Over the last 5 years, anti-programmed cell death protein (PD)-1 or anti-programmed death-ligand (L)-1 immunotherapy has proven to be highly effective in the treatment of various types of cancer. By releasing the immune brake on antitumor activity, immunotherapy can also trigger immune-related adverse events (irAEs) [1]. Checkpoint blockade by an anti-PD-1 or anti-PD-L1 will induce an irAE in around 40% of patients, and 8% of patients will experience a severe (grade 3 or 4) irAE [2, 3]. There are two main types of irAE: the first, most frequently observed type is an immune-mediated inflammation that in principle can affect any organ (the thyroid, lungs, skin, digestive tract, etc.) [1], and the second corresponds to a flare-up of a pre-existing autoimmune disorder [4–6].

Paraneoplastic syndromes (PNSs) are rare autoimmune disorders associated with cancer [7, 8]. The complex pathogenesis of these syndromes is mediated by either soluble factors (such as hormones or cytokines secreted by the tumor) or cellular immune mechanisms that target tumor cells displaying cross-reactivity with self antigens [8–10]. To date, little is known about the use of immune checkpoint inhibitor in patients with a PNS, although some recent data suggest that these syndromes are complications of immunotherapy [10–13]. It is noteworthy that PNSs are generally associated with specific tumor types that are not routinely treated with immune checkpoint inhibitors; these include small-cell lung cancers [14], gynecological cancers [15], and thymic carcinomas [16]. A recent review of neurologic PNSs that occurred following treatment with immune checkpoint inhibitors emphasized the clinical seriousness of these syndromes and the need for further investigation in the context of immunotherapy [11]. Here, we performed a French, nationwide, observational, multicenter study of patients with PNS having undergone anti-PD-1 or anti-PD-L1 immunotherapy.

To the best of our knowledge, the present study is the first to have described the tolerability of immunotherapy in patients with a pre-existing or newly diagnosed PNS. We studied patients with neurologic (34%), rheumatologic (28%) and connective tissue PNSs (25%). Half of patients with a pre-existing PNS experienced a worsening of the corresponding symptoms after the initiation of anti-PD-1 or anti-PD-L1 immunotherapy. Our observations also highlighted the seriousness of PNSs (especially neurologic PNSs), since four of the 32 patients (13%) died (paraneoplastic encephalitis: n = 3; Lambert Eaton syndrome: n = 1).

For the study population as a whole (i.e. cohorts 1 and 2), the overall response rate of 47% was relatively high. Although our study’s descriptive, retrospective design prevents us from drawing robust conclusions about responses rates, this high response rate emphasizes that the exacerbation or appearance of a PNS can be associated with an effective tumor response soon after the initiation of immunotherapy.

Patients with a PNS accounted for 2.53% of all the requests addressed to the ImmunoTOX committee; hence, these syndromes are rare but are likely be encountered in routine clinical practice. The relatively low prevalence might be explained by the current indications for immunotherapy; the tumor types most frequently associated with PNSs (such as small-cell lung cancers, gynecological cancers and thymic tumors) are not generally treated with immune checkpoint inhibitors. Furthermore, PNS are sometime difficult to diagnose and so their prevalence might be underestimated in routine practice [8, 19]. Graus et al. recently stated that the prevalence of neurologic PNS has probably been underestimated as a result of (i) the difficulty of diagnosing these conditions and (ii) the possible underreporting of neurologic irAEs [11]. A recent study of 216 patients with recurrent small-cell lung cancer treated with nivolumab (alone or in combination with ipilimumab) found that four (2%) patients had experienced neurologic irAEs, although the researchers did not state whether or not these events were associated with paraneoplastic features [14]. These data indicate that the neurologic safety of immunotherapy in patients with small lung cancer - a tumor potentially associated with neurologic PNSs - needs to be characterized in more detail.

Our present results showed that PNSs worsened or appeared quite soon after the initiation of immunotherapy; the median time interval was 1.4 months. This finding suggests that patients at risk of a PNS should be closely monitored during the initial immunotherapy cycles. In the present study, we also looked at the causal relationship between anti-PD-1 or anti-PD-L1 immunotherapy and the accentuation or appearance of a PNS. We found that 25% of the patients with a pre-existing PNS and 38% of the patients with a newly diagnosed PNS had obtained an objective tumor response – showing clearly that the PNS was associated with immunotherapy and not with tumor progression.

Neurologic syndromes were the most common and severe PNSs observed in the present study. These neurologic PNSs were of particular concern because of their severity; this might not have been apparent in individual studies or clinical trials but was revealed by our large-scale survey. With regard to the pathogenesis, most studies to date have found that neurologic PNSs have an immune etiology; cross-reactivity occurs when the immune system is misled by the unconventional, ectopic expression of neural proteins on tumor cells [19, 20]. We observed six patients with paraneoplastic encephalitis, and four of the latter were positive for anti-Ma2 autoantibodies. All six patients with encephalitis had severe disease, and three died as a result. The cases of encephalitis were generally difficult to treat, since all patients received various immunomodulatory or immunosuppressive treatments in addition to corticosteroids. Our present data suggest that paraneoplastic encephalitis may be a life-threatening complication of immunotherapy. Importantly, four of the six patients with encephalitis had an objective tumor response at the time when the PNS worsened or appeared, and the other two had stable disease. Hence, the encephalitis was probably triggered by immunotherapy and not by tumor progression. We therefore consider that in clinical practice, (i) all suspected cases of paraneoplastic encephalitis should be investigated extensively (including a screen for anti-neuron antibodies in the serum and cerebrospinal fluid), and (ii) immunotherapy must be discontinued as soon as signs of encephalitis are suspected.

Hypertrophic osteoarthropathy was the second most frequent type of PNS observed in our study; it was generally characterized by digital clubbing, periostitis (often affecting the tibia) and joint pain [21]. In four of the six cases in our study, hypertrophic osteoarthropathy was diagnosed after the initiation of immunotherapy. The condition was mild or moderately severe, and had a limited impact on the patients’ subsequent cancer immunotherapy, which was temporarily discontinued in three cases, and not discontinued in the other three cases. Most of the cases of hypertrophic osteoarthropathy responded well to steroids or nonsteroidal anti-inflammatory drugs, and none of these patients had painful rheumatologic symptoms at the time of analysis. It was not possible to formally establish a causal relationship between immunotherapy and hypertrophic osteoarthropathy, since none of the patients obtained an objective tumor response; hence, the PNS could have been related to tumor progression as well as to immunotherapy.

Dermatomyositis was the third most frequently reported type of PNS; this is a well characterized PNS, particularly when it is associated with anti-TIF1 antibodies [22, 23]. In the present study, we observed five cases of paraneoplastic dermatomyositis. Only one of the five patient with dermatomyositis was seropositive for anti-TIF1 autoantibodies. The paraneoplastic dermatomyositis was severe in all five cases, and had a significant impact on patient care; three of the five patients with dermatomyositis had to permanently discontinue immunotherapy. The cases of dermatomyositis recorded in our study difficult to treat, since four of the five patients required immunomodulatory or immunosuppressive treatments. It is noteworthy that two of the patients with dermatomyositis had an objective antitumor response, indicating that immunotherapy may have had a triggering effect in these individuals.

Overall, one can question the causal relationship between PNSs and immunotherapy. We believe that most of the cases observed in the present study were driven by the tumor, and that immunotherapy merely exacerbated or revealed the clinical symptoms. In cohort 2, the PNSs were detected after the initiation of anti-PD1 or anti-PD-L1 immunotherapy; this prompted us to describe them being incident or newly diagnosed. Given the retrospective nature of the study, however, it is not possible to say whether the PNSs in cohort 2 were present but latent (i.e. clinically asymptomatic) before the initiation of immunotherapy or whether they developed afterwards.

The study’s main limitations were the small number of patients in the final sample and the inherent sources of bias associated with the retrospective, descriptive design. The main bias was selection bias, since patient recruitment was declarative; hence, the most severe cases might have been over-represented in this setting. Consequently, we cannot draw any firm conclusions about the patient distribution. However, given the rarity of PNSs and the often challenging diagnosis, we consider that a declarative study is currently the best way to obtain detailed, useful information. Another limitation was related to data interpretation; it was difficult to distinguishing between a PNS on one hand and a conventional irAE on the other, as these entities can have several common signs and symptoms. To address this issue, we prespecified a list of the most frequent types of PNS (Additional file 1: Table S1) and centrally reviewed all cases of PNS.

Our present results showed that PNSs tend to be worsened or revealed by anti-PD-1 or anti-PD-L1 immunotherapy. Up to half of patients with a pre-existing PNS might experience a worsening of their symptoms following immunotherapy. Paraneoplastic encephalitis emerged as a potentially life-threatening complication of treatment with immune checkpoint inhibitors. When initiating immunotherapy, physicians should carefully monitor patients with pre-existing PNSs.