Biological toxicities as surrogate markers of efficacy in patients treated with mTOR inhibitors for metastatic renal cell carcinoma

Angiogenesis plays an important role in the invasion and dissemination of renal cell carcinoma (RCC), and dysregulation of genes encoding for the vascular endothelial growth factor (VEGF) and its receptors (VEGF-R) is the mainstay of this process [1]. Molecules targeting VEGF and VEGF-R have been developed. Tyrosine kinase inhibitors (TKI) (sunitinib, sorafenib, axitinib, cabozantinib) and monoclonal antibodies (bevacizumab) are routinely used as first- or second-line therapy [2–6].

The mammalian target of rapamycin (mTOR) pathway is downstream of the phosphoinositide 3-kinase and AKT pathway. Increased levels of phospho-AKT have been observed in RCC [7]. Therefore, the inhibition of the mTOR pathway can potentially inhibit the tumor progression. In phase III controlled trials, temsirolimus and everolimus have shown antitumoral efficacy in patients with mRCC. Temsirolimus was compared in first-line treatment with interferon-α, showing a statistically significant benefit in favor of temsirolimus for progression-free survival (PFS) and overall survival (OS) [8]. In a phase III trial against placebo (RECORD-1), everolimus was evaluated in patients with progressive disease after TKI failure resulting in a clear benefit in PFS, but not in OS [9]. Those results led to an approved use of temsirolimus and everolimus in mRCC. All of these targeted therapies are responsible for toxicities. Some of them are common for all of antiangiogenic therapy such as fatigue, anorexia, diarrhea and anemia whereas other adverse events are class or agent specific [3–9]. With mTORi, the most frequent clinical adverse events reported in phase III trials were skin rash, nausea, mucositis, and diarrhea [8, 9]. A class-specific pulmonary toxicity is observed in 35% of patients receiving an mTORi, going from cough to irreversible interstitial pneumonitis [10]. The most common metabolic toxicities were anemia, hypercholesterolemia, hypertriglyceridemia, increase in serum creatinine, and lymphopenia [8, 9].

A relationship between toxicities of antiangiogenics and their efficacy has been extensively studied. For instance, hypertension and hypothyroidism have been shown to be associated with a better disease outcome in patients receiving sunitinib or bevacizumab [11–17]. As regards mTORi, a relationship has been found between the occurrence of pneumonitis and a better clinical outcome for both everolimus and temsirolimus [18].

Some clinical toxicities have been shown to be correlated with efficacy of antiangiogenics and mTORi. For mTORi, some of these toxicities were directly related to the inhibition of the mTOR pathways, both in normal and tumoral tissue. Therefore, it has been proposed to investigate toxicity as a surrogate marker for mTORi efficacy [19]. The purpose of the present study was to evaluate retrospectively the relationship between metabolic toxicities and mTORi efficacy in mRCC patients.

To our knowledge, this is the first study analyzing retrospectively the relationship between metabolic toxicities and mTORi efficacy in patients with mRCC. Our findings showed that a CB was more frequent when liver transaminases and serum creatinine increased, and when absolute changes in glycemia, phosphatemia, and liver transaminases were higher. A short time to the highest lymphopenia grade and a long time to the highest hyperglycemia grade were correlated with a progressive disease. The onset rate of lymphopenia was faster in progressive patients compared with those having a CB. A longer PFS was associated with an absolute one-percent increase in glycemia and ASAT and an absolute one-percent decrease in phosphatemia, whereas an absolute one-percent decrease in lymphocytes was associated with a shorter PFS. Efficacy of mTORi in this series was better than expected maybe because of young age and good performance status (ECOG-PS) of patients.

As regards hypophosphatemia and increase in liver transaminases, the incidence rates of grade 3–4 toxicities reported in pivotal phase III trials were 4% and < 1%, respectively, which were close from our findings [8, 9]. In contrast, there were some discrepancies between our results and those published in the literature. The rates of grade 3–4 hyperglycemia and lymphopenia were higher in our everolimus subgroup than those described in the literature (13% vs. 4%; 20% vs. 3%, respectively) [9] whereas the rates of grade 3–4 hyperglycemia were lower in the temsirolimus subgroup than those reported in the literature (6% vs. 11%) [8]. These discrepancies may be due on the one hand to the small sample size, and on the other hand to a high proportion of patients (64%) who were treated in third-line or more.

Correlations between efficacy and toxicity were highlighted in our study by using two approaches: the classical CTC-AE developed by the National Cancer Institute, and the assessment of time to highest grade of toxicity and changes in the absolute rates, correlated with RECIST tumor response (RECIST 1.1 criteria) and PFS. The classical CTC-AE scale presents several limitations as, for instance, the assessment is clinician-based and the reporting is potentially incomplete [20]. Although our findings are of interest, one should keep in mind several limitations of the study. It was a retrospective study and, even if the data were prospectively recorded, there were several missing data. No correction for multiple testing has been made and relation with a p value close to 5% should be taken cautiously. Twelve patients withdrew treatment because of toxicity before efficacy assessment: this could be responsible for modification of estimation of relationship between efficacy and toxicity. Furthermore, the lack of significant relationship can be due to the small sample size (n = 75). Moreover, we included patients with 6 various RCC histological subtypes, knowing that these tumors have unequal sensitivity to mTORi inducing a bias in the efficacy assessment. Finally, patients received either everolimus or temsirolimus. If both drugs have close toxicities, their frequencies are not similar. This discrepancy could induce a bias in the estimate of toxicity rate. A final potential bias in this estimate was the line of therapy: some patients received mTORi as first line (n = 8), whereas others received mTORi in the salvage setting (n = 67).

The most common toxicities of routinely used antiangiogenics (sunitinib, sorafenib) are hypertension, fatigue, hand-foot syndrome, hypothyroidism, increased lipase, and lymphopenia [21]. Everolimus and temsirolimus present with a different toxicity profile, leading especially to frequent metabolic disorders such as hyperglycemia, hyperlipidemia, and hypophosphatemia [22]. Our choice was to not analyze the correlations between clinical toxicities and efficacy of mTORi, such as pneumonitis, because its incidence was low (around 14%) in previously published data for everolimus in the RECORD-1 trial [9]. Thereby, the monitoring of metabolic toxicities seemed more relevant, because of their frequency and more reproducible assessment methods.

In a series where patients with mRCC were treated with temsirolimus or interferon, treatment benefit of temsirolimus over interferon was associated with increase in serum cholesterol [23]. We did not find this relationship in our population. Several metabolic toxicities can predict efficacy of targeted therapies and should be explored, such as proteinuria for VEGF pathways inhibitors, hyperglycemia or hyperlipidemia for mTORi [24]. We hypothesized that metabolic toxicity could be related to the efficacy because the occurrence of toxicity can reflect a higher drug exposure. An interindividual pharmacokinetic variability has been described in patients receiving everolimus in case of hepatic impairment [25]. It has been also shown that the Japanese patients presented a modified pharmacokinetics of temsirolimus, allowing a maximum tolerated dose of 15 mg/m² once a week instead of 25 mg/m² [26]. Besides, intra-patient dose titration is routinely performed with axitinib in order to increase axitinib exposure in patients who do not develop hypertension. Indeed, it has been shown that higher drug exposure and diastolic blood pressure were independently associated with longer PFS and OS, and with a higher probability of partial response in mRCC patients [27]. This tends to demonstrate that diastolic blood pressure could be a potential marker of efficacy. Finally, a well-described and reproducible assessment of metabolic toxicity might be a useful biomarker to predict efficacy in the drug development process [28].

Our study suggested that biological toxicities of mTORi (hypophosphatemia, hyperglycemia, and increase in liver transaminases) were significantly correlated with mTORi efficacy and lymphopenia with lack of efficacy. Given this preliminary result, it seems difficult to use a single biological parameter or a single descriptive method to predict treatment efficacy on the basis of toxicity. The relationship between a greater number of toxicities and clinical benefit should be investigated but we did not assess these relationships because of the sample size in our series.

Our retrospective study suggested that hypophosphatemia, hyperglycemia, and increase in liver transaminases were significantly correlated with mTORi efficacy and lymphopenia with lack of efficacy. On the other hand, hypercholesterolemia and hypertriglyceridemia had no relationship with antitumoral efficacy of mTORi.

We noticed that the time to severe toxicity was short with a median time of 28 to 56 days, highlighting that toxicities occurred earlier than the first tumor assessment (84 days). These interesting results require to be prospectively confirmed on a larger and independent cohort of patients.