Prognostic Role of Serum Antibody Immunity to p53 Oncogenic Protein in Ovarian Cancer: A Systematic Review and a Meta-Analysis

Objective Serum p53 autoantibodies (p53-AAbs) are the product of an endogenous immune response against p53 overexpression driven by the ovarian tumour. The p53-AAbs are detectable only in a subset of patients. To date, the evidence of an association between the presence of p53-AAbs and ovarian cancer outcomes has been poorly investigated. Methods A systematic literature search was performed to identify eligible studies investigating the association of serum p53-AAbs and overall survival (OS) and disease free survival (DFS). Associations between presence of serum p53-AAbs and baseline tumour characteristics were also evaluated. Pooled hazard ratios (HRs) and corresponding 95% confidence intervals (CI) were computed to estimate the prognostic impact of serum p53-AAbs. Heterogeneity between studies was assessed. Results A total of 583 patients (7 studies) for OS and 356 patients (4 studies) for DFS were included in the meta-analysis. Presence of p53-AAbs was not associated to OS (pooled uni- multivariate HR = 1.09; 95% CI: 0.55–2.16), and a large heterogeneity was found. When only multivariate HRs were pooled together (4 studies), presence of p53-AAbs was significantly associated to a better OS (pooled HR = 0.57; 95% CI: 0.40–0.81), and no significant heterogeneity was observed. A reduced DFS was associated to p53-AAbs (pooled uni- multivariate HR = 1.37; 95% CI: 0.83–2.25), though not significantly and with a moderate heterogeneity. Conclusions The prognostic significance of serum p53-AAbs in ovarian cancer was diverging according to uni or multivariate models used. Since the results of this work were based on only few investigations, large prospective studies are needed to better define the role of antibody immunity against p53.


Literature Search
PUBMED, EMBASE, Cochrane library and Web of Science databases were comprehensively searched to identify eligible studies on the association between serum p53-AAbs and ovarian cancer prognosis, including OS, DFS, relapse free survival (RFS) and progression free survival (PFS). Furthermore, reported associations between serum p53-AAbs and baseline tumour characteristics were also commented. All articles were extracted by May 29, 2015. In order to search and include all potential studies, we applied various combinations of the following medical subject headings and key words in order to hold high sensitivity: p53 autoantibodies, or serum p53 autoantibodies, or p53-AAbs, or serum autoantibodies, or p53 immunity, or antiovarian antibodies; ovarian cancer, or ovarian, or ovarian tumour; survival, or disease free survival, or prognosis, or outcome, or clinical. As a search limit, only studies published in English and concerning humans were included. In addition, references of other narrative and systematic reviews were checked for relevant articles.

Eligibility criteria
All the retrieved records were independently screened by two distinct reviewers. Disagreements were resolved by consultation with a third reviewer. Firstly, all irrelevant records, reviews, case reports, studies on animals or cell lines, and studies on other cancers were excluded in addition to all papers in which presence of serum p53-AAbs was assayed only for a diagnostic purpose (early detection of ovarian cancer). Eligible studies meeting the following criteria were included to evaluate associations between p53-AAbs and ovarian cancer outcome: (1) proven diagnosis of ovarian cancer; (2) serum or plasma p53-AAbs detection by using multiple methods. Detection of autoantibodies against p53 in tissue samples by immunohistochemistry (IHC) techniques was not included, as well as evaluations of p53-AAbs in ascitic fluid or in other nonblood derived biological fluids; (3) reported associations with circulating p53-AAbs and survival data, including OS, DFS, progression free survival (PFS), and relapse free survival (RFS), in both univariate and/or multivariate analyses. Moreover, reported associations with circulating p53-AAbs and baseline tumour characteristics, i.e. p53 tumour overexpression, FIGO stage disease, tumour differentiation grade, histological subtypes, and residual tumour were also commented.

Data extraction
The following data from collected studies were independently extracted by two observers (MG; MM): last name of first author, publication year, country, definition of ovarian cancer diagnosis, age, ethnicity, number of patients enrolled, method for p53-AAbs detection, p53 tumour expression, FIGO stage, tumour grading, residual tumour, tumour histotype, cut-off values, statistical tests data (contingency tables, Kaplan-Meier, Cox models) in univariate and/or multivariate analysis with hazard ratio (HR) or relative risk (RR), 95% confidence interval (CI) and p values. When the above information were not reported in the original study, the items were treated as "Not Available (NA)". Multiple studies published by the same author(s) were checked for overlap of included case subjects. Inconsistencies in the research process, they were solved by discussion. The quality of the included studies was assessed by the Newcastle-Ottawa Scale (NOS) (http://www.ohri.ca/programs/clinical_epidemiology/oxford.asp). If a study did not clearly mention one of these key points, we considered that the point was not covered in the study, and the results may have underestimated the reported characteristics.

Statistical analysis
To estimate the association between the presence of serum p53-AAbs (+p53-AAbs) and/or the absence of p53-AAbs (-p53-AAbs) on survival of ovarian cancer patients, only studies reporting univariate and/or multivariate HRs or RRs, corresponding 95% CI and p values, were considered. If both multivariate and univariate analyses were present, we chose the former. We provided estimations from given data (Kaplan-Meier curves) when these statistical variables were not available in an article, using methods reported by Tierney et al. [21]. Overall survival was the primary outcome, and DFS was the secondary outcome. We decided a priori to run random effect models to calculate the pooled HRs estimated and 95% CIs. Heterogeneity between studies was evaluated by means of tau-squared (τ 2 ) and inconsistency index (I 2 ) statistics. The heterogeneity (I 2 <25%: no heterogeneity; I 2 = 25-50%: moderate heterogeneity; I 2 >50% or p>0.01: large or extreme heterogeneity) [22]. To verify the presence of publication bias, funnel plots and Egger's linear regression test were performed [23].
The robustness of the combined results was assessed by sensitivity analysis in which studies were removed one by one each time and the pooled HR was recalculated after the exclusion to identify the studies causing considerate fluctuation of HR estimates. All the p values were two sided, and p<0.05 was considered significant, except for Egger's test where we considered a p<0.1 as statistically significant. All statistical analyses were performed using STATA 10.1 (STATA Corp., college Station, TX).

Search results and study selection
By means of the above described search strategies, we initially identified 85 articles of which 68 were discarded after reading abstracts and/or full text. The remaining 17 papers were critically reviewed (systematic review). The flowchart of search strategy for articles is presented in Fig 1 and S1 Checklist. The main clinical-pathological characteristics of the 17 studies reporting associations between circulating p53-AAbs and ovarian cancer outcome are shown in Table 1.

Study characteristics
Overall serum samples were collected before surgical treatment or within 6 months from diagnosis [25]. In some studies, samples were collected also at different times during follow-up [26,27,33,38]. The presence of autoantibodies against p53 was measured in plasma samples in only one study [11], while it was also evaluated in ascitic fluid in another one [31].

Publication Bias Assessment and Sensitivity Analysis
Visual assessment of the funnel plot suggested some publication bias with a positive asymmetry for all the 7 studies included in pooled OS (Fig 4A).
The Egger's test showed some evidence of small study effect for pooled OS (p = 0.096) ( Fig  4B). Sensitivity analysis was performed to evaluate the stability of results. No study was found to significantly affect either the pooled HRs for OS (Fig 5A) or the pooled HR for DFS (Fig 5B). When studies concerning p53-AAbs and OS were grouped by ethnicity excluding the only Asian study [32] by sensitivity test, the heterogeneity was still high (I 2 = 78.1%, τ 2 = 0.629, p<0.001) and the +p53-AAbs was associated with a better OS though not reaching statistical significance (pooled HR = 0.89 for +p53-AAbs vs -p53-AAbs, 95% CI: 0.47-1.70). Due to the low number of studies, the Egger's test was not applied for publication bias assessment in pooled DFS. Nonetheless, the funnel plot showed some positive asymmetry (not shown). The sensitivity analysis for DFS, showed that results did not change significantly (Fig 5B).

Discussion
This paper summarised for the first time the prognostic role of serum p53 autoantibodies in ovarian cancer. To our knowledge, this is the first systematic review and meta-analysis on this topic. The usefulness of p53-AAbs as prognostic biomarkers in clinical outcome of ovarian cancer, in particular, has been poorly investigated; the role of p53-AAbs is still controversial but critical in understanding their function in the immune surveillance of cancer [40].
The presence of circulating autoantibodies against specific tumour-associated antigens (TAA) is generally found in less than 1/3 of cancer patients, resulting in poor diagnostic sensitivity [41]. Autoantibodies reflect both changes in the primary tumour as well as effective cancer immune surveillance, and may represent candidates for immunotheraphy development if they are associated to improved prognosis [40]. Nonetheless, AAbs can be detected up to 5 years before symptomatic disease, demonstrating that the human immune system recognizes the autologous TAAs as "non-self" producing an earlier humoral response in the patients [42]. Combination of panels of TAAs to detect multiple different and specific AAbs is the current goal to increase their diagnostic potential [20]. Anyway, recent evidences suggest that serum p53-AAbs can be considered as biomarkers to detect many types of cancer, including ovarian cancer [17]. p53-AAbs are usually IgG indicating a secondary response after a prolonged immunization process probably before the diagnosis of the malignancy [43]. The evidence that in healthy subjects p53-AAbs are extremely rare [17] explains the concept that cancer cells carrying mutations in TP53 gene and with p53 over-expression are the source of a self-immunization process. However, considering that only a subset of cancer patients (~20-50%) carrying TP53 somatic mutations have detectable p53-AAbs, genetic variants in TP53 alone are probably not sufficient to trigger the p53-AAbs secretion, but mutations of p53 regulators and nonmutative pathways are probably engaged as well [9]. Furthermore, in ovarian cancer the correlation between p53 accumulation, tumour grade and p53-AAbs detection has not always been found [44]. Overall, these observations suggest that the biological and immunogenetic background of individuals, such as the set of major histocompatibility complex (MHC) classes I and II molecules, should be considered in the induction of an anti-p53 specific humoral response [36,[44][45][46][47].
As represented in Fig 1, to examine the correlation between p53-AAbs and survival, only 7 studies were eligible for meta-analysis from 17 candidate articles (Table 1), via systematic review. These results confirmed the suggestion that the presence of autoantibodies against p53 in the serum of ovarian cancer patients has been poorly investigated in the last 20 years.
To define the prognostic role of p53-AAbs as biomarkers for clinical utility, they must be independent of known clinic-pathologic criteria, and well-established validation assays tested in large cohorts of patients and control populations are also required. Moreover, an accurate clinical history of patients and long term follow-up are necessary. Among the 17 studies (Table 1) initially selected, 53% (9/17) of them did not carry out an investigation in a control group of subjects, i.e age matched healthy donors and/or patients with cystadenomas and/or benign gynaecological diseases [12,25,27,29,31,33,35,36,38]. This point could affect the quality of p53-AAbs detection when a non commercial assay was used. Moreover, blood samples should be collected prior to treatment (first surgery and/or chemotherapy) even if an immunological memory exists. Finally, data should be analyzed also considering sub-groups of patients who received uniform therapies. We observed that in different studies the presence of p53-AAbs was significantly associated to overexpression of p53 in the tumour (67%), to III-IV stages (60%), and to G2-G3 tumours (50%). Among the 7 articles [12,25,29,30,32,35,37] selected for meta-analysis, in 5 [25,30,32,35,37] a non commercial ELISA test was used to measure serum p53-AAbs and a matched control group was considered only in 3 studies [30,32,37]. Data on the chemotherapy treatment were not reported in 2 papers [35,32] and not described in detail in another one [25]. Ten studies showed a tendency toward a poor outcome (Table 1): 8 articles were associated to shorter DFS/PFS/RFS [12,24,26,27,29,30,33,38] and 5 to decreased OS [12,[28][29][30]32]. Three studies showed an association with a favourable outcome [25,35,37]. The presence of serum p53-AAbs was statistically significantly associated to a reduced DFS in ovarian cancer only in univariate analysis in 2 studies [12,25]. In multivariate analysis, +p53-AAbs failed to be an independent prognostic factor in both studies, one was associated to a favourable DFS [25]. However, HR estimates were not consistent between the univariate and multivariate analyses in this paper [25] in both OS and DFS. In the final pooled analysis with 4 eligible studies [12,25,29,30], we found an association with a reduced DFS and a moderate heterogeneity.
Concerning OS, no associations between p53-AAbs and survival were found in 4 studies [11,31,34,36]; anyway, data about OS in univariate analysis (log-rank p value) were reported for 2 articles [11,36]. As described in the Results, for the final meta-analysis, 7 studies [12,25,29,30,32,35,37] were eligible for OS evaluation. Only 5 studies [12,25,30,35,37] presented results about multivariate estimations. Detectable serum p53-AAbs were significantly associated to survival after adjustment for ovarian cancer main prognostic factors in 3 studies [30,35,37], even if, one [30] did not report the estimated HR and 95% CI. Three studies [25,35,37] were associated to a better outcome, however, one of these did not reach significance [25]. When we considered the 7 studies overall, we found a large heterogeneity and no association with the presence of the p53-AAbs and OS. As suggested by the Egger's test and the funnel plot, we found evidences for publication bias and for some small study effect [32], which could explain the large heterogeneity in the OS analyses. Intriguingly, when the 4 studies reporting the multivariate HRs estimations [12,25,35,37], adjusted for known prognostic factors that may have influenced survival were pooled together, no significant heterogeneity was detected, and the presence of autoantibodies was significantly associated to a better OS. In particular, we observed that the latest studies restricted the survival analyses only to patients with advanced stage [35] and serous histotype [37] and adjusted the multivariate analyses for multiple factors (Table 2), finding an association with a better survival. The stratification used seemed logical since the presence of p53-AAbs proved to be associated to these pathological parameters [11,12]. Furthermore, authors adjusted multivariate analyses with a more complete panel of variables, compared to the other 2 studies. Some biologically plausible mechanisms may explain the p53-AAbs potentially direct or indirect protective role in ovarian cancer. p53-AAbs appearance in serum is a product of a natural immunization process detectable only in a subset of patients, particularly with advanced stage disease. An anti-p53-specific IgG autoantibodies may induce amplification of specific p53-T-cell memory response [48], but the function of these autoantibodies is still unknown [41].

Conclusion
This review included all studies reporting univariate or multivariate estimates of ovarian cancer prognosis linked to p53-AAbs. Our study suggested that serum p53-AAbs have a controversial prognostic role in ovarian cancer, although their presence was significantly associated to an improved OS only at multivariate analyses. Autoantibodies were associated with a worse, although not significant DFS, while no association was observed including only multivariate HRs. Nonetheless, the following limitations should be considered to interpret meta-analyses results: (1)-the low number of studies included in the meta-analysis, especially for DFS; (2)univariate and multivariate HRs were pooled together and this may explain the wide heterogeneity observed; (3)-presence of autoantibodies against p53 was determined by means of different types of ELISA assays (commercial or not) thus, misclassification is possible; (4)-the p53-AAbs detection was not validated in a matched control group in most of eligible studies; (5)-Caucasian/white female patients who may limit the comparison of our results to other populations.
However, the meta-analysis has several strengths, including the homogeneity of ovarian cancer diagnosis, the prospectively collected OS and DFS data, and the univariate and multivariate estimations of HRs. Furthermore, measures of p53-AAbs levels were based on blood samples collected shortly after diagnosis and prior to primary surgery. This review shows that, to date, p53-AAbs have had a limited clinical application. Our conclusions are based on few investigations thus, should be considered carefully. Further researches in large patient's cohorts are needed to explore the role of the natural immunity process against the oncogenic p53 protein in ovarian cancer.