Discontinuation risk from adverse events: immunotherapy alone vs. combined with chemotherapy: a systematic review and network meta-analysis

In our network meta-analysis evaluating first-line treatments for diverse solid tumors, we found that patients treated with Chemo-IO experienced a higher rate of treatment discontinuation due to TRAEs than those receiving mono-IO.

The advent of immunotherapy has ushered in a new era in cancer treatment. Starting with mono-IO, the field has since evolved to incorporate more complex regimens, such as IO-IO combinations and Chemo-IO, which have shown improved survival outcomes over Chemo-only or mono-IO [2, 6, 7, 14]. These advancements have led to approvals from the U.S. Food and Drug Administration (FDA) and the European Medicines Agency (EMA) for various cancer treatment settings [7, 8, 15, 16].

However, due to the additive nature of adverse effects, Chemo-IO is generally associated with a higher incidence of TRAEs compared to mono-IO [8, 9]. Consistent with these findings, our study observed a significantly increased rate of TRAE-induced discontinuation in the Chemo-IO group compared to the mono-IO group across various cancer types. Similarly, among NSCLC patients, Chemo-IO also demonstrated an elevated rate of discontinuation due to TRAEs compared with mono-IO.

The impact of treatment discontinuation due to AEs on clinical outcomes remains a subject of debate. Some studies have observed significantly poorer outcomes in the discontinuation group compared to the non-discontinuation group [17‐19], while some studies have found no significant difference in clinical outcomes between groups that discontinued treatment due to AEs and those that did not [20‐22], Although treatment discontinuation may occur because of AEs, several factors—including the timing of AEs, the patient’s response prior to discontinuation, the severity of the AEs or other variables—could be associated with clinical outcomes after discontinuation [17, 19, 23]. For instance, one study found that early AEs, occurring before 12 weeks and often leading to discontinuation, were associated with worse survival outcomes than later AEs [17]. Another study demonstrated that patients achieving complete response (CR) or partial response (PR) before the occurrence of the first irAE showed no difference in clinical outcomes between those who discontinued treatment and those who did not or those who resumed treatment. Conversely, patients who did not achieve CR or PR before the first irAE experienced significantly poorer outcomes in the discontinuation group compared to the non-discontinuation or re-treatment group [19]. Furthermore, mild irAEs might be indicative of a favorable response to immunotherapy, while severe irAEs (grade ≥ 3), often life-threatening or leading to treatment discontinuation, may not be indicative of favorable clinical outcomes [23]. TRAEs that occur during IO or Chemo-IO treatment could be linked to life-threatening conditions or mortality, leading physicians to generally consider them as potentially detrimental. In this study, we primarily suggested that IO-chemotherapy combination treatments are associated with a high rate of discontinuation. And there is a need for additional studies to investigate the association between AE- or TRAE-induced discontinuation and clinical outcomes.

Furthermore, the optimal duration for IO treatment continues to be a topic of ongoing debate, several trials, including KEYNOTE-189 [24], KEYNOTE-010 [25], and CheckMate-153 [26], suggest the possibility of poor survival outcomes or disease progression after discontinuing IO treatment following one or two years of administration [27]. Limited research has managed to follow patients who discontinue treatment due to AEs. However, one retrospective study that examined the clinical outcomes of patients who discontinued IO treatment due to immune-related AEs found that 20% of patients experienced disease progression within six months of discontinuation and 10% of patients died [28]. This evidence further emphasizes the potential impact of treatment discontinuation on patient outcomes.

In susceptible populations, such as elderly patients or those with a poor Eastern Cooperative Oncology Group performance status (ECOG PS), the administration of Chemo-IO could potentially be associated with a poorer prognosis due to an increased incidence of severe TRAEs and subsequent treatment discontinuation owing to these AEs [29]. Moreover, even with Chemo-only regimens, certain patient subgroups such as those of advanced age, poor performance status, individuals with anemia, impaired renal function, hearing impairment, or history of falls, are known to have an increased risk of toxicity [30]. This heightened toxicity could potentially lead to treatment discontinuation in these subgroups. However, enrolling these specific subgroups in oncological trials poses challenges, leaving unresolved questions and concerns about the extent of harm caused by treatment discontinuation in these populations.

For high-risk patients predicted to cease treatment due to TRAEs, a treatment strategy focusing on mono-IO might offer a lower discontinuation rate, thus improving treatment continuity. Furthermore, considering that several studies have reported comparable effects of IO to Chemo-IO in certain subgroups, such as patients aged 75 or over, or those who have received prior treatments [16, 31], there is a pressing need for further research into potential biomarkers or clinical factors that can aid in identifying patients who would benefit sufficiently from mono-IO instead of Chemo-IO, especially since these populations may experience higher rates of treatment discontinuation with Chemo-IO.

Several predictive biomarkers, including PD-L1, tumor mutation burden (TMB), and tumor-infiltrating lymphocytes (TILs), have been proposed to predict the efficacy of IO, but their roles remain inconclusive [3, 4, 8, 32]. Consequently, novel approaches are being explored to improve prediction accuracy, including the application of artificial intelligence (AI). Some studies have even reported that AI-assisted methods, such as the evaluation of pretreatment contrast-enhanced CT images, PD-L1 expression, or the spatial analysis of TIL, can yield better predictions of survival outcomes in NSCLC patients undergoing IO treatment [33‐35]. Advancing this line of research might not only refine predictions of IO efficacy but also help in anticipating and managing treatment discontinuation, thus facilitating more personalized treatment strategies and potentially improving patient outcomes.