Pain, emotional distress, and fear of recurrence or progression in people with endometriosis: a network approach

This study was preregistered on Open Science Framework ( https://osf.io/qh6kz ). Ethical approval for this study was obtained from the University of Sydney Human Research Ethics Committee (2023/541; August 30, 2023). Participants were recruited in October 2023 through advertisements on Endometriosis Australia's Web site and social media platforms. Currently, endometriosis can only be diagnosed by an expensive and invasive surgical procedure (laparoscopy), and there is on average a 7-to-10-year diagnosis delay for endometriosis. 2 , 61 Hence, to adequately capture people across the endometriosis disease spectrum, we also recruited participants who had received a provisional diagnosis by a medical professional based on clinical information and/or the results of an ultrasound or MRI. Accordingly, participants were eligible to complete the study if they were at least 18 years old, located in Australia, fluent in English, and self-reported a confirmed, or provisional diagnosis of endometriosis given by a medical professional (where endometriosis is highly suspected, but not yet confirmed). Participants were excluded if they reported being younger than 18 years and did not have a diagnosis (confirmed or provisional) of endometriosis. Participants were also excluded from the study if they did not complete the whole survey. Demographic questions assessed participants' age, gender, country of birth, ethnicity, and level of education attained. Participants also completed questions regarding the year, type, and method of their endometriosis diagnosis. Diagnosis delay was calculated by subtracting the year each participant reported first experiencing endometriosis symptoms from the year that each participant reported being diagnosed with endometriosis. Participants were asked to select the symptoms that they experienced, from a list of common symptoms experienced by people with endometriosis sourced from the literature, 49 comprising symptoms compiled within an endometriosis service and endorsed by Pelvic Pain Foundation Australia, 15 and those identified through a scoping review of clinical interviews. 37 Participants also had the option to include additional symptoms through a free text “other” box. The severity of these symptoms was rated by participants on a visual analogue scale (VAS; 0: none—10: severe). Participants indicated whether they were currently using hormonal medications to manage their endometriosis (eg, oral contraceptive pill, intrauterine device, GnRH analogues). Furthermore, participants were asked whether they underwent any surgeries to manage their endometriosis, and if so, how many surgeries they received and the year of their most recent surgery. Participants were asked to indicate whether they experienced endometriosis-related pain. For those who selected yes, pain severity was assessed using 4 items from the Brief Pain Inventory (BPI). 9 The original BPI was adapted to assess for levels of pain over the past month, rather than the past 24 hours to capture the cyclical nature of pain in endometriosis. Participants rated their current level of pain, average level of pain, and their least and worst level of pain over the last month using a VAS: 0 (no pain)—10 (pain as bad as you can imagine). As per Cleeland and Ryan's 9 guidelines, pain severity was calculated using the average of these 4 pain measures, which produced a score ranging from 0 to 10, with higher scores indicating greater pain severity. In this study, internal consistency of the BPI was good (α = 0.87). Pain-related interference in those who indicated experiencing endometriosis-related pain was assessed through the Pelvic Pain Impact Questionnaire (PPIQ). 8 Participants rated the extent to which their endometriosis-related pain impacted aspects of their daily lives, such as their mood, energy levels, and sleep quality over the past month using a 5-point Likert scale; 0: not at all to 4: a great deal. Participants ratings on each of these items were summed to obtain a total pain interference score, with higher scores indicating greater pain interference. Internal consistency of the PPIQ in the current sample was excellent (α = 0.93). Health-related quality of life was assessed through the Endometriosis Health Profile Questionnaire (EHP-30) 22 (The EHP is copyrighted material that can be obtained by contacting [email protected] and no unauthorised use of the EHP is allowed without a license). Participants rated the extent to which their endometriosis had affected various areas of their life, such as their social well-being and mood using a 5-point Likert scale; 0: never to 4: always. Participant ratings on each of these items were summed to obtain a total quality of life score, with higher scores indicating worse health-related quality of life. Internal consistency of the EHP-30 in the current sample was excellent (α = 0.97). Interpretation bias was measured using the Word Association Task . 32 Participants were presented with 11 ambiguous words. They were then prompted to type the first word that comes to mind after reading the target ambiguous word. The ambiguous words all had a possible negative health-related interpretation (eg, “terminal” could be followed by “illness” or “train”). Participant's responses were coded as either neutral (0) or negative health/pain related (1), based on a coding scheme developed by the research group. 30 , 49 To mitigate bias, coding was conducted blind to the rest of the data. Responses that were not already in the coding scheme were double coded, and any inconsistent codes were discussed and resolved with a third coder. These new words were then added to the existing coding scheme. Participant's coded responses on each of these items were summed to obtain a total interpretation bias score, with higher scores indicating greater negative health-related interpretation bias. The Word Association Task has been successfully used in previous studies to show significant differences in interpretation biases between chronic pain and healthy samples. 30 , 52 In a recent meta-analysis, it was 1 of 2 most commonly used measures of interpretation bias and was associated with the largest effect size when comparing those with and without chronic pain (Hedge g = 0.899). 50 Participants also completed the Worries about Recurrence or Progression Scale (WARPS) 40 to assess their fear of endometriosis recurrence or progression. Participants rated the extent to which they feared that their endometriosis may recur or progress using a 5-point Likert scale: 1: strongly disagree to 5: strongly agree. Participants responses on each of these items were summed to obtain a total FORP score, with higher scores indicating greater levels of FORP. Internal consistency of the WARPS in the current sample was excellent (α = 0.94). The WARPS demonstrated good validity in previous samples with various chronic illnesses. 40 , 45 To determine the proportion of participants in the current sample who met criteria for clinically significant levels of FORP, we used clinical cutoffs for the WARPS questionnaire, as reported in Smith et al., 45 where 54 is the cutoff for moderate FORP and 65 is cutoff for severe FORP. Participants completed the Existential Concerns Questionnaire (ECQ) 53 to assess their general existential anxiety, death anxiety, and avoidance of existential cognitions. Participants rated the extent to which they worry about various existential concerns on a 5-point Likert scale; 1: never to 5: always. Participants responses on each of these items were summed to obtain a total existential concerns score, with higher scores indicating greater levels of existential concerns. Internal consistency for the ECQ in the current sample was excellent (α = 0.95). The ECQ has demonstrated sound validity in previous studies. 53 Participants completed the short-form Metacognitions Questionnaire (MCQ-30) 59 to assess their metacognitive beliefs, judgements, and monitoring tendencies. Participants rated the extent to which they agree with beliefs/statements about cognitions using a 4-point Likert scale; 1: do not agree to 4: agree very much. Participants responses on each of these items were summed to obtain a total metacognitions score, with higher scores indicating greater levels of unhelpful metacognitive beliefs. In this sample, internal consistency for the MCQ-30 was excellent (α = 0.93). The MCQ-30 has demonstrated sound validity in previous samples. 59 Participants completed the Generalised Anxiety Disorder Assessment (GAD-7) 47 to assess for the severity of anxiety-related symptoms. Participants rated the extent to which they experienced anxiety-related symptoms on a 4-point Likert scale; 0: not at all to 3: nearly every day. Participants responses on each of these items were summed to obtain a total anxiety symptom score, with greater scores indicating greater levels of anxious symptomology. Internal consistency for the GAD-7 in the current sample was good (α = 0.89). Participants also completed the Patient Health Questionnaire (PHQ-9) 23 to assess for the severity of depression-related symptoms. Participants rated the extent to which they experienced depression-related symptoms on a 4-point Likert scale: 0: not at all to 3: nearly every day. Participants responses on each of these items were summed to obtain a total depression symptom score, with higher scores indicating greater levels of depressive symptomology. Internal consistency for the PHQ-9 in the current sample was good (α = 0.86). Participants completed the 8 neuroticism items from the Big Five Inventory 17 to assess for trait neuroticism. A neuroticism measure was included to test whether trait neuroticism and fear of progression/recurrence were distinct constructs. Participants rated the extent to which they generally saw themselves as having features of neuroticism on a 5-point Likert scale: 1: disagree strongly to 5: agree strongly. Participants responses on each of these items were summed to obtain a total neuroticism score, with higher scores indicating greater levels of neuroticism. The study was completed online via Qualtrics (Provo, UT), an online survey platform. Participants who saw the online advertisement and wished to participate in the survey either scanned a QR code or followed a link to the survey. Next, participants read the Participant Information Statement and provided their consent to participate in the study. If participants consented to partake in the study, they were asked various demographics questions, followed by the stated questionnaires and interpretation bias task. Upon completion, participants were fully debriefed. The study took approximately 25 minutes to complete. Correlational analyses, descriptive statistics, and internal consistency estimates were calculated using IBM SPSS 29.0.2.0. 21 Bivariate correlations were conducted to explore the relationships between all key constructs (FORP, interpretation bias, existential concerns, metacognitive beliefs, pain severity, pain interference, quality of life, depression, and anxiety). An analysis was deemed statistically significant if P < 0.05. The magnitude of correlations was assessed as small (r < 0.30), moderate (0.30 < r 0.05). 10 Network analyses were conducted to explore dynamic, and interdependent relationships between key pain, emotional distress, and theoretical constructs. The benefits of network analysis are that it offers a powerful compliment to regression analysis by focusing on relationships and structure, rather than the impact of isolated variables. In doing so, network analysis allows us to capture complex patterns of relationships between key variables that are more representative of how psychological factors interact with one another and hence allows us to better understand how variables may cluster within a model. 4 As such, network analysis addresses questions related to how connections matter, rather than addressing potential causal relationships that typically oversimplify the nature of relationships. Three network analyses were conducted. To explore the first aim of the study—which was to investigate whether theoretical constructs from models of FORP were related to FORP in people with endometriosis—we conducted a network analysis with constructs FORP, existential concerns, interpretation bias, and metacognitive beliefs ( theoretical network ). We conducted 2 further network analyses to explore the second aim of our study—which was to understand the nature of relationships between FORP and key theoretical, emotional distress, and pain constructs. The pain and emotional distress network sought to confirm relationships between FORP, relevant pain (pain severity, pain impact, quality of life), and emotional distress (anxiety, depression) constructs in people with endometriosis. Finally, the combined network explored the associations between FORP and all theoretical, pain, and emotional distress constructs. All network analyses were conducted using the statistical software R 4.4.1, 48 using the qgraph (version 1.9) 13 and bootnet (version 1.5) packages. 14 Within the network analyses, partial correlations between 2 variables (also referred to as nodes) are represented by edges (a line connecting the 2 nodes). These edges demonstrate the unique strength of association between 2 nodes after controlling for all other nodes in the model. Positive correlations are represented as blue edges, and negative as red. The edges are undirected, meaning causality cannot be inferred. The strength of the partial correlation (ie, edge weight) is represented by the width and saturation of the edges, with stronger associations having wider, and more saturated edges. Highly connected nodes appear closer together, and those less connected appear further apart. The accuracy and stability of each network model was estimated using non-parametric bootstrapping with the bootnet (version 1.5) package. 14 To evaluate the accuracy and stability of the edge weights within each network, 95% confidence intervals were estimated for each edge weight in the network with 1000 iterations, in accordance with Epskamp et al. 14 guidelines. To measure the stability of the edge weights within each network, we calculated the strength correlation stability (CS) coefficient. The CS coefficient represents the maximum proportion of cases that could be excluded such that, in at least 95% of 1000 bootstrap iterations, the correlation between the original and resampled centrality indices remains above 0.70. 14 As per Epskamp et al. 14 guidelines, the recommended cutoff for CS coefficients is suggested to be at least 0.25 but preferably over 0.50. To reduce the likelihood of finding spurious correlations, a regularisation method called the Least Absolute Shrinkage and Selection Operator (LASSO) 17 was used. The LASSO algorithm reduces small, spurious associations to zero. Consequently, the resulting network is sparser, and it captures only the more robust relationships. Given network analytical approaches are relatively new in the field of psychology, there is no clear consensus on how power should be calculated. 14 Generally, larger sample sizes allow for more stable and accurate networks to be estimated. 20 Hence, the current study aimed to recruit at least 250 participants, with recruitment left open for approximately 2 weeks. Additionally, we conducted a post hoc multiple regression analysis to predict FORP using existential concerns, metacognitive and interpretation biases simultaneously.

A total of 317 participants completed the study, of which 98.7% of participants identified as female, and 1.3% identified as nonbinary or other. Of these participants, 86.8% were born in Australia, the average age was 32.65 years ( SD = 8.06), and 60.87% of participants reported completing tertiary education. In our sample, 59.6% reported that their ethnicity was Oceanian, 35.6% North-West European, 10.4% South-East European, 1.3% North African and Middle Eastern, 2.2% South-East Asian, 0.63% North-East Asian, 2.6% Southern or Central Asian, 2.2% Peoples of the Americas, and 0.1% sub-Saharan African. Most participants were employed full time (48.6%), part time (22.4%), or casually (11.4%). Also, 6.3% of all participants reported being business owners, 14.5% were studying, and 18.0% were volunteering, retired, a homemaker, looking for work, not looking for work, or other. Most participants reported having a confirmed diagnosis of endometriosis through laparoscopic surgery (86.4%). Some participants reported having confirmed or provisional diagnosis based on ultrasound (33.1%), pelvic examination (17.0%), magnetic resonance imaging (MRI) (3.2%), and other (8.2%). The most common endometriosis diagnosis was deeply infiltrating endometriosis (46.4%), followed by a confirmed diagnosis but unsure of which type (30.0%), abdominal wall endometriosis (24.9%), superficial peritoneal endometriosis (21.5%), endometrioma (19.9%), other (13.9%), and diagnosis suspected by a medical professional but not confirmed (3.5%). The average diagnosis delay was 10.1 years ( SD = 7.13). Participants endorsed a range of symptoms associated with endometriosis, with the most common being abdominal pain (96.2% of the sample). Most participants reported taking hormonal medication to manage their endometriosis (58.4%), and 88.0% of participants reported receiving surgery for their endometriosis. The median number of surgeries was 2 (mean = 2.44), with 38 participants receiving 0 surgeries, 251 participants receiving between 1 to 4 surgeries, 18 participants receiving between 5 to 9 surgeries, and 10 participants receiving 10 or more surgeries for their endometriosis. Based on clinical cutoffs for the WARPS questionnaire in Smith et al. 45 study, 93.5% of participants demonstrated scores consistent with moderate levels of FORP, and 84.9% had scores consistent with severe levels of FORP. As per expectations, greater levels of FORP were significantly associated with greater levels of existential concerns and unhelpful metacognitions ( P 0.05) in people with endometriosis. As per expectations, greater levels of FORP were significantly associated with greater levels of pain interference, pain severity, depression, anxiety, and lower health-related quality of life ( P < 0.01), with correlations of moderate to large size. Table 1 provides bivariate correlations between variables, and descriptive statistics (means, standard deviations, and ranges) for each of the questionnaire measures. Correlations between fear of disease recurrence or progression, pain, and emotional distress constructs in endometriosis sample. N = 317, * P < 0.05, ** P < 0.01. Anx, anxiety; Dep, depression; EC, existential concerns; FORP, fear of disease recurrence or progression; IB, interpretation bias; MC, metacognitive beliefs; PI, pain interference; PS, pain severity; QOL, quality of life. Fear of disease recurrence or progression demonstrated a moderate correlation with neuroticism ( r = 0.34). Accordingly, FORP and neuroticism are related but distinct constructs. We also tested for multicollinearity of key constructs through regression models, which indicated low multicollinearity (VIF < 5). A network analysis was conducted to explore partial correlations between FORP and key constructs from models of FORP (interpretation bias, existential concerns, metacognitive beliefs). Of 6 possible edges, 4 were unequal to zero and visible in the network. Fear of disease recurrence or progression was positively associated with interpretation bias ( r = 0.05), unhelpful metacognitive beliefs ( r = 0.07), and existential concerns ( r = 0.26). Existential concerns were also positively associated with unhelpful metacognitive beliefs ( r = 0.51). Figure 1 provides visual representation of partial correlations within the network. Notes: Positive correlations are represented as blue edges, and negative as red. The strength of the partial correlation is represented by the width and saturation of the edges, with stronger associations having wider, and more saturated edges. Highly connected nodes appear closer together, and those less connected appear further apart. Network model demonstrating partial correlations between ORP, interpretation bias, existential concerns, and metacognitive beliefs. IB, interpretation bias; EC, existential concerns; FORP, fear of disease recurrence or progression; MC, metacognitive beliefs. The strength CS coefficient calculated for network 1 was 0.75, which is indicative of a robust and stable network model as per Epskamp et al. 14 guidelines. Accordingly, robust inferences can be made from the overall network (see Figures A1 and A2, http://links.lww.com/PAIN/C419 in the appendices for bootstrapped confidence intervals, and stability graphs for Network 1). To understand the relevant contribution of each of the 3 theoretical variables to FORP, we conducted a post hoc multiple regression analysis to predict FORP using existential concerns, metacognitions, and interpretation biases simultaneously. These constructs contributed 15% to the variance in FORP. The results confirmed that it was existential concerns (β = 0.301, P < 0.001), rather than metacognitions (β = 0.113, P = 0.083) or interpretation bias (β = 0.079, P = 0.130) that contributed independently to FORP. A network analysis was conducted to explore partial correlations between FORP, key pain-related variables (pain interference, pain severity, quality of life), and emotional distress (anxiety, depression) constructs. Of 15 possible edges, 10 were unequal to zero and visible in this network. Fear of disease recurrence or progression was positively associated with anxiety ( r = 0.20) and quality of life ( r = 0.26), and negatively associated with pain severity ( r = −0.07). Quality of life was positively associated with pain interference ( r = 0.44), pain severity ( r = 0.33), anxiety ( r = 0.03), and depression ( r = 0.21). Pain interference was positively associated with pain severity ( r = 0.34) and depression ( r = 0.13). Depression was positively associated with anxiety ( r = 0.54). Figure 2 provides visual representation of partial correlations within the network. Notes: Positive correlations are represented as blue edges, and negative as red. The strength of the partial correlation is represented by the width and saturation of the edges, with stronger associations having wider, and more saturated edges. Highly connected nodes appear closer together, and those less connected appear further apart. Network model demonstrating partial correlations between ORP, pain, and mental health constructs. Anx, anxiety; Dep, depression; FORP, fear of disease recurrence or progression; PI, pain interference; PS, pain severity; QOL, quality of life. The strength CS coefficient calculated for network 2 was 0.75, which is indicative of a robust and stable network model as per Epskamp et al. 14 guidelines. Accordingly, robust inferences can be made from the network (see Figures A3 and A4, http://links.lww.com/PAIN/C419 in the appendices for bootstrapped confidence intervals, and stability graphs for Network 2). A network analysis was conducted to explore partial correlations between all key constructs within a model. Of 36 possible edges, 21 were unequal to zero and visible in the network. Fear of disease recurrence or progression was positively associated with existential concerns ( r = 0.08), interpretation bias ( r = 0.01), pain interference ( r = 0.10), anxiety ( r = 0.14), and worse quality of life ( r = 0.22). After controlling for all other variables in the model, FORP did not demonstrate a direct relationship with metacognitive beliefs, depression, or pain severity in the network model. Within the network, the following variables demonstrated partial correlations of moderate effect sizes: metacognitive concerns and anxiety ( r = 0.32); depression and anxiety ( r = 0.33); health-related quality of life and pain interference ( r = 0.43); pain interference and pain severity ( r = 0.33); pain severity and health-related quality of life ( r = 0.31). The following variables demonstrated partial correlations of small effect sizes: metacognitive beliefs and existential concerns ( r = 0.26); existential concerns and depression ( r = 0.19); existential concerns and anxiety ( r = 0.22); depression- and health-related quality of life ( r = 0.20); and depression and pain interference ( r = 0.13). Figure 3 provides visual representation of partial correlations within the network. Notes: Positive correlations are represented as blue edges, and negative as red. The strength of the partial correlation is represented by the width and saturation of the edges, with stronger associations having wider, and more saturated edges. Highly connected nodes appear closer together, and those less connected appear further apart. Network model demonstrating partial correlations between ORP, pain, theoretical, and mental health constructs. Anx, anxiety; Dep, depression; EC, existential concerns; FORP, fear of disease recurrence or progression; IB, interpretation bias; MC, metacognitive beliefs; PI, pain interference; PS, pain severity; QOL, quality of life. The strength CS coefficient calculated for network 3 was 0.75, which is indicative of a robust and stable network model as per Epskamp et al. 14 guidelines. Accordingly, robust inferences can be made from the overall network (see Figures A5 and A6, http://links.lww.com/PAIN/C419 in the appendices for bootstrapped confidence intervals, and stability graphs for Network 3).

The first aim was to identify the prevalence of elevated levels of FORP in people with endometriosis. Most of our sample (94%) had moderate levels of FORP, whereas 85% were in the severe range. These rates of severe FORP are even higher than those observed in people with cancer. 33 Our second aim was to determine the relevance of 3 theoretical constructs implicated in fear of cancer recurrence. In bivariate correlations, FORP was positively associated with existential concerns and unhelpful meta-cognitions but not interpretation bias. However, in our theoretical network , FORP demonstrated a strong independent relationship with existential concerns, and small independent relationships with both metacognitive beliefs and interpretation bias after controlling for other theoretical constructs. Post hoc analyses confirmed that only existential concerns contributed independent variance in FORP. Our final aim was to understand the nature of relationships between FORP, theoretical, emotional distress, and pain-related constructs. In our pain and emotional distress network , greater FORP demonstrated strong relationships with poorer quality of life, anxiety-related symptoms, and pain outcomes, emphasising the importance of FORP. Interestingly, visual inspection of the combined network showed 2 distinct but interrelated symptom clusters: pain interference–related variables (pain impact, pain severity, and quality of life), and emotional distress–related variables (anxiety, depression, existential concerns, and metacognitive beliefs) in people with endometriosis. Both these symptom clusters were strongly related to FORP. Alarmingly, 85% of individuals in our sample scored above the clinical threshold for severe FORP. Furthermore, consistent with Todd et al., 49 people with greater FORP reported greater pain experiences and anxious/depressive symptomology. Our combined network analysis revealed 2 interrelated but distinct clusters, representing pain impact and emotional distress. Fear of disease recurrence or progression demonstrated moderate relationships with both clusters. Taken together, these results strongly support that FORP is an important factor in our understanding of pain and well-being in people with endometriosis because it is common and strongly associated with poor emotional well-being, quality of life, and physical functioning, as has been shown in other chronic illnesses. 41 Across all analyses, existential concerns demonstrated the strongest relationship with FORP. Post hoc multiple regression analyses confirmed existential concerns to be the only theoretical variable that added independent variance to FORP. Our findings suggest that people with endometriosis who experience greater worries about death, their identity, and the meaning of life also experience greater worry that their endometriosis will progress or recur and bring them closer to these negative outcomes. People who experienced greater existential concerns also rated themselves as being more impacted by their pain, having worse quality of life, and greater emotional distress. What is particularly interesting about this finding is that unlike cancer, endometriosis is typically not life threatening. However, endometriosis is a chronic disease without any known cure. 61 Furthermore, endometriosis has been linked with infertility, fatigue, insomnia, decreased participation in social activities due to pain, disruptions to education and career, and sexual difficulties. 6 , 11 , 12 , 26 , 29 , 31 Accordingly, individuals with endometriosis may experience greater uncertainty about the trajectory of their disease and heightened concerns about their future, identity, and ability to engage in activities due to the debilitating nature of endometriosis symptoms. Previous studies have similarly reported FORP to be strongly related to existential concerns in other non–life-threatening illnesses—eg, rheumatoid arthritis. 42 The role of other theoretical constructs was less clear. In correlational analyses and the theoretical network , metacognitive beliefs demonstrated a relationship with FORP, pain, and emotional distress. By contrast, FORP demonstrated no association with FORP in our combined network model; however, this is likely due to high correlations between existential concerns and metacognitive beliefs. Our study generally suggests that individuals who have unhelpful beliefs about their worry (eg, worrying about their endometriosis progression helps them be more prepared for potential health outcomes, or worrying itself is harmful) also experienced greater pain and emotional distress, and it adds evidence to suggest a relationship between metacognitive beliefs and FORP in people with endometriosis. This is consistent with other studies in chronic pain which have shown unhelpful metacognitive beliefs to be associated with higher experiences of pain and poor psychological functioning. 35 , 57 , 60 Evidence to suggest a relationship between FORP and interpretation bias is mixed. In our theoretical network , and combined network , FORP demonstrated a small relationship with interpretation bias; however bivariate correlations showed no significant correlation between the constructs. However, interpretation bias has been found to differ between those with and without clinically significant FORP in people with rheumatoid arthritis, 5 diabetes, 25 and cancer. 33 , 34 Because almost all the participants in this study reported severe worries about recurrence or progression, the range may have been restricted for both FORP and interpretation bias. Also, consistent with Pickup et al. 30 study, an association between interpretation bias and pain impact was replicated. Taken together, interpretation bias appears to make a small contribution to our understanding pain and well-being for people with endometriosis. However, contrary to the model of Heathcote and Eccleston 19 of FORP in cancer populations, it is not clear that interpretation bias is related to FORP and instead may be directly associated with worse pain and psychological functioning in people with endometriosis. A limitation of the current study is its cross-sectional nature, meaning the direction of influence between variables cannot be assumed. However, through our network analyses provided useful foundations for testing how key constructs may be related to each other. Longitudinal and intervention studies can better disentangle the direction of relationships between these constructs over time and better capture the nature of relationships between pain and other constructs, given that pain experiences may fluctuate in people with endometriosis. 16 The correlation between some variables (eg, pain interference and quality of life, r = 0.81) was high; however, variance inflation factor (VIF) scores were below 5, and thus, multicollinearity was unlikely to be an issue. Additionally, given our sample was recruited through Endometriosis Australia (an endometriosis-specific organisation), people without pain, or extreme pain, may be less likely to participate in this research. Thus, our sample can best be considered a convenience sample. Finally, we chose the Word Association Task because it has been used to successfully differentiate between people with and without pain conditions 30 , 52 ; however, responses on this may be confounded by factors such as lexical factors and prior experience. There are other more sophisticated approaches, such as the use of ambiguous scenarios 18 , 38 ; however, they require the generation of disorder-specific scenarios. Furthermore, in a recent meta-analysis, the word association task gave rise the largest interpretation biases, despite its lack of ecological validity. 38 There is a lack of high-quality interventions for addressing pain in people with endometriosis, 54 and these results are important because they identify potentially modifiable targets that could be considered for intervention. First, there is evidence that existential concerns can be modified through psychological intervention. A meta-analysis of interventions for death anxiety found that cognitive behaviour therapy was efficacious in reducing death anxiety. 24 Although evidence for psychological interventions of death anxiety in people with cancer was less strong, there are interventions such as meaning-making therapy that show promise. 46 Hence, managing existential concerns could be worthwhile pursuing for people with endometriosis. Likewise, metacognitive beliefs are another variable that can be targeted through intervention, such as metacognitive therapy. 58 Indeed, the ConquerFear program for fear of cancer recurrence 7 was based on a combination of metacognitive therapy, acceptance and commitment therapy, and behavioural therapy. The ConquerFear program has demonstrated significant improvements of fear of cancer recurrence in cancer populations, which were maintained over a 6-month period. Importantly, mediation analyses confirmed that it was change in metacognitions and intrusions which fully mediated the efficacy of treatment, indicating that change in metacognitions was likely a mechanism of treatment. 43 As such, an adapted version of the ConquerFear program for endometriosis-related pain could be useful in reducing the burden of pain and FORP in people in endometriosis. Although interpretation bias was not associated with FORP, it remained associated with other important pain outcomes. Interpretation bias can be modified, for example, Cognitive Bias Modification for Interpretation (CBM-I) is a highly scalable, low-cost psychosocial intervention, which may be of benefit at a population level even if effects are small. There is preliminary evidence to suggest that CBM-I is effective in reducing interpretation bias and pain outcomes (eg, pain severity, pain interference, and attention biases) in other chronic pain conditions. 39 Importantly, although these clinical implications are speculative, in an area with few current effective therapeutic interventions, these results give some direction for potential interventions that could be trialled. As such, the current study makes an important contribution to our understanding of FORP, pain, and well-being in people with endometriosis. Using correlations and novel network analyses, we found that 85% of individuals reported severe FORP. People with higher levels of FORP were also found to experience greater physical pain and poorer overall well-being. People with greater existential concerns and unhelpful metacognitive beliefs also experienced greater FORP, whereas evidence to suggest an association between FORP and interpretation bias was mixed. Thus, FORP and underlying constructs of existential concerns, metacognitive beliefs, and interpretation bias may have the potential to be useful treatment targets to reduce the impact of pain on the well-being of people with endometriosis, although our results would suggest that a focus on existential issues is most indicated.

Endometriosis is a chronic condition that affects approximately 1 in 9 individuals assigned female at birth. 61 Pain is a common symptom reported by up to 80% of individuals with endometriosis. 6 , 27 Endometriosis-related pain has been linked to poor psychosocial outcomes, including worse daily functioning and quality of life, and greater anxious and/or depressive symptomology. 36 , 56 Like other chronic pain conditions, endometriosis-related pain is not strongly associated with the level of disease pathology, 55 and psychosocial factors have a key role in determining pain impact. Fear of disease recurrence or progression (FORP) is a construct that has become recognised as common and important for people with chronic health and pain-related conditions. 41 A recent study found that people with endometriosis reported higher levels of FORP than has previously been reported in people with cancer. 49 This is perhaps unsurprising, given that endometriosis is a chronic disease without a cure, and current medical treatments are often inadequate in managing experiences of pain. 3 Given the relevance of FORP in various chronic pain and health-related conditions, 41 the current study drew on core theoretical constructs from models of fear of cancer recurrence to investigate mechanisms potentially relevant in developing or maintaining FORP in people with endometriosis. Specifically, we drew on 3 recent models of FORP and assessed the core construct that they argue has a causal role in FORP. Heathcote and Eccleston's Cancer Threat Interpretation Model argues that the interpretation of ambiguous sensations as painful and threatening is the key driver of FORP (ie, interpretation bias). 19 In Simonelli et al. 44 model, the emphasis is placed on existential concerns (ie, anxiety about the meaning of life, death, and one's identity), whereas in Fardell et al. 16 Cognitive Processing model, metacognitive beliefs (beliefs about thinking processes) are seen to be the primary causal mechanism for severe FORP. Within the chronic pain literature, research has shown greater health-related interpretation biases, existential concerns (eg, death anxiety, meaninglessness, guilt), and unhelpful metacognitive beliefs (eg, thoughts about bias and controllability of health-related thoughts) to be associated with greater emotional distress and experiences of pain. 35 , 42 , 50 , 51 , 57 These constructs may also play a role in endometriosis-related pain given endometriosis is a chronic health condition, where people typically experience high rates of pain over extended periods. 1 , 28 Accordingly, the aims were to identify the presence of severe FORP and determine whether these factors known to be associated with fear of cancer recurrence are also associated with FORP in people with endometriosis. We also sought to understand the nature of relationships between FORP, theoretical constructs (existential concerns, metacognitive beliefs, interpretation bias), emotional distress (depression, anxiety), and pain-related (quality of life, pain severity, pain impact) constructs using correlations and network analyses. Given their relevance as predictors of fear of cancer recurrence, we expected that greater levels of health-related interpretation bias, existential concerns, and unhelpful metacognitive beliefs would be associated with greater FORP. We also expected that higher levels of FORP would be associated with higher levels of anxious and/or depressive symptomology, pain severity, pain impact, and lower quality of life in people with endometriosis.

Supplemental digital content associated with this article can be found online at http://links.lww.com/PAIN/C419 .

The authors have no conflicts of interest to declare.