In this study, we constructed a network of physical symptoms, anxiety and depression symptom in older patients with advanced cancer by utilizing network analysis to examine data from a multicenter cross-sectional study in China. The results suggest that the stability of the centrality indices is reliable; notably, strength tends to be the most stable estimated centrality index in these networks, followed by closeness and betweenness. Overall, the network stability was relatively high. According to the observed network model, MDASI core symptoms and anxiety and depression symptoms were strongly correlated, with the latter two being more closely related. These findings also confirm previous findings [
24,
25] that anxiety and depression have a high level of coexistence. Our study visualized the complex interaction between anxiety and depression, and the results confirmed that there is a significant relationship between the burden of physical symptoms in cancer patients and anxiety and depression [
26,
27]. The majority of older cancer survivors are more likely to have additional risk factors for cancer and comorbidities [
28]. The results of this study showed that 64.9% of older patients with advanced cancer had one or more symptoms, and up to 80% had anxiety and depression symptoms. These rates are higher than the overall burden of symptoms in patients with advanced cancer that our previous study revealed [
14], indicating that the cancer burden of the oldest people is much greater than that of the other people and is worthy of attention. In particular, largely unaddressed comorbidities associated with cancer in China are common mental health disorders that are underrecognized and undertreated [
29].
Based on network theory [
7,
8], given their high centrality index scores, these symptoms may be targets for therapeutic interventions. In the whole network, ‘distress/feeling upset’ (MDASI 5) had the strongest edge connections and was the most important bridging symptom connecting different syndrome communities. Distress is common in cancer patients and survivors and may interfere with the ability to cope effectively with cancer, physical symptoms and treatment. The level of distress of patients varies according to age, sex, cancer site, treatment setting and disease progression [
30]. Distress screening is recommended in guidelines and is one of the focuses of psycho-oncology. Brief screening tools can be used to identify patients who are experiencing clinically important cancer-related distress. The most widely used of these is the Distress thermometer, for which the prevalence of distress varies between 39% and 60% [
31,
32]. A score of four or more on the Distress Thermometer is suggested to prompt further review of symptoms of anxiety or depression [
33]. ‘Disturbed sleep’ (MDASI 4) is the most closely associated physical symptom with ‘distress/feeling upset’. The prevalence of sleep disturbance in older adults with cancer was 40%, and this symptom was associated with daily living impairment and physical activity limitations [
34]. In addition, people with sleep disturbance are usually prone to comorbid anxiety and depression [
35,
36]. Sleep hygiene and cognitive behavioral therapy are currently recommended for patients with sleep problems [
37]. For all three centrality indices, ‘I lost interest in my appearance’ (HADS-D4) had the lowest scores. One way to understand these results is that females, who made up a larger portion of this study (71.1%), may have less awareness and concern about beauty and cosmetics [
38]. Identifying the driving symptoms in the symptom cluster is an equally important question in cancer symptom management. While our network analysis based on cross-sectional data does not demonstrate causality, the centrality indices of the network provide some insights into symptom clusters. For example, we found strong direct associations between ‘drowsiness’ (MDASI 9), ‘fatigue’ (MDASI 2) and ‘shortness of breath’ (MDASI 6). The nodes are on the edges of one symptom community network in Fig.
3. If our findings are confirmed in an independent sample, future research could explore causality and evaluate interventions to help clinical workers provide individualized symptom management strategies.
Several limitations need to be considered. First, this network reveals only partial correlations and does not define causal associations. Prospective clinical studies can be conducted to verify the results of this study in the future. Second, because the participants had six various types of cancer, the impact of different cancers on the burden of symptoms was ignored; for example, lung cancer patients were more likely to have shortness of breath. Third, the strength of certain associations needs to be understood cautiously because the CIs of some edges between symptoms are large, suggesting that additional samples are needed to elucidate the strength of these associations.