Recently, despite the advancement in micro-neurosurgical techniques and neuroimaging modalities to improve preoperative planning and postoperative care, CP recurrence has remained a well acknowledged conundrum in many treatment centers. There is vast variability in the definition of CP recurrence and regrowth throughout the literature due to the differing methods of assessing and defining recurrence but also variance in the growth potential of the tumor in different individuals. In this paper, we define CP recurrence as evidence of remnant tumor observed on postoperative image studies of the follow-up period despite undergoing previous gross total resection (GTR) or subtotal resection (STR) according to the operating neurosurgeon; or despite an initial negative control postoperative follow-up MRI study [5] . In addition, the term regrowth refers to the growth of a known tumor remnant following intentional STR or partial resection in follow-up imaging studies [6] .

Age at the time of treatment is also an important factor, with higher morbidity being assumed more in younger patients than adults, although a number of studies have reported conflicting results with no consensus its association with recurrence rate (RR). According to the literature, exclusive pediatric series have reported recurrence rates ranging from 9% to 24.5% [17] [18] [19] [20] [21] Whereas some authors have noted young age [17] [19] as a recurrence risk factor, others adult age [22] , most cohorts have not found a significant difference in the recurrence rate between children and adults [4] [6] [16] [23] [24] [25] [26] . Similarly, role of sex as a predictive factor is still under debate; inasmuch as some studies found a significant correlation between the rate of CP recurrence and the male sex [6] [27] , others reported no association with sex whatsoever [16] [19] [25] [26] [28] .

The biological behavior of CPs varies from patient to patient in that while some remain stable even up to 30 years before symptomatic recurrence [29] , others may grow rapidly within an erratic period of time [2] It has been reported that most recurrences appear during the first 5 years following treatment [6] [30] . Similar to primary CPs, the clinical symptoms of RCPs are inconstant, on account of the variable topographical location [31] . CPs of suprasellar region commonly present with headache, visual field defects, and endocrine dysfunctions [32] . A few studies have reported visual symptoms at presentation [24] , symptoms of intracranial hypertension at presentation [27] and severe hydrocephalus [17] as significant factors associated with recurrence. However, other cohorts reported no significant differences between patients with primary CP and RCP in the neurological, endocrinological, visual, or functional symptoms [20] [23] [33] nor the presence of hydrocephalus [19] [25] .

An issue of specific interest is the relevance of tumor morphology in CP recurrence to which current literature bears inconsistent findings. Tumor size is likely to be a predictive factor because increased recurrence rates have been shown in tumors with a diameter larger than 3 cm [5] [17] [20] [34] [35] ; other studies did not support these findings [27] [36] . Aside from tumor size, other physical attributes such as tumor adherence [37] , intrasellar location [28] , third ventricle remnants [38] and cystic tumors [34] [39] have been proposed to be clinical predictors of recurrence. In other studies, none of tumor brain invasion [21] [34] , tumor extension [19] , tumor consistency [26] nor tumor location [19] [22] [25] [35] and presence of xanthogranulomatous tissue [16] , were found be of significant importance in tumor recurrence. Clinically, the presence of calcifications could suggest a possible involvement with CP recurrence given the fact that tight adherences to neurostructures have often led to incomplete resection even in some the most skilled hands [40] . On the contrary, Elliot et al. in a cohort of eighty-six pediatric patients showed that the absence or presence of minimal residual calcification does not have an impact on the risk of CP recurrence after GTR [41] .

Ohmori et al. in a pediatric series of twenty-seven pediatric patients surgically treated for CPs observed a significant low recurrence rate for patients in whom early postoperative MRI reveals complete CP removal [42] , in agreement with Mortini’s findings [21] . On the other hand, Eldevik et al. [43] studied the radiologic and histologic characteristics of CPs and could not identify any imaging characteristics of tumors that corresponded to a high or low rate of recurrence, as supported by other studies. [23] [44]

Over the past 30 years, studies into the histopathologic nature of CPs have thrown more light into understanding its pathogenesis. Still, the genetic and molecular basis of the recurrence CP is yet to be well elucidated. However, advances in immunohistochemical studies and direct genetic sequencing have attempted to exemplify the molecular pathogenesis of CP initiation, growth, and recurrence [3] . There are two subtypes of CPs, adamantinomatous (ACP) and papillary (PCP) that may differ clinically and histologically but also share overlapping characteristics. Histologically, ACPs contain nodules of wet keratin, regressive changes like fibrosis, calcifications, old hemorrhages, cholesterol deposits; a mixture of cystic and solid portions, whorl-like structures, ghost cells, intense surrounding gliosis, and profuse rosenthal fiber formation [1] [45] and nucleo-cytoplasmic accumulation of b-catenin [45] [46] . On the other hand, the PCP subtype is distinctly characterized by a solid, papillary growth pattern missing the wet keratin, cystic appearance and regressive elements of the ACP subtype [44] .

At the molecular level, extensive studies of whole-exome sequencing data of craniopharyngiomas have revealed that the ACP and PCP subtypes have distinct molecular underpinnings, each driven characteristically by mutual exclusivity of mutations in a single well-established oncogene: CTNNB1 (P-catenin gene) in the ACP form and BRAF in the papillary form as a result of activation of the MAPK pathway [47] [48] [49] . Tena-Suck et al. observed that the presence of whorl-like arrays was associated with recurrence/regrowth of CP [16] which could likely be caused by mutations of the P-catenin gene. Interestingly, several studies have reported no significant differences in recurrence rate and between adamantinomatous type and papillary type [6] [21] [24] [26] [34] [50] . Whilst no definitive tumor markers that predict recurrence have not been identified, several recent studies (Table 1) which have attempted to shed more light on some important prognostic factors that may correlate with CP recurrence have shown that RCPs displayed a significantly higher expression of these postulated biomarkers than primary tumors.

Table 1 summarizes some of the potential biomarkers of CP recurrence.

Other studies have supported the role of Ep-CAM overexpression in tumor cell proliferation and recurrence in other brain tumors [58] [59] . Conversely, others studied molecular proteins which have been associated with recurrence but have not been found significant include laminin 8 and BCL-2 [51] . Apart from its likely involvement in maintaining the proliferative activity of tumoral cells, increased p53 expression has been purported as a marker of malignancy transformation [60] . Of particular note, Prieto et al. [5] and Tena-Suck [16] reported a positive association of p53 expression with CP recurrence, though other studies obviate these findings [60] [61] .

The use of proliferation-associated antigen Ki-67 and mitotic index for the histological evaluation of different tumors has been widely reported [5] [18] [62] . On the contrary, other studies found no correlation between Ki-67 and risk of recurrence of CPs [22] [28] [63] . Duo et al. [64] proposed that the very low level of MIB-1-LI in residual tumor remnants negates their role in initiation of recurrence. In disagreement, Hussain et al. [3] purported that rapid recurrence may be a function of the immune response in the tumor microenvironment. Lefranc and colleagues [61] attributed the adhesiveness of CPs to the proximity tissues such as optical chiasm and pituitary stalk to interactions between vitronectin, CP-alpha (2beta1) integrin and collagens, proteins expressed by peritumoral tissue. Furthermore, they demonstrated that galectin-4 expression was associated with a significant delay in recurrence. In another study, Xu and colleagues found that minichromosome maintenance protein 6 (MCM6) label

index (LI) was significantly higher in primary CPs than in RCPs [65] . Whether angiogenesis-related factors play a part in CP recurrence is still a matter of contention. In line with this observation, Sun et al. compared the expression of angiogenesis-related factors in 4 recurrent and 6 nonrecurrent tumors, they found that expression of PDGFR-alpha and FGF-2 were significantly higher for recurrent tumors (P = 0.02 and P = 0.01), thus suggesting that selective PDGFR-alpha blockers may offer a novel therapeutic option for CP treatment [66] , although they no significant correlation with VEGF, in parallel with Xu et al. findings [65] , but in disagreement, Agozzino [22] found a significantly higher VEGF expression in recurrences than in primary CP.

Despite their reportedly benign nature, a few cases of malignant transformation of primary CP diagnosis on recurrence [67] [68] [69] have been reported but the exact pathogenesis and the biological behavior of malignant change in recurrent CP are not well elucidated in although a possible causative association with radiation therapy has been postulated [68] [70] . Ectopic recurrence of CP after primary surgical management is no longer a rare phenomenon as several reports have been documented [71] [72] [73] [74] with a widely accepted pathophysiological hypothesis which evolves around tumor seeding occurs either along the surgical route, or at a distal location in the subarachnoid space [71] [73] .

The optimal management of recurrent craniopharyngioma (RCP) is still disputable [75] . However, several authors have stressed that radical surgery should be the first therapeutic option for RCP with acceptable morbidity and mortality risks similar to that of primary tumors [6] [18] [20] [76] [77] [78] . Conversely, others have reiterated that secondary surgery for RCP is associated with a low cure rate and a high risk of complications [19] [37] [75] . In addition, recurrence rates following radical tumor resection vary widely and have been reported as 9% to 65% [19] [21] [23] while others noted that the incidence of recurrence did not differ significantly with radicality at surgery [4] . In general, several authors have echoed that surgical treatment for RCP is considered more challenging than primary surgery due to scarring, absence of gliosis reaction in RCP, tight adherences and morphological changes due to irradiation [2] [37] [75] . Most published literature has consistently reported residual tumor and the extent of resection as the strongest predictive factor for CP recurrence [6] [16] [17] [20] [25] [34] [37] [63] . In contrast, other studies have downplayed the role of residual CP as a predictive of recurrence [4] [40] [64] .

Accordingly, due to the variable location of CPs or RCPs in proximity with neurostructures, some surgeons have advocated for intentional subtotal or partial removal for reasons not limited to; hypothalamic adherence of CP which is associated with postoperative morbidity and mortality [36] ; an inadequate view of the tumor due to large tumor extensions and inappropriate surgical approach; presence of major calcifications [37] which poses a risk of damage to the hypothalamus and the optic chiasm [37] [77] ; firm adherence of the tumor capsule to relatively large perforating vessels or to large basal arteries as it is safer to leave minor residual portions of tumor in anticipation of vascular damage [32] [79] . Prieto demonstrated that in an effort to preserve the hypothalamus, the CP remnants may lead to erratic tumor recurrence [38] , whereas studies by Li et al. [80] mentioned otherwise. It has been noted significant number of patients with residual tumors remain stable for a long time [36] . Similarly, Hoffman et al. recommended that radical surgery is not an appropriate treatment strategy in patients with hypothalamic involvement [45] . In a recent study, Clark et al. reviewed a total of 109 studies describing the extent of resection resulting in a cohort of 531CP patients and 377 recurrences. They found no difference in the progression-free survival (PFS) of 1 year (89% vs 84%) or 5-year (77% vs 73%) between the groups who underwent GTR and STR combined with radiation. Their results suggest that STR + XRT of pediatric CP is associated with similar rates of tumor control as GTR [7] . These findings were similar to those of Schoenfeld et al. [26] and Yang [81] and colleagues. What remains unclear is whether the preservation of pituitary stalk at resection represents a recurrence prognostic factor because some studies have demonstrated a correlation with recurrence [82] [83] whereas other studies reported no significant association with CP recurrence [33] [83] . In some population of CP patients, although Growth hormone replacement therapy (GHRT) may pose a concern for stimulating tumor regrowth, several studies have demonstrated that long-term GHRT is not associated with risk of CP recurrence [84] [85] [86] .

Throughout literature, CPs and RCPs have been widely reported to present in the variable locations [31] . On these grounds, several surgical approaches in recurrent CP series including transsphenoidal, subfrontal, trans-lamina terminalis, subtemporal and transcallosal approach [2] [6] [75] [78] [87] [88] have been advocated with the effort to reduce morbidity and mortality risks, although the optimum surgical approach is still debatable. In a recent publication, Morisako et al. attempted to devise an anatomical subclassification of CPs for achieving aggressive surgery [89] . In a similar vein, Prieto et al. [5] stressed that evasion of hypothalamic injury should be the primary goal in any surgical planning, including the choice of approach and preplanned extent of removal. A few authors have strongly recommended the use of endonasal route [2] [90] for recurrent tumors especially patients in whom the initial surgery was transcranial, asserting the advantage it offers of facing the tumor upon dural opening without brain retraction while optimizing visualization of the relevant anatomy via a direct surgical trajectory.

The role of adjuvant radiotherapy (RT) in the management of RCP is well documented in literature with numerous studies showing lower recurrence rates after GTR or STR [2] [8] [18] [26] [43] . In contrast, other studies have reported that the incidence of recurrence did not differ significantly with respect to postoperative radiotherapy [4] [16] [91] . In a review of pediatric CP surgical series, Tomita et al. demonstrated that RT was effective for recurrent tumors and should be considered being the primary treatment for recurrences or difficult tumors, which are not amenable to total resections [19] . The optimum timing of RT for RCP is still controversial. Fisher et al. observed that short times to recurrence may result from the tendency to delay radiotherapy [40] . Additionally, Elliot et al. the authors observed that withholding irradiation or other adjuvant therapy in the setting of minimal residual calcification without enhancing tumor [41] .

The efficacy of GKS as a valuable adjuvant treatment modality for RCP has been well documented [11] [91] [92] despite its limitation to small size tumors (<3 cm). A study by Park et al. [9] showed that STR followed by GKS results in a lower recurrence rate than neuroendoscopy-GKS and GKS alone group much as the latter provided better preservation of endocrine function. Others studies have proposed CP histological subtype [93] intratumoral homogeneous irradiation using multiple isocenters [94] , distance from the tumor to the optic nerves, and tumor radiosensitivity [10] as prognostic factors that favor successful treatment outcome in RCP. In a large cohort of one hundred CP cases, Kobayashi et al. [95] demonstrated that tumor diameter and radiation dose were the only the significant prognostic factors related to recurrence of CP.

1) Intracavitary Brachytherapy: Recent studies investigating the efficacy of intracavitary brachytherapy have shown that inasmuch as it may offer a treatment option for cystic CPs, its use is limited by the fact that it does not limit growth of solid tumor parts or deter formation of new CP cysts [12] [96] .

2) Intracystic Bleomycin Therapy: Hader et al. [97] reported significant tumor control with use of intratumoral bleomycin for cystic CPs, while avoiding potentially harmful surgical or radiotherapy related risks in the pediatric population. Reported complications related with this treatment include bleomycin leakage and toxicity, vasculopathy [98] and fatal toxic effects [99] . However, importantly, Zhang [14] and colleagues argued the rational use of intracystic bleomycin due to lack of RCTs, quasi-randomised trials or CCTs of the treatment of cystic CPs in children.

3) Pegylated interferon-a-2b: Yet another systematic conventional therapy [100] reportedly used in treatment of cystic recurrent CPs and some studies have shown its efficacy in children [101] , however, its toxicity and optimum dosage obviate its importance and suggestively large series are necessary to establish its long term benefit.