The present systematic review is registered on PROSPERO (https://www.crd.york.ac.uk/PROSPERO/; ID: CRD42020184812).

The relevant studies (published between 01 April 2000 and 01 May 2020) were identified by searching MEDLINE, PubMed, Embase, and PsycINFO databases, adhering to The Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) systematic review and meta-analysis guidelines (Shamseer et al., 2015). We also included any study known to the research team that had been submitted or in press in peer-review journals.

The key search terms were (“fear” [MESH] OR worry OR concern OR anxiety) AND (“neoplasm” [MESH] or cancer or carcinoma) AND (“recurrence” [MESH] OR “neoplasm recurrence” [MESH] OR progression OR return OR relapse OR remit) AND (gender OR male OR female OR men OR women).

Papers selected for inclusion had to (a) be published in peer-reviewed journals between 01 April 2000 and 01 May 2020; (b) be written in English; (c) be quantitative studies and (d) report the association between patient gender and FCR or fear of progression (FoP) level in their results. Qualitative studies, dissertations, editorials, conference abstracts and commentaries were excluded. In addition, studies that reported fear of recurrence of nonneoplastic or noncancerous diseases were excluded.

After removing duplicate studies, the titles and abstracts of potential references were reviewed, and unsuitable ones were excluded. Then full texts were acquired and examined. The papers that completely fulfilled the inclusion criteria were conserved and recorded. Data extraction was conducted by CP and overviewed by GH. The following data were extracted for each study: (a) Authors' names, (b) year of publication, (c) sample size of the study, (d) mean age of the sample, (e) cancer types, (f) statistical data on gender and FCR/FoP association, (g) difference direction (males or females that have higher FCR/FoP), (h) FCR/FoP measure utilized (i) country of study, and (j) study design.

For papers incorporated more than one wave of valid data collection from different samples, each data collection was, respectively, extracted as independent studies in our review protocol. Where studies incorporated longitudinal waves of data collections from the same sample, we decided a priori to extract the association statistic from the baseline wave. The logic of selecting the first instance of patient assessment in a panel study was that it would likely have the largest sample size and hence favorable statistical power.

Based on the cancer types included in each of the study samples, all the included papers were divided into two groups. Group 1 (pure group) included studies that exclusively have cancers of one site or one system without gender-specific types. Hence the patients in this group will have contracted cancer such as head and neck cancer that affects both men and women. Whereas, group 2 (mixed group) included the studies with a number of cancer types including, or not including, gender-specific cancers, such as studies with mixed ovarian, prostate and other cancers. Although breast cancer can be contracted by men in rare cases, it is also regarded as a gender-specific type because the vast majority of the patients are women. Only one study in the mixed group does not have gender-specific cancers (Langeveld et al., 2004). Thus, in the pure group, we will be able to exclude the potential bias caused by gender-specific cancers or other various cancer types, while the studies with mixed cancer types were also analyzed so that it could be compared with the pure group to see if significant differences do exist. The cancer type will also be analyzed globally as a moderator in the following meta-regression to further investigate its possible effect on our results.

The Joanna Briggs Institute (JBI) critical appraisal checklist for analytical cross-sectional studies was modified and applied to the cross-sectional, longitudinal, and RCT studies included in the present review (Aromataris et al., 2015; Lim and Humphris, 2020). The JBI checklist was an 8-item quality assessment tool. We excluded two items for our review which were not applicable: “was the exposure measured in a valid and reliable way” and “were objective, standard criteria used for measurement of the condition.” The assessment selections of “unclear” and “not applicable” were merged into one. The 6-item tool was manually applied to all 29 studies included in this review. The modified checklist is presented in the Supplementary Material.

The search process is shown in Figure 1. In total, 3,216 references were identified from the four databases (EMBASE, MEDLINE, PubMed, and PsycINFO). After the duplicates were excluded, the remaining 2,971 titles and abstracts were scrutinized for relevance. Preliminary screening identified 45 papers to be relevant. The full texts were obtained, and the data regarding gender and FCR level were extracted. Finally, 26 papers were included. As described in the Data Extraction section, two included papers each incorporated two waves of data collection of different samples (Humphris et al., 2003; Deimling et al., 2006a). These four waves of data were extracted as four independent studies in the present review for analysis, entitled as Humphris^a,b and Deimling^a,b in Table 1. Additionally, one set of data was extracted from the Head & Neck 5000 study database, entitled as (Head Neck 5000 - NHS Foundation Trust, 2016^d) and is being submitted. Hence the total number of studies for analysis is 29, out of which 15 were included in pure group and 14 were included in mixed group (Tables 1A,B).

Studies were all satisfactory as they were rated positively in half or more of the six criteria (Tables 2A,B).

The N size of pooled participants was 33,339, 16,493 in the pure group and 16,846 in the mixed group. Two thirds of the included studies (66% i.e., 19 out of 29) reported that females have significantly higher FCR level. Two studies reported a higher average mean FCR score for males. These were conducted with separate samples from the Guangzhou Hospital Region (Yang et al., 2019; Luo et al., 2020). The forest plot (Figure 2) shows the effect sizes of the pure group and mixed group are, respectively, 0.28 (95% confidence interval: 0.24–0.32) and 0.29 (95% confidence interval: 0.18–0.40), with an overall effect size of 0.30 (95% CI: 0.23–0.36). The tests for the overall, pure, and mixed group effect size are, respectively, z = 9.16 (p < 0.001), z = 13.90 (p < 0.001), z = 5.34 (P < 0.001). With the “one study removed” analysis it was found that the minimum overall effect size reported was 0.284 demonstrating the very limited effect of any single study to influence the overall results. A classic Fail-safe N calculation resulted in 2,750+ studies would need to be found to bring the effect size to a not significant value (p > 0.05).

The overall Q value for heterogeneity is 127.75 (df = 27, p < 0.001), I² is 78%, and Tau² is 0.034. The Q values of the pure group and mixed group are 14.65 (df = 14, p = 0.403) and 105.40 (df = 13, p < 0.001), respectively. The difference between the two groups is significant (p = 0.006).

The Egger and Begg tests found no consistent evidence of reporting bias (z = −0.07, p = 0.94 and z = −0.99, p = 0.32, respectively). Likewise, the funnel plots showed little evidence of consistent bias (Figure 3) as shown by an approximate symmetric pattern on either side of the 95% CI boundary.

A meta-regression was performed, including four moderators of the publication year (range = 2,003–2,020), cancer type (0 = pure or 1 = mixed), mean age (in years) of the sample (range = 24–74) and the length of the FCR measurement (0 = single-item or 1 = multi-item) (Table 3). No statistically significant effects of these moderators were found (all p levels > 0.4). In addition, we investigated the potential difference between single cancer type, such as the head and neck cancer, and the other cancers but again no reliable effect was shown (z = −0.26, p = 0.79).

An explanation of the gender difference of FCR is likely to be multifactorial. The etiology of FCR development is complex and is poorly understood. The basis of how gender might influence FCR can be speculated upon from various sources. In general, gender differences are common and well-described in mental and psychological conditions among common people as well as cancer patients, and negative conditions are more prevalent in women than men (Faravelli et al., 2013; Salk et al., 2017; Aminisani et al., 2021). A recent study demonstrated that the frequency of psychological distress is especially high among women with colorectal cancers (Aminisani et al., 2021). Another study reported that female sarcoma patients have lower health-related quality of life (HRQoL) (Eichler et al., 2020). Specifically, first, gender differences of social and psychological indicators are associated to factors of income (Reiss, 2013), exposure to violence (Koss et al., 1995) and the division of labor by gender (Wood and Eagly, 2012). These social factors also apply to cancer patients, and women cancer patients usually need more social and mental support (Ozbayir et al., 2019). These have been explained as largely stemming from the gender inequality against women within society but vary widely on a cross-nation level (Salk et al., 2017). Second, women cancer patients are more inclined to express their problems and seek for help, while men patients may abstain from expressing fear or worry when suffering a negative condition because of feelings of shame (Clover et al., 2015; Anuk et al., 2019). This implies that women may tend to report higher FCR level.

Many studies with mixed cancer types included gender-specific cancers when assessing the effect of gender. This may have potential impact on the gender distribution of study samples and gender-related analysis, and lead to bias of study results. Given this concern, we divided all the included studies into “pure” and “mixed” groups. The cancer type seems to have a very limited effect, and less than expected on effect sizes of the two groups, resulting in almost identical values (0.28 vs. 0.29). However, the heterogeneity of the mixed group is much greater than the pure group, leading the overall heterogeneity to a similarly high level. This was to be expected because studies of the mixed group have more diverse samples with various types of cancer including gender-specific cancers and in addition the authors adopted more complex data analytical approaches. They also used, on inspection, a more variable set of FCR measurement tools. Most of the studies of the pure group used standardized questionnaires such as the Fear of Progression Questionnaire (FoP-Q) or the 4 item version of the Fears of Cancer Recurrence Scale (FCR4), while in the mixed group, almost all the studies used study specific FCR measures ranging from a self-defined single question, 4-item scale, to subscales extracted from various other questionnaires such as the Cancer Problems in Living Scale (CPILS), the Health Worries subscale of the Impact of Cancer scale (IOC-HWS) or Health-related Quality of Life (HRQOL). The exception to this rule were three studies that consistently used the FoP-Q short form (Hinz et al., 2014; Götze et al., 2019; Luo et al., 2020). This possible complication, may reflect itself in the observed level of heterogeneity of the mixed group as well as the overall statistics. We consider this may not influence our conclusion substantially because the heterogeneity of the pure group stays at a very low level, and the pure group possessed an almost identical effect size as the mixed group and subsequently the overall result.

We performed a meta-regression including four moderator variables of: cancer type, publication year, mean age of the sample and single- or multi-item FCR measurement, because these factors have potential effects on FCR level. FCR level differs among patients with different cancer types (Crist and Grunfeld, 2013; Simard et al., 2013), which was also discussed in the gender-specific cancer relevant section in this review. In a meta-analysis about the relationship between age and FCR, the authors found the more recent the study, the strength of the effect size decreases (Lim and Humphris, 2020). Several studies demonstrated a relationship between patient age and FCR. The older the cancer patients, the lower the reported FCR level (Simard et al., 2013; Lim and Humphris, 2020). Furthermore, given that there is no consensus on the measurement tool of FCR, the methods in literatures vary from the 43-item Fear of Progression Questionnaire (FoP-Q) (Erim et al., 2013) to a single-item question like “I have fear about my cancer coming back” (Fisher et al., 2016), which generate different FCR results. However, no significant result of the moderator variables was found in the present study, demonstrating they might have less effect than expectation. But in consideration of the previous positive evidence mentioned above, these factors are still important in future studies to be explored further.

In clinical communication, clinicians play a crucial role and often occupy an important position from the patient's perspective. It has been recommended that clinicians should pay attention to the discussion of FCR with cancer patients. There are studies showing that non-mental health trained clinicians can provide effective interventions for FCR (Liu et al., 2019). But this is still challenging because non-mental health trained oncologists often have difficulties in deciding which patients need such assistance. There are tools developed to assist clinicians to be aware of cancer patients' individual needs, such as the Patient Concerns Inventory (PCI), a checklist where patients can select the items they want to discuss at a clinic, including the fear of cancer recurrence (Rogers et al., 2018). However, it may be of benefit if the more easily obtained demographic characteristics, such as gender and age, can be used as easily available indicators to help clinicians to focus on the discussion of FCR with patients at high risk. The result of this review fits our prediction and supports clinicians who consider gender as a useful indicator of FCR level. It might help clinicians to focus discussion of FCR with patients attending follow-up clinics when combine their gender with other easily obtained demographic characteristics, and help improve the efficiency of doctor-patient communication.

This review has a number of strengths including the systematic search for all quality studies that reported the relationship between fears of recurrence or progression with gender. In addition, this is the first study to utilize meta-analytical and meta-regression methodology to assist interpretation of a substantial pooled sample of reported investigations. However, this systematic review and meta-analysis has several limitations. First, to completely avoid the potential bias from gender-specific cancers, only papers about both-gender cancers should be included. However, only sixteen studies have met this criterion, which provides some evidence but not extensive. Therefore, we included gender-specific cancers and divided these studies into two groups (“pure” and “mixed”) as outlined already with the additional analysis to study the effect of cancer types, that showed no statistical significance. Second, this review excluded papers not written in English, which may reduce its generalisabilty to some extent.

In the literatures on FCR so far, gender has been usually regarded as a “not-so-important” factor out of various demographic characteristics. The effect of gender should be investigated with greater consideration and authors in the FCR field are encouraged to report FCR levels by gender. When attempting to assess the factor of gender, cancer types that only a single gender can contract should be considered carefully in their interpretation when intending to investigate gender FCR differences. An argument can be voiced to exclude such cancer types in assessing gender and FCR association to avoid possible confusion. In terms of the high-level heterogeneity observed in the mixed group of this review, while the mixed group of studies may have included a wider selection of patient samples the varied measurements employed indicated that an international consensus on the FCR measurement may be helpful to reduce the number of measures and possibly strengthen consistency. In addition, authors of future studies are commended to include in their manuscript results a breakdown or association statistic of gender and FCR to assist any review update and strengthen our understanding of this key relationship.