Introduction
Immune checkpoint inhibitors (ICPIs) have radically changed the history of a wide spectrum of different neoplasms, including both solid and hematologic tumors [
1‐
3].
Indeed, the recent history of non-small cell lung cancer (NSCLC) treatment has seen a dramatic turn with the introduction of ICPIs as well [
4]. These agents exert their activity by blocking inhibitory signaling and enhancing T-cell activity against tumor cells. Moreover, these results have been obtained with tolerable toxicity profile, gaining them approval for numerous different types of cancer and paving the way for designing new trials in rarer ones [
5,
6]. The first ICPI approved in the treatment of advanced NSCLC is nivolumab, an inhibitor of programmed cell death protein 1 (PD-1) [
7]. The signaling cascade following the binding between PD-1 and its ligands (PD-L1 and PD-L2) is responsible for the inhibition of positive signaling, proliferation, and survival of peripheral tissue activated T cell, in order to prevent autoimmunity during inflammatory-mediated immune response. The same pathway is engaged by tumor cells to dampen early activation of the immune system and forestall tumor rejection [
8].
Results from clinical trials with nivolumab in previously treated NSCLC have demonstrated an increase in median overall survival (OS) and overall response rate (ORR) in patients with pretreated non-squamous and squamous NSCLC, as well as an increase in median progression-free survival (PFS) in squamous histology [
9,
10]. Nivolumab, together with the other ICPIs, shows great differences in term of frequency, types and grade of toxicities compared to chemotherapy. Due to their particular action on the immune system, these drugs present a significantly lower grade of hematologic toxicities, nausea, neuropathy, fatigue, and asthenia; however, a peculiar spectrum of adverse events (AEs) may be recorded. Indeed, their mechanism of action might result in impaired self-tolerance with subsequent development of immune-related AEs (irAEs). Previous experiences with ipilimumab (a CTLA-4 inhibitor) in advanced melanoma showed potential unconventional endocrine toxicities, and other data have been collected about nivolumab, although within ICPIs class some differences have been noticed [
11,
12]. Compared with other irAEs caused by immune checkpoint blockade, endocrinopathies are unique because these manifestations are often irreversible. While hypophysitis is related to ipilimumab therapy with an estimated frequency around 11%, thyroid disfunction is the most frequent endocrine irAE reported during nivolumab treatment (hypothyroidism in 7% of cases) [
13]. As far as nivolumab, results from Checkmate 057 and 017 trials, which explored the efficacy of nivolumab versus docetaxel in pretreated non-squamous and squamous NSCLC, reported up to 7 and 4% cases of hypothyroidism, respectively [
9,
10].
Some recent evidence showed an improved overall survival in patients treated with ICPIs developing irAEs, suggesting that the onset of an autoimmune disease may be marker of a more potent immune activation consequent to the treatment, including the tumor-directed immune response [
14‐
19].
In the last 5 years, after the arrival of nivolumab in pretreated NSCLC patients, new anti-PD-1 drugs have been added to the armamentarium of oncologists. Nowadays, all patients receive an ICPI in the first or in the second line of treatment. This prospective study was born at the dawn of the era of immunotherapy in NSCLC. The aim of this study was to prospectively evaluate the endocrinological toxicity of treatment with Nivolumab, the first of the available PD-1 antibodies, in the treatment of NSCLC. The long follow-up (> 60 months) allowed us to investigate the connection between toxicity and overall survival of patients in a real world setting.
Discussion
Pivotal studies with nivolumab in patients with non-squamous and squamous NSCLC reported, as endocrine AEs, only hypothyroidism in about 7% and 4% of patients, respectively [
9,
10]. In line with the published data on endocrine irAEs associated with nivolumab, in our cohort, thyroid function abnormalities were confirmed to be the major endocrine irAE [
21‐
25]. Conversely, in this cohort, we did not record any hypophysitis, confirming that this life-threatening event, in contrast with the evidence observed with CTLA-4 blockade, is rare during PD-1 inhibitors treatment [
22‐
24,
26].
Development of impaired thyroid function in the present study was reported in 23.4% of patients (
n = 15). Hypothyroidism seems to be the most frequent irAE and occurred in 17% of cases (
n = 11), often without a concomitant increase in thyroid antibodies. This incidence of hypothyroidism was superior when compared to that reported in the registration trials [
9,
10]. However, in these latter studies, the features of hypothyroidism and the presence of a previous ATT were not specified. The 6.2% of patients (
n = 4) developed a transient thyrotoxicosis, evolving to hypothyroidism on autoimmune basis in 75% of cases. Another 6.2% of patients (
n = 4) showed only a transient thyrotoxicosis with negative thyroid autoimmunity, which turned into euthyroidism without requiring any treatment. In these cases, since the ultrasound examination performed in this subgroup of patients revealed the presence of thyroid nodules or goiter, it is possible to postulate that even an excess of iodine intake due to contrast enhanced radiological exams (CT-scan every 4–8 weeks) could have played a role in the recorded transient thyroid abnormalities.
Considering the entire cohort of patients, no significant correlation was observed between functional thyroid abnormalities and ultrasound morphology.
Furthermore, even though the limited number of cases with previous and concomitant thyroid alterations in our trial (9/74), we did not record any effect of nivolumab on thyroid function in this subgroup of patients. Indeed, our results confirm that a preexisting thyroid alteration does not represent a major exclusion criterion for nivolumab therapy, if combined with appropriate management of thyroid disease, consistently with other publications [
27]. We preferred to include “prior thyroid disease” as exclusion criteria to reduce possible bias. This may be seen as a limitation of the study, but it is aimed to assess the impact of therapy on the occurrence of endocrine adverse events. We know that this reduces the possibility of extending these results to the entire population.
As regards HPA, data are confounded by concomitant therapies known to interfere with this hormonal assessment. However, in our cohort, we did not identify any case of isolated ACTH deficiency induced by autoimmune-related mechanism, as recently reported in few cases with metastatic malignant melanoma or other cancers treated with nivolumab [
28‐
34].
However, it is interesting to note that 32% of our patients presented with elevated IGF-I levels at baseline. This finding is in line with published data correlating high IGF-I serum levels to an increased risk of lung malignancy [
35‐
37], as well as colon and breast cancer [
38,
39], and showing that patients with lung cancer have significantly increased circulating IGF-I levels compared to healthy controls [
40]. On the contrary, reduced IGF-I levels at baseline in specific settings of patients can be explained by the presence of advanced metastatic disease, while the increase in IGF-I levels during therapy may reflect the improvement of metabolic conditions and performance status [
41‐
43]. Recent evidences in the literature, regarding other different neoplasms [
43‐
45], have shown that lower levels of IGF-I were predictive of worse outcome. According to these observations, even in our study, IGF-I low levels were associated to shorter OS compared with patients with normal/high IGF-I. In fact, survival curve of patients with reduced IGF-I at baseline, or displaying a reduction during follow-up, showed a significantly reduced median survival compared to those of patients with normal/high IGF-I levels (
p = 0.031). Therefore, low IGF-I at baseline, or a decrease of IGF-I levels during nivolumab treatment, could be considered as a negative prognostic factor in this setting of patients. We could speculate whether the reduction in IGF-I levels was secondary to treatment or disease progression and, due to the distribution of cases with low IGF-I at baseline and during therapy in our series, we could hypothesize that this phenomenon is more related to disease progression. Furthermore, it is known that IGF-I is an expression of the nutritional status and is related to the metabolic state of the patient; therefore, it could be considered as a surrogate of performance status (PS). In our series, interestingly, low IGF-I performed better than PS in predicting worst patient outcome, being significant in univariate (
p = 0.037) and at the limits of significance in multivariate (
p = 0.054; see Table
5). However, the size of our sample does not allow us to draw clear conclusions. Therefore, this observation needs to be further validated in a larger study cohort of patient with advanced NSCLC.
Finally, regarding the connection between OS and new onset of hypothyroidism as irAE, we observed that patients developing hypothyroidism showed a lower mortality rate as compared with non-developers. Median OS in the first subgroup of patients was 23.2 months, whereas in the second was 6.2 months (
p = 0.023). Univariate and multivariate analyses confirm these data [Fig.
2 and Table
5].
The onset of endocrine irAEs has already proved to be a favorable response predictor to ICPIs therapy in recent studies in different cancers [
18,
46‐
48], and such findings are even of older knowledge when regarding the treatment of melanoma with ICPIs, due probably to the greater experience acquired so far with tumor, for which ICPIs were first approved [
10,
11]. When considering monotherapy with nivolumab in lung cancer, most of the studies available in the literature are retrospective [
14,
17,
19,
49]. Many of them appear to confirm our findings: Baldini et al. and Ricciuti et al. [
19,
49] found that the irAEs-developing group showed a substantial difference in OS as compared to non-developers (
p < 0.0001 in both studies). Ricciuti et al. even observing that patients who developed ≥ 2 irAEs during treatment had a significantly longer median PFS and OS compared to those with one irAE. Campredon et al. [
50] and Sbardella et al. [
51] did not find statistical significance, still they observed a trend toward an improved OS in the group of patients developing AEs (
p = 0.069 and
p = 0.27, respectively), though the sample size was relatively small.
On the contrary, Ksienski et al. [
52] observed a decreased survival in patients developing severe AEs, though the considered patients had to suspend treatment at very early stages (landmarks are 6 and 12 weeks).
Among few prospective studies [
53‐
56], Kim’s group was the only one to specifically evaluate the association of thyroid dysfunction during PD-1 blockade with the treatment efficacy in patients with NSCLC. They observed a significantly longer OS for patients developing thyroid irAES as compared with the non-developers (
p = 0.025). However, Kim et al. in their prospective trial enrolled patients received PD-1 blockade (nivolumab or pembrolizumab) without making a distinction in outcomes between the two drugs [
57].
Our results, consistent with data of the vast majority of the cited papers, show significantly longer OS in patients who developed hypothyroidism as irAEs, suggesting that the dampening of PD-1 signaling pathway may be responsible for both endocrine toxicity and antitumor response, so that hypothyroidism may be seen as a marker of an enhanced therapeutic effect. Indeed, to the best of our knowledge, this represents the first prospective study proving a better OS in NSCLC patients developing hypothyroidism as irAES during treatment with nivolumab in monotherapy.
In conclusion, routine monitoring of thyroid function should be performed at baseline and after each cycle of nivolumab. Global pituitary function, including HPA axis, should be investigated and monitored in all case and, obviously, in case of the onset of signs and symptoms of hypophysitis or cortisol deficiency.
Reduced IGF-I levels in specific settings of patients can be explained by the presence of advanced disease and have shown to be predictive of worse outcome in our cohort.
Our findings regarding the connection between hypothyroidism onset and better OS seem to indicate that the development of hypothyroidism as irAE is a positive predictive biomarker of nivolumab antitumor efficacy in patients with NSCLC.
These results need to be confirmed by further studies. However, a close collaboration between oncologist and endocrinologist is of crucial importance to provide patients with the best care and avoid unnecessary therapy discontinuation.
Publisher's Note
Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.