Here, we showed that grades 2b and 2a (invasive) PAs were associated with poorer treatment-related clinical outcomes and required more often complex therapeutical regimens, including multiple and multimodal treatments. Grade 2b and 2a PAs in comparison to the non-invasive grades 1b and 1a had significantly higher rates of persistent remnant within 1-year after operation (93 and 78% vs 18 and 30%) and active disease at last follow-up (40 and 27% vs 12 and 10%), as well as required more often re-operation (27 and 16% vs 0 and 5%), radiotherapy (53 and 38% vs 12 and 7%), multimodal treatment (67 and 49 vs 18 and 25%), multiple treatment (33 and 27% vs 6 and 9%), as well as a higher number of treatments (2.6 and 2.1 vs 1.2 and 1.4). These findings support the usefulness of Trouillas’ classification in predicting PAs that are more likely to be refractory to conventional treatments and may require more often multimodal and multiple treatment approaches.
Previous studies confirmed the prognostic value of this classification in predicting PA recurrence/progression [8, 12, 14, 15], with grade 2b PAs having considerably higher risk of recurrence/progression than grade 1a tumors, ranging from 3.72–4.8 [11, 15] up to 25-fold [8]. In a study including only NFPA patients, the risk of recurrence was 8.6 times higher in grade 2b tumors than those of grade 1a [14]. However, only one recent study assessed more in detail the impact of Trouillas’ classification in terms of treatment strategies, beyond its usefulness to predict recurrence/progression [15]. In this study involving 607 patients with PAs, 114 required an additional first-line adjuvant therapy, 17 had a second-line therapy, and 5 received multiple treatment. Conventional radiotherapy, in particular, was given to 11 patients, all of them having invasive PAs (grade 2a and 2b tumors) [15]. Consistently, in our series, radiotherapy was predominantly given to patients with invasive PAs (78.6% of the irradiated patients had a grade 2b or 2a tumor), which means that an invasive PA has a likelihood 3–7 times greater of needing radiotherapy than the non-invasive grades 1a and 1b counterparts. Sahakian et al. also reported an overall risk of needing multiple treatment as 9.5-fold higher in grade b cases as compared to grade a tumors. However, worths mentioning that in this study, out of the 5 patients with proliferative grade b PAs who had multiple treatment, four (80%) were grade 2b and only one case corresponded to a grade 1b tumor [15], suggesting that proliferation, but also invasion, may be critical for refractoriness and multi-step treatment approach in patients with PAs. In our study, proliferative and invasive (grade 2b) PAs more frequently needed a multiple treatment approach (3.6x more comparing to grade 1a PAs), and when directly compared to grade 1a tumors, grade 2b PAs required 2-times more treatments, and had higher rates of multimodal treatment and re-operation. Additionally, our study showed that the rates of persistent remnant within 1-year after operation and active disease at last follow-up were 3–4 times higher in grade 2b PAs than in grade 1a tumors. Taken our data together with the findings reported by Sahakian and coworkers, grade 2b PAs are more refractory and difficult-to-treat exposing the patient to more complex multiple and multimodal treatment regimens, which may justify a more thight monitoring protocol, and eventually, a lower threshold to administer precocious adjuvant therapy to grade 2b PAs.
Published studies assessing Trouillas’s classification have been trying to dissect which one of the processes –invasion or proliferation- is the most critical for tumor recurrence/progression. In two studies proliferative tumors (grades 1b and 2b) were independent predictors of recurrence/progression, suggesting that proliferation was possibly more relevant than invasion in determining poorer outcomes [12, 15]. In contrast, in another study, a more prominent prognostic role for invasion than proliferation alone was suggested [11], which align with the findings from our study, as we found poorer treatment-related outcomes in grade 2a (invasive, non-proliferative) PAs than in those classified as grade 1b (non-invasive, proliferative). Such discrepancies across the published studies may be related to several factors, such as the heterogeneity of the PAs, as well as the differences in the study populations, follow-up durations and in PA management approaches across the studies. Nevertheless, it may be irrelevant to discuss as to whether proliferation or invasion alone are more critical for recurrence and poorer outcomes in PA patients. Instead, as proposed in Trouillas’ classification [8], both these criteria combined (possibly also integrating other factors such as large tumor volume, high-risk histopathological subtype or residual tumor after operation [13, 25]) are much more useful and valuable in predicting recurrence/progression and poorer treatment-related outcomes [11, 12, 14, 15].
Invasion has been recognized as a critical factor for surgical failure in resecting the entire pituitary tumor, and an important determinant of recurrent disease in PA patients [2, 26, 27]. In two series reporting invasion as less important than proliferation, the criteria followed to consider a PA invasive relied on both MRI and histological demonstration of cavernous sinus and/or dura matter invasion [12, 15]. This approach, contemplated in the Trouillas’ classification [8], may result in grading many PAs as invasive that otherwise would be considered non-invasive based on MRI alone, including tumors with histological microinvasion that might be biologically or clinically not relevant, and/or may be fully resected at the operation, which ultimately, may lead to an underestimation of the overall prognostic role of invasion. In contrast, we and others [11] defined invasion relying only on the radiological evidence of cavernous or sphenoid sinus invasion, an approach that may fail to identify cases with histological invasion not apparent on imaging, or misclassify cases with dubious radiological invasion. To overcome such issues and uniformize the performance of this prognostic clinicopathological classification, particularly for grade 2a PAs, perhaps it would worth distinguishing grade 2a tumors with microscopic invasion from those with gross radiological cavernous or sphenoid sinus invasion [15]. This approach could further refine the accuracy of the predictions based on this classification, and allow a more precise and personalized management and follow-up of PA patients. However, the histological invasion is often not routinely reported in the pathology reports [14], thus complicating its application in real-life clinical practice.
Ki-67, mitotic count and/or p53 expression are the most commonly used proliferation markers in PAs, but each one of these has important limitations, and divergence still exists concerning which one is the most relevant [2, 26]. Ki-67 has been regarded as probably the best single marker, however is insufficient to predict tumor behavior or recurrence/progression when considered in isolation [11, 27]. Mitotic activity is one of the most routinely used histopathologic method in evaluating tumor proliferation and aggressiveness, however its role in predicting recurrence/progression in PAs, as well as benign versus malignant disease, is very limited [28], while a reliable quantification method and definitive cut-offs for p53 expression predicting recurrence/progression in PAs are still lacking [5, 8, 11]. In fact, it is still unknown whether the p53 staining or the mitotic count provide any added value to the Ki-67 index alone. Trouillas’ classification defined proliferation as the presence of at least 2 of the 3 markers: i) Ki-67 ≥ 3% for formalin fixative (or > 1% for Bouin-Hollande fixative); ii) more than 2 mitosis per 10 high power fields; iii) positive p53 staining (> 10 strongly positive nuclei per 10 high power fields) [8]. However, in the original study, the predicted ROC curve analysis for recurrence/progression showed very similar results when Ki-67 ≥ 3% (or > 1% for Bouin-Hollande fixative) were considered alone or in combination with mitotic count and p53 expression [8], suggesting that the recurrence/progression predictive performance of this classification would be similar relying on the currently proposed criteria, or instead only on Ki-67 alone, for assessing the proliferative status. Moreover, in further studies, Ki-67 appeared to be the key proliferation marker, with Ki-67 ≥ 3% found in nearly all PAs graded as 1b and 2b, and rarely in non-proliferative 1a or 2a tumors; whereas, for instance, p53 positivity may be detected in only about a third of proliferative PAs [11, 14].
Contrarily to the published studies validating Trouillas’ classification [11, 12, 14, 15], the criteria to define proliferation in our study was only based on Ki-67 (with a PA being considered proliferative if Ki-67 ≥ 3%) because the mitotic count and p53 expression data were not routinely available. However, we also considered relevant to investigate the usefulness of this classification when proliferation status would rely only on Ki-67 as this marker is routinely used by most centers, in contrast to mitotic count and/or p53 expression which are often unavailable, thus facilitating the implementation of Trouillas’ classification in many centers. Our approach in defining proliferation based only on Ki-67 may eventually overestimate grade b tumors, thus skewing the distribution of PAs across the different grades. However, in general, the proportion of each grade in our cohort (1a: 45.7%; 1b: 13.2%; 2a: 29.5%; 2b: 11.6%) was similar to other studies, with grade 1a PAs usually representing most cases (ranging from 32–67%), followed by grade 2a PAs (varying from 14.2–42%), whereas 1b tumors are typically the less common grade (4–15%) [8, 11–15]. This may also aid explaining, at least in part, the lack of significant differences we observed between grade 1a and 1b PAs concerning different treatment-related clinical outcomes.
Our subanalysis focused only on NFPAs identified high rates of tumor remnant within 1-year post-operatively in invasive grade 2b and 2a tumors (80 and 87.5%, respectively) in comparison to the non-invasive grades 1b and 1a (9.1 and 35.1%). Such rates are similar to those reported by Lelotte et al. in a series of 120 operated NFPA patients, where postoperative residues were noted in 82 and 22% of invasive and non-invasive tumors, respectively [14]. Almost 40% of patients with grade 2b and 2a NFPAs required multimodal treatment, including radiotherapy, a percentage four and eight-times greater than those observed for patients with tumors graded as 1b (9%) and 1a (5%). Grade 2a NFPAs also required a higher number of treatments than grade 1a tumors, further strengthening the notion that cavernous sinus invasion is critical for the refractoriness and recurrence of NFPAs, which is not surprising as cavernous sinus invasion is a key determinant for incomplete surgical resection of PAs [2, 26, 27]. In the acromegaly subgroup, patients with invasive tumors (grade 2a and 2b) required about twice as much treatments than grade 1a or 1b tumors, and medical therapy after operation was needed in 80% and 100% of acromegaly patients with grade 2a and 2b PAs, in contrast with 28.6% and 33.3% for patients with grade 1a and 2b tumors, respectively. We also observed noteworthy trends for grade 2a and 2b cases requiring more often multiple treatment and re-operation, as well as having active disease at last follow-up, but no statistical significance was reached due to the low number of cases, particularly for grades 2b (n = 4) and 1b (n = 3). Currently, there are no studies validating this classification specifically on acromegaly patients, however in the original study was reported a modest prognostic contribution of invasion alone in GH-secreting PAs (OR = 1.54), while both invasion and proliferation (grade 2b) was found of utmost relevance in predicting recurrence/progression (OR = 5.99 [1.97; 18.26]) [8].
Our study has some limitations related to the study design and population. It is a retrospective monocentric study conducted in Portugal, with a relatively low number of cases comparing to other series [12, 13, 15], which limited the statistical subanalysis, particularly for functioning PAs. Also, our study population includes mainly Portuguese patients, which limits the generalization of our findings to other populations or ethnicities; however, our study shows a valuable prognostic role of Trouillas’ classification when applied to a different population, in this case a Portuguese cohort, hence adding further knowledge on this classification from the other previous national cohorts from France [8, 11, 15], Italy [12, 13], and Belgium [14]. Moreover, to the best of our knowledge, our study is the second assessing this classification from the angle of treatment-related clinical outcomes, specifically looking into the performance of this classification in anticipating PAs that are more refractory and may need a more complex multiple and multimodal treatment approaches, beyond its value for predicting recurrence/progression already well-established from previous studies [8, 11–15]. Another limitation of our study is related to the criteria we used for proliferation, discussed in detail above, which can simultaneously be regarded as a novel element and a strength of our study, as we reported a remarkable usefulness of this classification relying only on Ki-67 as proliferative marker, allowing a more broad and universal application of Trouillas’ classification in real-life clinical practice as in many centers p53 expression or the exact mitotic count is not routinely available. Other important strength of our study is the long mean follow-up of 93 months (almost 8 years), similar to the follow-up duration of the original study [8], but much longer than the mean follow-up from all the subsequent studies, which ranged from 3.5 up to 5.8 years [11–15].
In conclusion, we confirmed that the Trouillas’ clinicopathological classification is a suitable grading system to identify individual PAs that may be more refractory to conventional treatment, particularly surgery, and might require more complex multimodal and multiple therapeutical approaches. Invasive pituitary tumors, especially those categorized as grade 2b (invasive and proliferative), are those more likely to have postoperative residues, and more likely to require a multimodal treatment approach, including radiotherapy, as well as may display higher rates of active disease at last follow-up despite the higher mean number of treatments offered to grade 2b PA patients during the disease course.