Impact of Hormone Replacement Therapy on the Overall Survival and Progression Free Survival of Ovarian Cancer Patients: A Systematic Review and Meta-Analysis

Simple Summary The standard treatment course for ovarian cancer virtually always induces menopause with subsequent symptoms. This systematic review and meta-analysis strived to further elucidate the safety of Hormone Replacement Therapy in the setting of ovarian cancer treatments by investigating its effect on Overall Survival and Progression-Free Survival. The results highlighted a slight benefit in terms of survival and recurrence rates in favour of the hormone replacement therapy groups, pooling respective Hazard Ratios (HR) of 0.66 and 0.73. However, detailed subgroup analyses revealed no statistically significant results in terms of recurrence for the treated groups, while data were sequenced based on stages, grade of differentiation, the radicality of surgery, and the age of participants. Even so, in null outcomes regarding progression-free survival, hormone replacement therapy remains advantageous in lessening menopausal symptoms and improving the quality of life for these patients. Abstract Background: Frequently, patients treated for Ovarian Cancer (OC) undergo menopause with subsequent symptoms. This review scrutinised the impact of Hormone Replacement Therapy (HRT) on the Overall Survival (OS) and Progression-Free Survival (PFS) of patients diagnosed with OC. Methods: A systematic literature search was conducted in the most popular English databases. Inclusion and exclusion criteria were applied to select publications that evaluate OS and PFS in these patients. End-point analysis targeted values of log(HR) and its Standard Error (SE). Results: Up to 1 September 2022, 11 studies were included in the qualitative synthesis. Eight publications, totalling 4191 patients, were included in the meta-analyses. Eight studies were considered for the OS analysis and pooled an HR of 0.66 with respective 95% CI between 0.57 and 0.76, with a p-value < 0.00001 at a Z value of 5.7, in favour of the HRT group. Results for PFS showed an overall HR of 0.73 in favour of the HRT group; CI between 0.57 and 0.95, p = 0.02 at a Z value of 2.36. Further subgroup analyses highlighted the non-inferiority of this treatment. Conclusions: Patients treated for OC that receive HRT for menopausal symptoms after various treatments appeared to have better OS than never-users.


Introduction
The vast majority of ovarian cancers (OC) are diagnosed at stages FIGO (The International Federation of Gynecology and Obstetrics) III and IV [1]. This makes OC one of the most lethal gynaecological neoplasias, with very high mortality rates compared to its relatively low incidence [2]. This is because most cases produce little to no symptoms or have very unspecific manifestations [3]. Moreover, no advantage has been ascertained from effect measure of HR was used, as per Cochrane's recommendations [22]. The ratio summary statistics can take the lowest value of 0 and go up to infinite values, while one is usually thought to be the null effect. Accordingly, the log transformation is usually undertaken. This will render confidence intervals appearing symmetric and is the preferred method for analysing data [23]. Peto's method [24] is acceptable for fixed effects meta-analyses, and the pooled lnHR = ∑ logrank(O−E) ∑ logrank (V) , where O − E represents the Observed minus Expected and V stands for Variance. For this meta-analysis, however, the inverse variance approach was used-the pooled lnHR = ∑ lnHR V ∑ 1 V , where V represents the variance of lnHR. Cox proportional hazard models usually provide lnHRs and SE and, therefore, are compatible with the inverse variance pathway. Sometimes, studies might provide HRs and respective 95% CI. In this case, V can be obtained by the formula V = UppCI−LowCI . The UppCI and LowCI are the upper and lower 95% CI, and φ stands for the cumulative distribution function of the normal distribution. Hence, φ −1 1 − ∝ 2 = 1.96 for 95% CI. Regardless, the most suitable statistics are not always presented in primary publications; therefore, transformations need to be done to obtain the lnHRs and their variance [25]. Tierney et al.'s paper accurately depicts 10 ways to obtain the desired summary measures from individual trials, while Cochrane's handbook depicts 3 main derived reliable methods. While the first was discussed above, another possibility is obtaining HR estimates from the log-rank analysis, a direct method. Here, Simmonds [26] describes the lnHR = (O−E) V , where O stands for the observed events in the research groups, E is the log-rank expected number of events and V stands for the variance of the test. However, sometimes data are presented as just the number of analysed patients, events in both groups and perhaps a p-value from the log-rank test [27]. For these instances, the following were used: in the case of equal In this scenario, the variance will be estimated by V ≈ O r ×O c O . If the randomisation is not equal in both groups, another formula was employed: Here, the R c stand for the number of patients. The variance will be estimated accordingly to V ≈ O×R r ×R c (R r +R c ) 2 . Regardless of the method used, the (O − E) and V values were imputed in lnHR = (O−E) V , and the var(ln(HR)) is easily deduced by var(ln(HR)) = 1 V . These values were finally used in the inverse variance method for obtaining pooled HRs. The third option will be to reconstruct data from the Kaplan-Meier survival curves [28] if neither of the above is presented in the original papers.

Planned Analysis Method
Data were computed into the Fixed or Random Effects model depending on the levels of heterogeneity. Heterogeneity was assessed using the Chi-squared (Chi 2 ) test. The cut-off values for p were set at 0.10. I 2 values of 25, 50 and 75% were considered low, medium and high heterogeneity.

Study Characteristics
Essential data extracted from primary publications are shown in Table 1. Two RCTs [15,62] and one non-RCT [60] were included, alongside seven cohort studies and one retrospective case-control study. Only five publications reported adverse side effects of HRT [15,60,61,63,68], and two reported second primary malignancies [15,63].

Study Characteristics
Essential data extracted from primary publications are shown in Table 1. Two RCTs [15,62] and one non-RCT [60] were included, alongside seven cohort studies and one retrospective case-control study. Only five publications reported adverse side effects of HRT [15,60,61,63,68], and two reported second primary malignancies [15,63].

Risk of Bias within Studies
The NOS [21] was used for the included cohort studies [61,63,[65][66][67][68][69], case-control studies [60] and non-randomised trials [60] [15] study had a low risk of bias, while the other RCT presented some concerns. These were primarily due to the deviations from the intended interventions, as some participants in the experimental group stopped taking the HRT. Results are highlighted in Table 3.

Risk of Bias within Studies
The NOS [21] was used for the included cohort studies [61,63,[65][66][67][68][69], case-control studies [60] and non-randomised trials [60] [15] study had a low risk of bias, while the other RCT presented some concerns. These were primarily due to the deviations from the intended interventions, as some participants in the experimental group stopped taking the HRT. Results are highlighted in Table 3.

Results of Individual Studies
Eight studies totalling 3578 patients were included in the OS quantitative analysis. Four studies [15,66,67,69] provided unadjusted HRs with corresponding 95% CI; therefore, the direct method for extracting corresponding logHRs and SEs was used. Three other studies [61][62][63] provided observed events numbers in both groups; consequently, the indirect transformation formula using the provided log-rank p-values was used to determine logHRs and SEs. One study [68] presented both options; hence, the direct method was used again. Detailed results can be found in the supplementary material, the Summary Statistics sheet.

Risk of Bias within Studies
The NOS [21] was used for the included cohort studies [61,63,[65][66][67][68][69], case-control studies [60] and non-randomised trials [60] [15] study had a low risk of bias, while the other RCT presented some concerns. These were primarily due to the deviations from the intended interventions, as some participants in the experimental group stopped taking the HRT. Results are highlighted in Table 3.

Results of Individual Studies
Eight studies totalling 3578 patients were included in the OS quantitative analysis. Four studies [15,66,67,69] provided unadjusted HRs with corresponding 95% CI; therefore, the direct method for extracting corresponding logHRs and SEs was used. Three other studies [61][62][63] provided observed events numbers in both groups; consequently, the indirect transformation formula using the provided log-rank p-values was used to determine logHRs and SEs. One study [68] presented both options; hence, the direct method was used again. Detailed results can be found in the supplementary material, the Summary Statistics sheet.

Risk of Bias within Studies
The NOS [21] was used for the included cohort studies [61,63,[65][66][67][68][69], case studies [60] and non-randomised trials [60] [15] study had a low risk while the other RCT presented some concerns. These were primarily due to the de from the intended interventions, as some participants in the experimental group taking the HRT. Results are highlighted in Table 3.

Results of Individual Studies
Eight studies totalling 3578 patients were included in the OS quantitative Four studies [15,66,67,69] provided unadjusted HRs with corresponding therefore, the direct method for extracting corresponding logHRs and SEs was use other studies [61][62][63] provided observed events numbers in both groups; conse the indirect transformation formula using the provided log-rank p-values was determine logHRs and SEs. One study [68] [15] stud while the other RCT presented some concerns. These were primari from the intended interventions, as some participants in the exper taking the HRT. Results are highlighted in Table 3. Table 3. Rob scores for the included RCT.

Study ID
Year

Deviations from the Intended Interventions
Missing Outcome Data

Results of Individual Studies
Eight studies totalling 3578 patients were included in the O Four studies [15,66,67,69] provided unadjusted HRs with co therefore, the direct method for extracting corresponding logHRs a other studies [61][62][63] provided observed events numbers in both the indirect transformation formula using the provided log-rank determine logHRs and SEs. One study [68] presented both opt method was used again. Detailed results can be found in the supp Summary Statistics sheet.

Risk of Bias within Studies
The NOS [21] was used for the included coh studies [60] and non-randomised trials [60] to ass excellent ratings, ranging between 6 [60,64], 8 [65stars, showing the sound quality of the included st  Table 3. Table 3. Rob scores for the included RCT.

Study ID
Year

Results of Individual Studies
Eight studies totalling 3578 patients were inc Four studies [15,66,67,69] provided unadjusted therefore, the direct method for extracting correspon other studies [61][62][63] provided observed events nu the indirect transformation formula using the pro determine logHRs and SEs. One study [68] pres method was used again. Detailed results can be fou Summary Statistics sheet.  Table 3. Table 3. Rob scores for the included RCT.

Study ID
Year

Results of Individual Studies
Eight studies totalling 3578 patients were inclu Four studies [15,66,67,69] provided unadjusted therefore, the direct method for extracting correspond other studies [61][62][63] provided observed events num the indirect transformation formula using the prov determine logHRs and SEs. One study [68]

Risk of Bias within Studies
The NOS [21] was used for the included cohort studies [61,63,[65][66][67][68][69], case-control studies [60] and non-randomised trials [60] [15] study had a low risk of bias, while the other RCT presented some concerns. These were primarily due to the deviations from the intended interventions, as some participants in the experimental group stopped taking the HRT. Results are highlighted in Table 3.

Results of Individual Studies
Eight studies totalling 3578 patients were included in the OS quantitative analysis. Four studies [15,66,67,69] provided unadjusted HRs with corresponding 95% CI; therefore, the direct method for extracting corresponding logHRs and SEs was used. Three other studies [61][62][63] provided observed events numbers in both groups; consequently, the indirect transformation formula using the provided log-rank p-values was used to determine logHRs and SEs. One study [68] presented both options; hence, the direct method was used again. Detailed results can be found in the supplementary material, the Summary Statistics sheet.

Risk of Bias within Studies
The NOS [21] was used for the included cohort studies [61,63,[65][66][67][68][69], case-control studies [60] and non-randomised trials [60] [15] study had a low risk of bias, while the other RCT presented some concerns. These were primarily due to the deviations from the intended interventions, as some participants in the experimental group stopped taking the HRT. Results are highlighted in Table 3.

Results of Individual Studies
Eight studies totalling 3578 patients were included in the OS quantitative analysis. Four studies [15,66,67,69] provided unadjusted HRs with corresponding 95% CI; therefore, the direct method for extracting corresponding logHRs and SEs was used. Three other studies [61][62][63] provided observed events numbers in both groups; consequently, the indirect transformation formula using the provided log-rank p-values was used to determine logHRs and SEs. One study [68] presented both options; hence, the direct method was used again. Detailed results can be found in the supplementary material, the Summary Statistics sheet.

Risk of Bias within Studies
The NOS [21] was used for the included cohort studies [61,63,[65][66][67][68][69], case studies [60] and non-randomised trials [60] [15] study had a low risk while the other RCT presented some concerns. These were primarily due to the de from the intended interventions, as some participants in the experimental group taking the HRT. Results are highlighted in Table 3.

Results of Individual Studies
Eight studies totalling 3578 patients were included in the OS quantitative Four studies [15,66,67,69] provided unadjusted HRs with corresponding therefore, the direct method for extracting corresponding logHRs and SEs was use other studies [61][62][63] provided observed events numbers in both groups; conse the indirect transformation formula using the provided log-rank p-values was determine logHRs and SEs. One study [68] [15] stud while the other RCT presented some concerns. These were primari from the intended interventions, as some participants in the exper taking the HRT. Results are highlighted in Table 3. Table 3. Rob scores for the included RCT.

Study ID
Year

Deviations from the Intended Interventions
Missing Outcome Data

Results of Individual Studies
Eight studies totalling 3578 patients were included in the O Four studies [15,66,67,69] provided unadjusted HRs with co therefore, the direct method for extracting corresponding logHRs a other studies [61][62][63] provided observed events numbers in both the indirect transformation formula using the provided log-rank determine logHRs and SEs. One study [68] presented both opt method was used again. Detailed results can be found in the supp Summary Statistics sheet.

Risk of Bias within Studies
The NOS [21] was used for the included coh studies [60] and non-randomised trials [60] Table 3. Table 3. Rob scores for the included RCT.

Study ID
Year

Results of Individual Studies
Eight studies totalling 3578 patients were inc Four studies [15,66,67,69] provided unadjusted therefore, the direct method for extracting correspon other studies [61][62][63] provided observed events nu the indirect transformation formula using the pro determine logHRs and SEs. One study [68] pres method was used again. Detailed results can be fou Summary Statistics sheet.
Eeles et al.' s [15] and Guidozzi' of the RoB2 program [20] for parallel R while the other RCT presented some c from the intended interventions, as so taking the HRT. Results are highlighte Table 3. Rob scores for the included RCT.

Study ID
Year

Results of Individual Studies
Eight studies totalling 3578 patie Four studies [15,66,67,69] provided therefore, the direct method for extrac other studies [61][62][63] provided obser the indirect transformation formula determine logHRs and SEs. One stu method was used again. Detailed resu Summary Statistics sheet.

Risk of Bias within Studies
The NOS [21] was used for the included cohort studies [61,63,[65][66][67][68][69], case-control studies [60] and non-randomised trials [60] [15] study had a low risk of bias, while the other RCT presented some concerns. These were primarily due to the deviations from the intended interventions, as some participants in the experimental group stopped taking the HRT. Results are highlighted in Table 3.

Results of Individual Studies
Eight studies totalling 3578 patients were included in the OS quantitative analysis. Four studies [15,66,67,69] provided unadjusted HRs with corresponding 95% CI; therefore, the direct method for extracting corresponding logHRs and SEs was used. Three other studies [61][62][63] provided observed events numbers in both groups; consequently, the indirect transformation formula using the provided log-rank p-values was used to determine logHRs and SEs. One study [68] presented both options; hence, the direct method was used again. Detailed results can be found in the supplementary material, the Summary Statistics sheet.

Risk of Bias within Studies
The    Table 3.

Results of Individual Studies
Eight studies totalling 3578 patients were included in the OS quantitative analysis.
Four studies [15,66,67,69] provided unadjusted HRs with corresponding 95% CI; therefore, the direct method for extracting corresponding logHRs and SEs was used. Three other studies [61][62][63] provided observed events numbers in both groups; consequently, the indirect transformation formula using the provided log-rank p-values was used to determine logHRs and SEs. One study [68]

Results of Individual Studies
Eight studies totalling 3578 patients were included in the OS quantitative analysis. Four studies [15,66,67,69] provided unadjusted HRs with corresponding 95% CI; therefore, the direct method for extracting corresponding logHRs and SEs was used. Three other studies [61][62][63] provided observed events numbers in both groups; consequently, the indirect transformation formula using the provided log-rank p-values was used to determine logHRs and SEs. One study [68] presented both options; hence, the direct method was used again. Detailed results can be found in the supplementary material, the Summary Statistics sheet.

Synthesis of Results
The meta-analysis investigating the effect of HRT on OS of OC patients included 3578 patients, out of which 912 received HRT. The fixed effects model pooled an overall

Synthesis of Results
The meta-analysis investigating the effect of HRT on OS of OC patients included 3578 patients, out of which 912 received HRT. The fixed effects model pooled an overall HR of 0.66, with a 95% CI of 0.57 to 0.76, showing statistical significance (p < 0.00001) at a Z value of 5.70 in favour of the HRT group. Heterogeneity was 25%. The results are shown in Figure 2.

Publication Bias
Publication attrition was assessed using funnel plots. The graphic in Figure 3 shows a slight tendency to asymmetry. No significant bias was detected.

Result of Additional Analyses
Sensitivity analyses were achieved by excluding one study at a time from the quantitative synthesis, and results can be conferred in Table 4. All results remained in the initial 95% CI, and the overall effect was kept throughout, showing robust sensitivity. Table 4. Sensitivity analysis for the OS statistics.

Publication Bias
Publication attrition was assessed using funnel plots. The graphic in Figure 3 shows a slight tendency to asymmetry. No significant bias was detected.

Synthesis of Results
The meta-analysis investigating the effect of HRT on OS of OC patients included 3578 patients, out of which 912 received HRT. The fixed effects model pooled an overall HR of 0.66, with a 95% CI of 0.57 to 0.76, showing statistical significance (p < 0.00001) at a Z value of 5.70 in favour of the HRT group. Heterogeneity was 25%. The results are shown in Figure 2.

Publication Bias
Publication attrition was assessed using funnel plots. The graphic in Figure 3 shows a slight tendency to asymmetry. No significant bias was detected.

Result of Additional Analyses
Sensitivity analyses were achieved by excluding one study at a time from the quantitative synthesis, and results can be conferred in Table 4. All results remained in the initial 95% CI, and the overall effect was kept throughout, showing robust sensitivity. Table 4. Sensitivity analysis for the OS statistics.

Result of Additional Analyses
Sensitivity analyses were achieved by excluding one study at a time from the quantitative synthesis, and results can be conferred in Table 4. All results remained in the initial 95% CI, and the overall effect was kept throughout, showing robust sensitivity.
None of the included studies reported results for the OS based on staging, resectability, differentiation or histological subtypes of the disease or age of the participants. Moreover, none of the publications reported results based on type, inception timing or HRT duration. Therefore, the only feasible subgroup analysis was based on the type of included publications, RCTs vs other types. As expected, sequencing results from RCTs only lessened the heterogeneity to 0%. The overall effect was kept at an HR of 0.69. Nevertheless, the other types of studies pooled a higher 41% I 2 value. Even so, the effect carried over at an HR of 0.64. No subgroup differences were highlighted. Results can be consulted in Figure 4. None of the included studies reported results for the OS based on staging, resectability, differentiation or histological subtypes of the disease or age of the participants. Moreover, none of the publications reported results based on type, inception timing or HRT duration. Therefore, the only feasible subgroup analysis was based on the type of included publications, RCTs vs other types. As expected, sequencing results from RCTs only lessened the heterogeneity to 0%. The overall effect was kept at an HR of 0.69. Nevertheless, the other types of studies pooled a higher 41% I 2 value. Even so, the effect carried over at an HR of 0.64. No subgroup differences were highlighted. Results can be consulted in Figure 4. Unremarkable findings were observed in the corresponding funnel plot for this analysis, which is shown in Figure 5. Unremarkable findings were observed in the corresponding funnel plot for this analysis, which is shown in Figure 5.

Results of Individual Studies
Five studies totalling 613 patients were included in the PFS quantitative analysis. A summary of the results is shown in the Supplementary Material, in the Summary Statistics sheet. One publication [15] provided unadjusted HRs with corresponding 95% CI; therefore, the direct method for extracting corresponding logHRs and SEs was used. Two other studies [62,63] provided observed events numbers in both groups; thus, the indirect transformation formula using the provided log-rank p-values was used to determine logHRs and SEs. One study [68] presented both options, and the direct method was preferred.
Interestingly, one study [61] presented an unadjusted HR outside the 95% CI. However, the publication also provided the number of observed events in each group and logrank statistics. Hence, the indirect method was used again. Details are provided in the supplementary material, in the Summary Statistics sheet. Cancers 2023, 15, x FOR PEER REVIEW 13 of 26

Results of Individual Studies
Five studies totalling 613 patients were included in the PFS quantitative analysis. A summary of the results is shown in the Supplementary Material, in the Summary Statistics sheet. One publication [15] provided unadjusted HRs with corresponding 95% CI; therefore, the direct method for extracting corresponding logHRs and SEs was used. Two other studies [62,63] provided observed events numbers in both groups; thus, the indirect transformation formula using the provided log-rank p-values was used to determine logHRs and SEs. One study [68] presented both options, and the direct method was preferred.
Interestingly, one study [61] presented an unadjusted HR outside the 95% CI. However, the publication also provided the number of observed events in each group and log-rank statistics. Hence, the indirect method was used again. Details are provided in the supplementary material, in the Summary Statistics sheet.

Synthesis of Results
The meta-analysis examining the effect of HRT on the PFS of OC patients included 613 patients, out of which 266 received HRT. The fixed effects model pooled an overall HR of 0.73, with a 95% CI of 0.57 to 0.95, showing statistical significance (p = 0.002) at a Z value of 2.36 in favour of the HRT group. Heterogeneity was trivial at 0%. The results are shown in Figure 6.

Synthesis of Results
The meta-analysis examining the effect of HRT on the PFS of OC patients included 613 patients, out of which 266 received HRT. The fixed effects model pooled an overall HR of 0.73, with a 95% CI of 0.57 to 0.95, showing statistical significance (p = 0.002) at a Z value of 2.36 in favour of the HRT group. Heterogeneity was trivial at 0%. The results are shown in Figure 6.

Publication Bias
Publication attrition was assessed using funnel plots. The graphic in Figure 7 shows a slight tendency to asymmetry again. No significant bias was detected. Wen Guidozzi Eeles Figure 6. Forrest plots for the HRT vs. No-HRT groups' PFS analysis.

Publication Bias
Publication attrition was assessed using funnel plots. The graphic in Figure 7 shows a slight tendency to asymmetry again. No significant bias was detected.

Publication Bias
Publication attrition was assessed using funnel plots. The graphic in Figure 7 shows a slight tendency to asymmetry again. No significant bias was detected.

Result of Additional Analyses
Sensitivity analyses were achieved by excluding one study at a time from the quantitative synthesis, and the results can be conferred in Table 5. All results remained in the initial 95% CI, and the overall effect was kept throughout, showing robust sensitivity.

Result of Additional Analyses
Sensitivity analyses were achieved by excluding one study at a time from the quantitative synthesis, and the results can be conferred in Table 5. All results remained in the initial 95% CI, and the overall effect was kept throughout, showing robust sensitivity. Multiple subgroup analyses were feasible. The first subgroup analysis was based on the type of included publications, RCTs vs other types. Heterogeneity remained trivial, and no subgroup differences were highlighted. At a closer look, results based on RCTs did not seem to carry over the effect, pooling an HR of 0.73, showing no statistically significant effect at a p-value of 0.05. Other studies failed to highlight any differences. Results can be consulted in Figure 8.
Multiple subgroup analyses were feasible. The first subgroup analysis was based on the type of included publications, RCTs vs other types. Heterogeneity remained trivial, and no subgroup differences were highlighted. At a closer look, results based on RCTs did not seem to carry over the effect, pooling an HR of 0.73, showing no statistically significant effect at a p-value of 0.05. Other studies failed to highlight any differences. Results can be consulted in Figure 8. The corresponding funnel plot for this analysis, shown in Figure 9, did not highlight publication attrition. The corresponding funnel plot for this analysis, shown in Figure 9, did not highlight publication attrition.
Multiple subgroup analyses were feasible. The first subgroup analysis was based on the type of included publications, RCTs vs other types. Heterogeneity remained trivial, and no subgroup differences were highlighted. At a closer look, results based on RCTs did not seem to carry over the effect, pooling an HR of 0.73, showing no statistically significant effect at a p-value of 0.05. Other studies failed to highlight any differences. Results can be consulted in Figure 8. The corresponding funnel plot for this analysis, shown in Figure 9, did not highlight publication attrition. Four studies [61][62][63]68] presented results based on the stages of the disease. A subgroup analysis was carried out, and data were split into stages from I to IV. While HRT carried over a tendency to improve recurrences, it was not statistically significant at any stage of the disease. Results are shown in Figure 10. Two studies, however, failed to include any stage IV patients [61,63].  Four studies [61][62][63]68] presented results based on the stages of the disease. A subgroup analysis was carried out, and data were split into stages from I to IV. While HRT carried over a tendency to improve recurrences, it was not statistically significant at any stage of the disease. Results are shown in Figure 10. Two studies, however, failed to include any stage IV patients [61,63].    Differentiation grade subgroup analysis included 429 patients. Neither wellmoderated nor poorly differentiated subtypes showed any benefit in recurrence rates from HRT. One publication failed to report data for moderated and poorly differentiated disease, but mixed the results [63]. Numbers can be consulted in Figure 12. Differentiation grade subgroup analysis included 429 patients. Neither well-moderated nor poorly differentiated subtypes showed any benefit in recurrence rates from HRT. One publication failed to report data for moderated and poorly differentiated disease, but mixed the results [63]. Numbers can be consulted in Figure 12. Differentiation grade subgroup analysis included 429 patients. Neither wellmoderated nor poorly differentiated subtypes showed any benefit in recurrence rates from HRT. One publication failed to report data for moderated and poorly differentiated disease, but mixed the results [63]. Numbers can be consulted in Figure 12.  Trivial publication bias is found in the funnel plots shown in Figure 13. Trivial publication bias is found in the funnel plots shown in Figure 13. Exploring whether HRT would affect recurrences based on the resectability of the disease yielded no statistically significant results either. The comparison between optimally debulked and suboptimally debulked patients is presented in Figure 14. Exploring whether HRT would affect recurrences based on the resectability of the disease yielded no statistically significant results either. The comparison between optimally debulked and suboptimally debulked patients is presented in Figure 14. Exploring whether HRT would affect recurrences based on the resectability of the disease yielded no statistically significant results either. The comparison between optimally debulked and suboptimally debulked patients is presented in Figure 14. The funnel plot presented in Figure 15 is unremarkable. The funnel plot presented in Figure 15 is unremarkable.  Finally, two studies presented PFS results based on the participants' age. The analysis proved no benefit in terms of PFS for any age categories, as shown in Figure 16. Finally, two studies presented PFS results based on the participants' age. The analysis proved no benefit in terms of PFS for any age categories, as shown in Figure 16. Finally, two studies presented PFS results based on the participants' age. The analysis proved no benefit in terms of PFS for any age categories, as shown in Figure 16.  The corresponding funnel plot can be consulted in Figure 17.
Cancers 2023, 15, x FOR PEER REVIEW 2 The corresponding funnel plot can be consulted in Figure 17. Results from individual studies based on respective subgroup categories c further consulted in the Supplementary Material, in the PFS Subgroups sheet, alon the detailed workflow. All subgroup analyses were based on the indirect metho pooling logHRs and SEs due to the nature of the data presented in the or publications. Fixed effect methods were employed, given the trivial heterogeneity Results from individual studies based on respective subgroup categories can be further consulted in the Supplementary Material, in the PFS Subgroups sheet, alongside the detailed workflow. All subgroup analyses were based on the indirect method for pooling logHRs and SEs due to the nature of the data presented in the original publications. Fixed effect methods were employed, given the trivial heterogeneity in all the analyses.

Summary of Evidence
The results of the current meta-analyses align with those presented in the primary included publications [15,[61][62][63][66][67][68][69] and emphasise a benefit in the OS of OC patients receiving HRT compared to never-users. The overall pooled HR of 0.66 showed statistical significance with a p-value < 0.00001, while the analyses pooled a trivial 25% heterogeneity and used a fixed effects model. The relatively low number of publications was mitigated by a robust population size of 3578 patients, of which 912 received HRT. No substantial publication attrition was disclosed while performing the funnel plots for the included publications. Results align with previous spottings that exhibit an OS benefit for the HRT groups. To the best of our knowledge, up to this date, there are only three other metaanalyses published in the literature that investigate the subject of HRT in the setting of OC. Li et al.'s meta-analysis [70], published in 2015, included 1448 patients and pooled an HR = 0.69 (95% CI: 0.61-0.79), also statistically significant. However, their study did not investigate the PFS in these settings, but analysed the RR (Relative Risk) of occurrence. Pergialiotis' meta-analysis [71], published in 2016, based their statistics on the OR of cancerrelated deaths and recurrences in 1521 women. Their study found no statistically significant discrepancies for these groups regarding OS and recurrence. Finally, the 2020 Cochrane systematic review [72] focused on the QoL indicators. Nevertheless, the study did enclose 350 patients in an OS analysis that pooled a favourable HR of 0.71 for the HRT group. The present meta-analysis included 4191 OC patients in two analyses regarding the OS and PFS (measured by HRs) of patients treated or not with HRT, the largest to date. Adding to the novelty, comprehensive subgroup analyses were undertaken to evaluate the actual effect size of HRT treatment. Interestingly, when analysing HRT based on the age of participants and the stage, differentiation and resectability of the disease, all analyses were deemed insignificant in terms of recurrences. It is worth mentioning that all included studies were cohorts [61][62][63]68], as the only RCT [15] investigating the PFS only provided overall PFS, rather than results based on the categories above. Thus, the results must be interpreted with caution. Regardless, even in the context of a null effect on the recurrence rates, HRT can still be deemed a viable option for these patients in terms of improving QoL and lessening climacteric symptoms.
Mixing RCTs, cohort studies or other studies might introduce bias and potentially become problematic. However, due to the limited number of identified publications, this was performed as a necessity. Accordingly, a subgroup analysis was undertaken to sequence the data from RCTs [15,62] and other types of studies. This was possible for both the OS and the PFS analyses. The analysis proved helpful, as it diminished the heterogeneity from 25% down to 0% in OS analysis when only considering RCTs, proving that mixing the results was the cause of the heterogeneity in the initial analysis. Cochrane also states that when pooling the desired effect measures from non-randomised studies, HR can be obtained from adjusted analyses, such as Cox multivariate regression analyses [73]. This will indeed lessen the risk of bias pooled from such publications; yet, some might consider these HRs incompatible for meta-analyses with the unadjusted HRs or those extracted directly from (O − E) events [27]. The present meta-analyses only used HRs pooled from unadjusted univariate analyses or obtained via a direct method described in the materials and methods subsection.
While there is still some disagreement regarding the risk-to-benefit ratios of HRT in treating menopausal symptoms [11], the advantages are well documented in selected patients and scenarios [12]. However, there is still apprehension as to the mechanism of action of HRT in OC. Interestingly, the primary included publications and the present meta-analyses showed a benefit in OS and PFS for the patients treated with HRT. Although E and P are known for reducing overall mortality and morbidity in general menopause, the mechanism of action in OC might be more complex; otherwise, the effect would not have carried over in the PFS analysis. One might argue that clinicians are more inclined to prescribe HRT for OC patients who are younger, fitter and have an earlier-stage disease and, therefore, a better prognosis. However, most studies controlled for such variables and provided measures of dissimilarities between groups. This was also marked in the quality assessment of cohorts and RCTs that yielded promising results, potentially mitigating the selection bias for the treatment and control groups. Consequently, a mechanism of action needs to be elucidated for HRT in OC, and perhaps future molecular models and studies will strive to explain.
Multifarious treatments were applied in the primary included publications, ranging from E only or P only to various combinations and doses. Some differences were also highlighted in the moment of inception for the HRT relative to the treatment of OC and the duration. Hence, sound conclusions cannot be outlined regarding the exact dosage, drug, moment of inception and course of treatment necessary for HRT to have a positive effect. Even so, it appears from these primary included publications that most regimens begin within the first year of oncologic treatments and last for at least one year. Although side effects reporting was scarce in the included studies [15,60,61,63,68], mammary gland hyperplasia seemed to be one of the most common side effects [61,68] of the treatment that demanded treatment to be halted. More concerning were the secondary primary malignancies reported in two studies [15,63]. Eeles' publication [15] disclosed two breast malignancies, alongside one colon and one jejunum malignant tumour, while an older study highlighted breast carcinoma [63].

Limitations and Strengths
The relatively low number of included publications for the meta-analyses can be an intrinsic limitation. However, it is partially mitigated by the more significant population sizes derived from these studies. The limited number of databases can also influence the outcomes through missing potentially eligible reports. However, a large number of 7814 studies were initially identified, while no publication attrition was detected. A slight warping in the funnel plots can also be attributed to the smaller studies that tend to magnify the effect. While no language criteria were applied for the search formula, articles that did not present an English full-text were excluded from this review, potentially marking language as a limiting factor. I 2 was used for heterogeneity assessment. Although this is not an absolute measurement of heterogeneity, I 2 is a valuable tool for highlighting the proportion between variances in true effect size or sampling errors. Chi 2 and p-values were also added. Trivial heterogeneity was regarded in all meta-analyses with satisfactory Chi 2 and p values for I 2 . This must also be interpreted with caution, as it can also highlight an over-selection of studies.
HRs were the choice parameters, and analyses were based on the logHR values and their respective variance measures, SEs. Cochrane advises against using continuous data outcomes for time-to-event measures (mean or median time of survival or until recurrence), as they usually exclude the censored data and produce bias [22]. One study reported the outcomes in this manner [65] and was not included in the present meta-analyses. HR is continuously changing; however, for the time-to-event analyses, most publications will assume a constant HR for participants through their contribution time. This is called a simplified HR, and this assumption will, nevertheless, be carried over in the pooling of data when performing a meta-analysis [73]. It is virtually impossible to overcome, but must be stated as a potential limitation to the present analyses.

Conclusions
The current systematic review and meta-analysis indicated that HRT could be safely and efficiently administered to patients treated for OC, who invariably experience menopause and subsequent symptoms. Furthermore, a statistically significant advantage in the OS and PFS has been marked in the HRT-treated groups compared to never-users, potentially implying a role of the HRT in managing OC patients with menopause-related manifestations. More detailed analyses based on the age of participants and the stage, grade of differentiation and resectability of the disease failed to disclose any benefit in terms of PFS for HRT users. However, even in this setting of non-inferiority, HRT can be safely considered for lessening symptoms and improving QoL for these patients. Institutional Review Board Statement: Ethical review and approval were waived for this study because it used data previously published in other papers.

Data Availability Statement:
The data presented in this study are available in supplementary material.