Small number of studies about PGAD is the reason of the wide confidence interval of incidence in this group. It, however, indicates that PGAD might be associated with TCs. Therefore, case-control studies would be of value for calculation of the risk ratio. Based on the findings of this study, it is suggested that more studies on incidence of TCs be conducted in South America, Africa, and Australia. Moreover, results of this meta-analysis and the emphasized gaps in knowledge can be used in future research on Tarlov's cysts. As all but one of the studies assessed only lumbosacral spine MRI and not whole-spine MRI, there exists some risk of bias regarding counting the vertebrae in individuals with transitional abnormalities. Since only a handful of papers delineated morphological features of the cysts such as size or shape, as for now, it is difficult to seek link between these characteristics and clinical picture. Especially, it is still problematic to determine quantitative relation between TCs and the most frequent complaints expressed by patients, which is radicular pain and local back pain. Moreover, there are not enough data to univocally answer the question whether spinal cord or cauda equina might herniate into the TCs. Hence, more precise depiction is encouraged in upcoming studies.

Conclusion
Spinal perineural Tarlov's cysts are found in a minority of population. Women tend to develop TCs more often than men. Perineural cysts in pediatrics seem to be rare findings. Incidence in Europe appears to be higher than in other continents. S2 level of the sacral bone is affected most frequently. Possibly high incidence in PGAD requires confirmation in case-control studies for the risk-ratio calculation. Studies with evidence stronger than level 3 are needed to corroborate the results of this meta-analysis.