Validity and screening capacity of the FCR-1r for fear of cancer recurrence in long-term colorectal cancer survivors

As more people are living longer with and after cancer, there is a growing proportion of older survivors, raising the importance of conducting research with these populations. While many long-term cancer survivors do not face the acute stressors of treatment and follow-up, they continuously deal with the uncertainties of survivorship, including potential recurrence. Fear of cancer recurrence (FCR), defined as “fear, worry or concern that cancer will come back or progress” [1] is a top health-related concern for almost all cancer survivors [2], and help managing FCR is one of the most frequently reported unmet supportive care needs [3‐5].

Identifying survivors with clinically significant levels of FCR is important, as it can greatly affect quality of life and may persist for many years without treatment [6, 7]. On average, older long-term cancer survivors report less FCR compared to younger survivors [8], independent of cancer type [7, 9, 10], but high levels of FCR are reported in a small but significant group of older (mean age 77.6 years), long-term (average time since diagnosis 9.5 years) cancer survivors (15.9% according to a Cancer Worry Scale cut-off ≥ 14) [7].

Psychosocial screening of older cancer survivors for multiple psychosocial issues with numerous long questionnaires is problematic, as age and anticancer treatment may together and separately affect cognitive function [11]. Besides this, screening requires administration, scoring, and interpretation by a healthcare provider, all of which may be complicated by lengthy questionnaires.

Efforts have been made to create short, simple, and valid screening items. The single-item FCR measure (FCR-1) [12] with the verbal formulation “On a scale from 0 to 100, what is your subjective level of fear of cancer recurrence at this time?” was developed for use in group sessions aimed at reducing FCR. The FCR-1 has been psychometrically tested [12] and then revised and validated in written form, FCR-1r [13].

In both studies, validation was performed in relatively small samples (FCR-1; N = 69 and FCR-1r; n = 107). The psychometric properties of the FCR-1 were evaluated only in women with breast and gynecological cancer with a mean age of 55, and the FCR-1r was evaluated in a group of patients with mixed cancers where 15% had experienced a cancer recurrence. There is a need to further validate the FCR-1r in additional populations with different types of cancer and at various stages of survivorship to extend generalizability.

The primary objective of this cross-sectional study was to examine the psychometric properties of the FCR-1r, by examining its concurrent, convergent, and divergent validity in a sample of long-term colorectal cancer (CRC) survivors and to investigate the screening performance of the FCR-1r in older versus younger CRC survivors. Results can be used to support the integration of a short and simple FCR screening item into routine psychosocial screening for CRC survivors.

Of the 4,483 CRC survivors, 2,128 (47.5%) returned the survey. 474 responses were excluded during data verification (Fig. 1). Among these were 157 cases with a response pattern of “0” to every FCRI-SF item. 1,654 participants had complete data on both FCR-1r and FCRI-SF after imputation, and were included in the analyses (median age 76 years (range 38–98), 47% females). (See Table 1 for socio-demographic and clinical characteristics for the two age groups (< 65 years and ≥ 65 years) and for the overall sample.) Mean FCR-1r score was 2.6/10 and mean FCRI-SF score was 10.1/36. Non-responders and those who were excluded were most prominent in paper surveys, but CRC survivors willing and eligible to respond were no different from CRC survivors not willing to respond regarding clinical and demographic characteristics (see Table 1a (paper population) and Table 1b (e-Boks population) in supplementary material). No unexpected differences were seen between paper and e-Boks survey respondents. E-boks responders tended to be younger and reported higher FCR levels. E-Boks responders were more likely to have rectal cancer, and a higher percentage were married (as opposed to widowed) and were still employed (see Table 2 in supplementary material).

Valid psychosocial screening instruments with clinically relevant cut-offs are crucial to personalizing survivorship care. This study aimed to assess the validity of the FCR-1r and evaluate its screening performance in older versus younger CRC survivors with no recurrence of cancer. The FCR-1r demonstrated satisfactory concurrent, convergent, and divergent validity compared to the FCRI-SF and measures of constructs related to FCR. A FCR-1r cut-off score ≥ 5 provided an optimal balance of sensitivity, specificity, NPV, and PPV for identifying likely cases of clinically significant FCR (FCRI-SF ≥ 22) in practice. The validity of the FCR-1r was comparable to the original FCR-1 [12], and the classification ability in this population seems adequate.

While this study provides important evidence regarding the validity and screening performance of the FCR-1r in a population of cancer survivors with lower rates of clinical FCR compared to previous findings in CRC survivors [17, 35], further work is needed to confirm the unusually low prevalence of clinical FCR in older populations with different health care systems and to confirm the utility of the FCR-1r as a screening tool for FCR across different survivorship stages, especially in patients with recurrent disease.

Short screening questionnaires can advantageously be followed by longer and more detailed measurements like the FCRI-SF in a two-step screening process to confirm FCR severity, as recommended for depression [36]. In the current context, the high specificity is particularly reassuring, as only 20% of patients (i.e., those reporting false positives) would go through this step unnecessarily. The NPV is extremely high regardless of cut-off, due to the low number of cases in this cohort [37], but the low PPV underpins the need for a two-step approach.

Other globally used psycho-oncology screening measures like the Distress Thermometer (DT) and the ESAS assess fears and anxiety, but not specifically FCR and FCR might occur as an isolated focus of worry not captured by these. For example, high overall distress and a positive fear item on the problem list on the DT [38] showed low NPV and sensitivity to detect high FCR (AUC around 0.70 depending on cut-off on the Cancer Worry Scale (CWS-6)) [39], indicating that a high score could be due to high FCR, but could also reflect other causes of distress, and a low proportion of patients with high FCR would be identified using the DT. No subsequently detailed measurements can detect cases if they do not proceed to this next step, and the DT was deemed unsuitable for FCR screening in routine practice.

In contrast, the ESAS-r [22] anxiety item showed surprisingly good screening performance (AUC = 0.87) in detecting likely cases of clinical FCR in the FCR-1r validation study by Smith et al. [13], comparable to the FCR-1r (AUC = 0.89) and the FCR-1r in our study (AUC = 0.92). Not all patients might recognize FCR as fear, as this generic wording could also be interpreted more broadly as related to the impact of cancer on health, family, or work rather than recurrence in particular, arguably anxiety and FCR seems closer related.

If FCR-1r was expected to be a stand-alone question, it could be just as easy to ask the survivor about FCR orally at the medical encounter, but this rarely happens [40, 41]. Therefore, routine screening is recommended. FCR is only one of many other late effects present among long-term cancer survivors, such as general distress, cognitive impairment, sexual dysfunction, financial toxicity, and fatigue [42], and a larger battery of screening measures should be included in the psychosocial screening. The FCR-1r is valid, time-efficient, and simple to interpret. It can be easily integrated into existing screening procedures in routine survivorship care, and it gives a clear indication of whether it is relevant to spend time at the consultation discussing FCR and possible treatment options.

Missing data was an issue for the paper-and-pencil questionnaires and we also excluded more than 150 responses, with a response pattern of “0.” The fairly large number of responses excluded due to consistent responses despite the reversed item is not ideal. The strategy of imputation was to test two different methods and decide which one to proceed with, based on the impact on the datasets and on the analyses, and compared to complete case analyses. Multiple imputation might be a more powerful and elegant strategy, but besides being more comprehensive, the subsequent STATA prefix “mi estimate” limits the variety of analyses supported by STATA. Mean imputation is easily applicable in clinical and research perspectives and our results confirm the validity of this more accessible strategy, as no difference was observed between imputation styles on the interrelationship between FCRI-SF and FCR-1r. Mean imputation of the participant’s own mean can be used in future studies with missing FCRI-SF data.

As all CRC survivors in Denmark diagnosed between 2014 and 2018 were invited to participate in this study without indicating any sign of interest, the response rate was expected to be low. We do not have registry data on recurrence, but estimate, that about 25% of our population was not eligible for participation due to cancer recurrence. If these were excluded a priori we would reach a response rate of around 60%, consistent with comparable research in this field. Future studies could overcome this by seeking insight into medical records and only approach CRC survivors with no recurrence and no residual disease, or by including people living with or after recurrence of cancer.

Denmark is a highly digitalized country, and all strictly personal communication from the authorities is electronic unless you apply to receive messages by post. Digital literacy is closely, but not completely, related to age, and to comprehensively investigate the FCR level in all CRC survivors, we needed to distribute both paper and electronic surveys. Differences in respondents according to survey type were expected and corresponded to age and age-related characteristics like FCR level, employment status, marital status, education, and the clinical characteristics of cancer type and chemotherapy. Perhaps more importantly there were no differences in registry data between respondents and non-respondents, suggesting that our sample was representative. We have no reason to believe the severity of FCR is related to digital literacy, but this could be further investigated in future studies, especially as there are a growing number of online FCR interventions (e.g., iConquerFear).

Choosing another self-report measure as a gold standard is troublesome, as this measure too has limited accuracy for correctly classifying patients with clinical FCR. Investigating the capacity of the FCR-1r alone compared to a two-step model incorporating the FCR-1r against a gold-standard clinical interview would shed some light on which strategy is most suitable when considering burden on both health professionals and patients.

We focused on validating and testing the screening performance of a short FCR screening tool with high sensitivity and NPV for identifying clinically significant FCR in long-term CRC survivors with minimal patient burden, considering the relatively low prevalence of clinical FCR in this population. However, we note that the FCR-1r also demonstrated good classification ability in identifying less severe levels of FCR, which may be relevant to contexts where patient burden and resources available to provide intervention for lower FCR levels are less of a concern.

Although the prevalence of patients reporting likely clinical FCR (FCRI-SF score ≥ 22) was low among long-term CRC survivors in our sample, particularly those who were older, the FCR-1r screening tool appears to be useful in identifying those who are likely to have clinical levels of FCR. The results suggest that using a cut-off score of ≥ 5 on the FCR-1r tool is appropriate for selecting survivors who may require further assessment and treatment.