Hypomagnesemia and clinical benefits of anti-EGFR monoclonal antibodies in wild-type KRAS metastatic colorectal cancer: a systematic review and meta-analysis

Hypomagnesemia is a recognized side-effect of cetuximab- or panitumumab-based chemotherapy for metastatic colorectal cancer (mCRC). The clinical relevance of hypomagnesemia is under debate. Thus, a systematic review and meta-analysis of retrospective studies and randomized clinical trials (RCTs) comparing hypomagnesemia with normal magnesium levels in wild-type KRAS mCRC was performed. One RCT, two retrospective studies, and two American Society of Clinical Oncology (ASCO) and European Society for Medical Oncology (ESMO) conference presentations from phase III RCTs involving 1723 patients were included in this study. Patients with hypomagnesemia demonstrated better progression-free survival (PFS) (Hazard ratio [HR]: 0.64; 95% confidence interval [CI]: 0.47–0.88), overall survival (OS) (HR: 0.72; 95% CI: 0.53–0.92), and objective response rate (ORR) (Risk ratio [RR]: 1.81; 95% confidence interval [CI]: 1.30–2.52). By subgroup analysis, frontline, later lines or combination therapy with hypomagnesemia were associated with PFS benefits (HR: 0.78; 95% CI: 0.62–0.98; HR: 0.60; 95% CI: 0.40–0.90; HR: 0.62; 95% CI: 0.41–0.94, respectively). In patients with wild-type KRAS mCRC, hypomagnesemia is associated with better clinical benefits of PFS, OS and ORR when treated with cetuximab- or panitumumab-based chemotherapy. Future clinical trials should corroborate its predictive role.

does not always guarantee clinical benefits with anti-EGFR MoAs 4,[11][12][13] . Besides, there are proposed positive predictive markers in patients with wild-type KRAS, such as EGFR amplification and overexpression of EGFR ligands [14][15][16] . Through both negative and positive selection that makes it possible to define the population that benefits more from these anti-EGFR MoAs.
Hypomagnesemia is an appreciated side effect of cetuximab and panitumumab treatment, as reported in randomized trials and meta-analysis [17][18][19][20] . Though magnesium is essential to organism, how hypomagnesemia relates to efficacy of these anti-EGFR MoAs and what effect magnesium levels have on cancer progression is an uncertain area of research with contradicting results [21][22][23][24][25][26] . Whether hypomagnesemia is associated with superior or inferior outcomes remains equivocal in these retrospective or prospective clinical trials. To elucidate the role and clinical benefits of hypomagnesemia in wild-type KRAS mCRC, a systematic review and meta-analysis of current clinical trials was performed.

Methods
Data source and search strategy. This study involved a comprehensive search of all published eligible studies and conference presentations that reported the PFS, OS, and ORR for wide-type KRAS mCRC patients with hypomagnesemia after treatment with cetuximab-or panitumumab-based chemotherapy. The searched included MEDLINE, EMBASE, and the Cochrane Central Register of Controlled Trials (CENTRAL) for relevant trials from inception to 01 April 2017. The search strategy were "(hypomagnesemia OR magnesium reduction) AND (cetuximab OR panitumumab OR EGFR) AND (survival OR outcome OR efficacy) AND (colon cancer OR rectal cancer OR colorectal cancer)" without restrictions for language and gender. To identify unpublished studies, the US National Institutes of Health trials register (http://clinicaltrial.gov) and conference abstracts from proceeding of the American Society of Clinical Oncology (ASCO) and the European Society for Medical Oncology (ESMO) were also searched. The ethical approval was waived because all analyses were based on previous published studies.
Inclusion and exclusion criteria. Studies or conference presentations that investigated wild-type KRAS mCRC treated with cetuximab-or panitumumab-based chemotherapy were included. Those with available data for the events and incidences of hypomagnesemia and those with available Hazard ratio (HR), 95% confidence interval (CI), or Kaplan-Meier survival analysis that compared outcomes in the hypomagnesemia and normal magnesium level groups were also included. The references from included trials were also checked to identify relevant trials. Studies that did not meet these criteria were excluded.
Data extraction and quality assessment. Two authors independently extracted the available data from the included trials using a standardized data collection form: first author, year of publication or conference presentation, study design, intervention type, study population per group, clinical setting (type of chemotherapy, definition of hypomagnesemia), and outcome data (PFS, OS, and ORR).
Hypomagnesemia was defined according to the CTCAE (Common Terminology Criteria for Adverse Events) or magnesium reduction level 27 . Grade 1 hypomagnesemia was from the lower limit of normal to 1.2 mg/dL, Grade 2 was <1.2 to 0.9 mg/dL, Grade 3 was <0.9 to 0.7 mg/dL, and Grade 4 was <0.7 mg/dL. Those with grades 1-4 hypomagnesemia or > 50% magnesium reduction was arbitrarily classified as the hypomagnesemia group. Observed differences in ORR that could also predict clinically relevant therapeutic benefits and evaluations were done in most trails, according to the RECIST (Response Evaluation Criteria in Solid Tumours) guidelines 28 .
A third author resolved discrepancies in opinion. The quality of methodology was assessed using the Cochrane Collaboration tool for randomized controlled trials (RCTs) and methodological index for non-randomized studies (MINORS) 29,30 .

Statistical analysis
The Review Manager 5.3 (The Nordic Cochrane Centre, The Cochrane Collaboration, 2014) was used for meta-analysis. The HRs and 95% CIs were extracted for all included trials and subgroups. Data from independent assessment of PFS and OS were pooled together for time-to-event outcomes analysis when both types of reviews were available. A random-effects (RE) model was used to calculate pooled HRs, 95% CIs, and p values 31 . A generic inverse variance (IV) meta-analysis was analyzed using log HRs and standard errors (SEs), estimated as described in the Cochrane handbook for systemic reviews of intervention, if the HRs is quoted in a trial together with a 95% CIs or p values 30 . If the HRs, 95% CIs, and p values were not reported, the HRs and SEs were estimated from Kaplan-Meier survival curve, patient number, censored data, or number at risk table reported 32 . For ORR analysis, risk ratio (RR) and 95% CI for binomial data and dichotomous meta-analysis were calculated with Mantel-Haenszel test under the random-effect model. A two-sided p < 0.05 was set as statistically significant.
For heterogeneity tests, the chi-square (chi 2 ) and I 2 inconsistency statistics were used 33,34 . A p < 0.10 was considered significant heterogeneity. Moreover, I 2 values of 0-24.9%, 25-49.9%, 50-74%, and 75-100% were considered as none, low, moderate, and high heterogeneity, respectively 33,34 . Data availability. All data generated or analyzed during this study are included in these published article or unpublished meeting abstracts (ref 22

Results
Eligible studies. The PRISMA study flow diagram were shown in Fig. 1 35 . A total of 157 records were identified after searching the MEDLINE, EMBASE, and CENTRAL databases. Another 16 records were identified from the United States National Institutes of Health trials register databases and from conference abstracts of ASCO and ESMO. After removing 29 duplicates and 139 exclusions due to irrelevant results of hypomagnesemia, lack of placebo, or mixed mutant type KRAS analysis, only five trials were incorporated in this meta-analysis. These were two conference presentations of phase III RCTs 22,25 , two retrospective studies 23,26 and one phase III RCT 24 .
The original setting of CO. 17 5,6 , PRIME 7 and ASPECCT 8 were multiple-center and open label phase III RCTs that had a low risk of bias when appraised using the Cochrane Collaboration's tool 36 . As two non-randomized control trial, Vincenzi et al. 23 and Fujii et al. 26 , also had a low risk of bias when appraised using the MINORS appraisal tool 29 .

Discussion
The KRAS mutation is a deeply-rooted negative predictive marker in mCRC treated with cetuximab or panitumumab 9 neuroblastoma RAS viral oncogene [NRAS]) is proposed as an even more potent negative predictive marker 13 .
Stepwise strategy with negative 4,11-13 and positive selection [14][15][16] [17][18][19][20] . In  24 . Although they performed the analyses in a phase III trial, RCT CO.17, substantial cases (10.6%) were excluded due to missing serum magnesium levels and KRAS status compared to the ASPECCT or PRIME trials (<2.8%) 22,25 . Thus, it is immature here to interpret the differences of these two anti-EGFR MoAs in view of hypomagnesemia and clinical efficacy. Nonetheless, cetuximab and panitumumab can lead to hypomagnesemia in varying extent 18,20 . Although the mechanism of hypomagnesemia may lie in the interaction of anti-EGFR MoA, with transient receptor potential cation channel, subfamily M, and member 6 (TRPM6) in the intestine and distal collecting tubules, there is little known about the difference between cetuximab and panitumumab, such as immunogenic properties, half-life, and dosing intervals 38 . Analyses of frontline, later-lines, monotherapy, and combination therapy suggest that cetuximab and panitumumab may possess different properties related to hypomagnesemia and efficacy. The underlying mechanism of hypomagnesium and this favorable response toward cancer treatment is lacking and not well elucidated 21 , and it is inferred that hypomagnesemia induces upregulation of EGFR ligands as This study has some limitations. First, not all included trials are RCTs and some are non-full-published meeting posters or abstracts 22,25 . However, CO.17, PRIME, and ASPECCT are randomized, controlled, phase III trials in original design [5][6][7][8] . Future full publications with unpublished data are likely to be relatively modest and are unlikely to substantively change the results reported here. In addition, the trials included here are not based on an individual patient data (IPD) but abstracted data from published literatures 40 . Despite attempts to contact the authors for original data, data from the studies by Vickers et al. and Vincenzi et al. are missing. Regulatory authorities should play an important role in support such an IPD meta-analysis and re-evaluate the extent of hypomagnesemia and its predictive role in anti-EGFR MoAs. Funnel plots, Egger's test and sensitivity analysis were not performed for each pairwise comparison due to less than 10 studies included in this meta-analysis. Instead, all available studies were enrolled for analysis. Lastly, hypomagnesemia is associated with better outcomes in patients with the wild-type KRAS mCRC in this study, but this warrants further validation in extended RAS mutation population. Prospectively designed study is needed to elucidate the magnitude of hypomagnesemia and its predictive role on cetuximab and panitumumab, and how magnesium supplementation can influence these outcomes.

Conclusions
In patients with wild-type KRAS mCRC, hypomagnesemia is associated with better clinical benefits of PFS, OS and ORR when treated with cetuximab-or panitumumab-based chemotherapy. Future clinical trials should corroborate its predictive role.