Our results underlined that the experience of PNS can be categorized into three severity levels. Up to 40% of our sample suffered from moderate to severe PNS, but a more nuanced look reveals that oral pain and sleep problems were present in all three classes and may therefore be considered symptoms intrinsically related to HNC [40]. In turn, both the moderate and the severe classes stood out for their depression levels, while the moderate was especially affected by anxiety and the severe by fatigue.

Therefore, this study supports the high prevalence of PNS among HNC patients [1, 2]. Indeed, if we focus on previous LCA studies, Tometich et al. [8] featured a two-class model with only 16% of breast cancer patients after diagnosis assigned to the class affected by PNS (mainly represented by anxiety, depression, and fatigue). After treatment, Lewson et al. [15] included one class with no symptoms (52% of the patients), one class with mild anxiety and moderate fatigue (34%), and a class with mild to moderate levels in all PNS except pain (14%) for patients with diverse cancer diagnoses. Han and colleagues [16] distributed colorectal cancer patients in one class with low severity in all PNS (28%); one class with high depression and anxiety (25%); and another with high fatigue, sleep disturbance, and pain (47%). Finally, the three classes retrieved by Harris et al. [17] again in a sample of breast cancer women were labeled as lower (72%), moderate (24%), and higher (4%) PNS.

Our results align with these findings in retrieving three profiles ordered according to PNS severity, with most participants allocated to a class with either mild or no symptoms, and only a minority assigned to the most severe. The only exception is the study by Han et al. [16]. However, this could be explained by the addition of cognitive impairment as an inclusion criterion and later as an indicator variable in the LCA. As for specific symptoms, in these studies, pain and sleep problems were generally less frequent, which reinforces their specific role in HNC. On the contrary, the clear prominence of anxiety, depression, and fatigue supports their general usefulness for mental health screening in cancer.

Regarding our previous network analysis on PNS in HNC [14], poor sleep quality, fatigue, and (to a lesser extent) depression were the core symptoms. Our current study shows that these three symptoms are not experienced equally by all patients, with sleep problems being present in all classes, depression in the moderate and severe, and fatigue only in the severe. This finding may suggest a progressive activation of the network, with more severity associated with sleep problems development into other PNS.

Concerning the covariates analyzed, we found that being a woman, having oral cavity cancer (rather than larynx), worse performance in daily activities, history of depression or anxiety, daily smoking, higher CRP, and flatter cortisol slope were associated with more PNS. Extensive research has shown that women experience more PNS such as anxiety and depression after a cancer diagnosis [41], also for HNC specifically [13, 42]. In turn, pain is the main complaint among oral cavity cancers [40]. Thus, considering the psychosocial implications of this specific diagnosis is important, although the variance explained by tumor site might be shared with other covariates as it lost significance in the multivariable logistic regression. The same applies to worse performance, already associated with higher PNS in previous LCAs [8, 17], but for which mixed results are found in HNC literature [40, 42]. Such inconclusive findings can be reasonably argued to depend on multiple interactions, like tumor site itself [8], which invites to frame this variable in the context of each person.

In the regression, lifetime history of anxiety and depression increased the probability of being assigned to the moderate class, with no differences compared to the severe. This result can be interpreted as such previous experience acting as a proxy to react with anxiety and depression again after the cancer diagnosis [42]. The relevance of both variables suggests that considering previous mental health episodes in screening may be more efficient than asking about specific disorders. As for smoking, it is a widely known risk factor for developing HNC but also for mental health burden after diagnosis [41, 42]. It may be hypothesized that patients who were daily smokers engage in self-blame, leading to higher PNS [43]. However, attention should be given to smoking-related metabolic changes as well, since both daily smoking and flatter cortisol slope were attributable to the severe PNS class. This is aligned with previous literature reporting that smokers experience dysregulations in their HPA activity [44], this reflecting in cortisol alterations, an inflammatory response, and ultimately the experience of depression and other PNS [6, 14, 45]. Indeed, this pathway would also account for the CRP differences between classes.

It is important to highlight that these two biological markers, cortisol and CRP, do not only show a discriminant value between classes here but they were also included in the same cluster as PNS in our previous network analysis [14], underlining their relationship. In fact, they stand out despite the general inflammation observed in the sample: all classes showed mean CRP values > 5 mg/L, when the cut-off flagging inflammation is > 3 mg/L [46]. Therefore, both are proposed as useful PNS biomarkers in cancer. Since systemic inflammation is believed to have prognostic value in HNC [47], and considering that it has also been related to the pathogenesis of a wide range of cancers [48], the interaction between cancer, PNS, and inflammation requires further research.

This study has several limitations. Even if listwise deletion is considered appropriate for MCAR data, our sample may be biased due to the differences between patients included and excluded from analysis. Such differences suggest that our participants have less complexity compared to the original NET-QUBIC sample. Therefore, results need to be interpreted cautiously. Indeed, although this was a multicenter project, it was conducted in a single country, recruited participants diagnosed with a specific group of tumors, and some variables (e.g., ethnicity) were not considered. All these constitute limitations to the generalizability of our results. Additionally, the cross-sectional nature of the study provides an initial description of profiles. However, future longitudinal studies are necessary to assess their evolution over time. Building upon evidence in breast cancer [49], they can be hypothesized to remain stable or even worse during treatment, particularly the severe profile, but such anticipation needs to be tested specifically for HNC.