Genetic polymorphisms and platinum-induced hematological toxicity: a systematic review

Background Platinum-based chemotherapy bring severe hematological toxicity that can lead to dose reduction or discontinuation of therapy. Genetic variations have been reported to influence the risk and extent of hematological toxicity; however, the results are controversial and a comprehensive overview is lacking. This systematic review aimed to identify genetic biomarkers of platinum-induced hematological toxicity. Method Pubmed, Embase and Web of science database were systematically reviewed for studies that evaluated the association of genetic variants and platinum-related hematological toxicity in tumor patients with no prior history of chemotherapy or radiation, published from inception to the 28th of January 2022. The studies should have specific toxicity scoring system as well as defined toxicity end-point. The quality of reporting was assessed using the Strengthening the Reporting of Genetic Association Studies (STREGA) checklist. Results were summarized using narrative synthesis. Results 83 studies were eligible with over 682 single-nucleotide polymorphisms across 110 genes. The results are inconsistent and diverse with methodological issues including insufficient sample size, population stratification, various treatment schedule and toxicity end-point, and inappropriate statistics. 11 SNPs from 10 genes (ABCB1 rs1128503, GSTP1 rs1695, GSTM1 gene deletion, ERCC1 rs11615, ERCC1 rs3212986, ERCC2 rs238406, XPC rs2228001, XPCC1 rs25487, MTHFR rs1801133, MDM2 rs2279744, TP53 rs1042522) had consistent results in more than two independent populations. Among them, GSTP1 rs1695, ERCC1 rs11615, ERCC1 rs3212986, and XRCC1 rs25487 present the most promising results. Conclusion Even though the results are inconsistent and several methodological concerns exist, this systematic review identified several genetic variations that deserve validation in well-defined studies with larger sample size and robust methodology. Systematic Review Registration https://www.crd.york.ac.uk/, identifier CRD42021234164.


Introduction
Platinum agents, including cisplatin, carboplatin and oxaliplatin, are used effectively against various tumor diseases either as monotherapy or in combination with other chemotherapeutics, radiation therapy and/or surgery.However, they display a range of severe side effects due to their poor selectivity for cancerous tissue over normal tissue.Hematological toxicity caused by platinum drugs are those that affect bone marrow function and blood cell production, characterized by leukopenia, neutropenia, thrombocytopenia, and anemia (Oun et al., 2018).Leukopenia or neutropenia, can leave patients susceptible to infections.Platinum-induced anemia is persisting erythropoietin deficiency state correlated with renal tubular dysfunction (Wood and Hrushesky, 1995).Acute myelosuppression occurs shortly after chemotherapy, while residual bone marrow injury manifested by a decrease in hematopoietic stem cell reserves or a myelodysplastic syndrome (Wang et al., 2006).
All three platinum drugs can cause some form of hematological toxicity, and myelosuppression is the doselimiting toxicity of carboplatin.In the majority of cases, neither cisplatin nor oxaliplatin is associated with severe myelosuppression (Rabik and Dolan, 2007).Carboplatin induced myelosuppression resulted in neutropenia and thrombocytopenia.Severe (grade 3 or 4) neutropenia occurs in approximately 18% of carboplatin-treated patients, whereas severe thrombocytopenia occurs in approximately 25% of cases (Go and Adjei, 1999).Cisplatin-induced hematological toxicity is usually mild at intermittent doses of 50-60 mg/m 2 and myelosuppression presents in 25%-30% of patients (Prestayko et al., 1979).Myelosuppression caused by oxaliplatin is generally mild.Grade 3/4 anemia, neutropenia and thrombocytopenia are observed in only 2%-3% of patients (Hartmann and Lipp, 2003).Hematological toxicity aggravates when platinum agents were combined with other cytotoxic drugs.The degree of hematological toxicity varies upon different chemotherapy regimen.For example, hematological toxicity was more profound in lung cancer patients treated with platinum agents plus gemcitabine (Fisher and D'Orazio, 2000;Schiller et al., 2002).
Inhibition of cell proliferation is one of the major causes of platinum-induced myelosuppression and related complications.The cytotoxicity of platinum on hematopoietic stem cells is attributed to its highly reactive hydrated platinum complex that binds to DNA and form intra-and inter-strand crosslinks; thereby produce subsequent interference with DNA transcription and/or DNA replication (Das et al., 2008).The generation in oxidative stress is also responsible for platinuminduced bone marrow toxicity (Basu et al., 2015).Increased platinum influx, decreased platinum efflux, impaired cell detoxification, low or absent DNA damage repair and activated cell death signaling may be the reasons of platinum-induced hematological toxicity (Shaloam and Paul, 2014) (Figure 1).
Identifying patients at greatest risk for these complications would be clinically useful for selecting patients for chemotherapy and planning the frequency of monitor and clinical treatment with colony-stimulating factor.Risk factors for hematological toxicities include kidney function, age, drug doses, combination chemotherapy, a poor performance status and prior chemotherapy exposure (Hartmann and Lipp, 2003;Ouyang et al., 2013).Furthermore, genetic variations in genes encoding proteins involved in pharmacokinetic and pharmacodynamic processes influence the occurrence and extent of adverse reactions (Zheng et al., 2020).Although several genetic polymorphisms have been identified to influence platinuminduced hematological toxicity, a comprehensive overview is lacking.We here provide an overview to identify which genetic variants consistently associated with hematological toxicity and discuss limitations of current pharmacogenetic analyses and formulate directions for further research.

Study eligibility
The systematic review was reported according to the Preferred Reporting Items for Systematic reviews and Meta-Analyses (PRISMA) checklist (Page et al., 2021) (Supplementary Table S1).The protocol was registered in the international prospective register of systematic reviews (PROSPERO; Registered number: CRD42021234164).The inclusion criteria were (Oun et al., 2018): studies that focus on the association between hematological toxicity and genetic polymorphisms (Wood and Hrushesky, 1995); studies including cancer patients using platinum-containing chemotherapy (Wang et al., 2006); studies that have specific toxicity scoring system and defined toxicity end-point.The exclusion criteria include the followings (Oun et al., 2018): preclinical studies (animal experiment or in vitro studies) (Wood and Hrushesky, 1995); studies in which patients were treated with concurrent radiotherapy (Wang et al., 2006); studies that were non-English, case report, review or metaanalysis (Rabik and Dolan, 2007); studies in which patients have prior history of chemotherapy and/or radiation.

Search strategy
PubMed/MEDLINE, EMBASE and Web of Science were searched for publications from inception to the 28th of January 2022.The literature search was conducted using Medical Subject Headings and combinations of relevant keywords.The detailed search strategy can be found in Supplementary Table S2.Additional research papers were identified by screening the reference sections of included articles.Two authors (YZ and MT) independently performed the data screening.Disagreements were consulted with a third arbiter (ZD).

Quality assessment
The quality of the studies will be assessed using a scoring system modified from a previously published study (Leusink et al., 2016) based on STREGA recommendations (Little et al., 2009) Supplementary Table S3.The scoring system contains ten items on five domains: clinical information, genotyping, study population origin, sample size and statistical correction for multiple testing, and study analysis.Each study included in this review was assessed for quality as good (overall quality score:7-10), moderate (overall quality score:4-6), or poor (overall quality score≤3) based on scores.Two reviewers (YZ and MT) will assess the quality independently and a third reviewer (PC) will be consulted in case of disagreement.

Data collection and analysis
The following data were extracted from each publication by two authors (YZ and MT): author, year, source of study (reference), sample size, ethnicity, type of cancer, number of treatment cycles, treatment schedule, toxicity scoring system, defined toxicity end-point, genetic polymorphisms involved and main study results.
Due to the heterogeneity in the patient population, treatment schedule, outcome definitions, meta-analysis was not appropriate.Studies were analyzed using a narrative synthesis approach.

Study selection
The initial search delivered 2057 articles; after removal of duplicates, 1156 abstracts were primarily screened of which 207 full-text articles remained.After reading the full-text, 83 studies were eventually included in the present systematic review.The article selection process is shown in Figure 2.
64 studies used mixed combination chemotherapy: cisplatin/ carboplatin-based chemotherapy (n = 51), cisplatin-based chemotherapy (n = 10), carboplatin-based chemotherapy (n = 3), 19 study uses single platinum-based chemotherapy.Platinum dosage and cycles varied according the tumor type: mostly cisplatin 75 mg/m 2 or carboplatin AUC 5, both administered on day 1 every 3 weeks.The dose of oxaliplatin was 85 mg/m 2 or 130 mg/m 2 29 studies did not mention the dose of platinum agents.

Metabolism
Platinum compounds can be detoxified by conjugation with glutathione through the aid of glutathione S-transferases (GSTs) (R et al., 2000).GSTP1, GSTM1, and GSTT1 belongs to Human GSTs and were mostly analyzed for the functional polymorphisms in gene regions.

NER pathway
The bulky DNA intra-strand adducts generated by platinum are mainly repaired by the nucleotide-excision repair (NER) pathway (Clarissa Ribeiro Reily et al., 2018).

BER pathway
The base excision repair (BER) pathway is mainly responsible for removing the oxidative DNA lesions generated by platinum drug exposure (Jana et al., 2018).

DSB pathway
Homologous recombination (HR) and non-homologous and joining pathways are responsible for repairing the DSBs generated by platinum-induced ICLs which are the most hazardous type of DNA damage.XRCC3 are one of the crucial proteins involved in mediating the HR pathway (Giovanna and Massimo, 2019).XRCC3 rs861539(C>T, Thr241Met) was show to reduce DNA damage repair capacity (Matullo et al., 2001).However, no association was detected in three analyses on XRCC3 rs861539 (Thr241Met) (Ludovini et al., 2011;Ruzzo et al., 2014;Zheng et al., 2017).

High throughput researches
Four studies are derived from the same cohort in Sweden population of 215 NSCLC patients treated with gemcitabine/ carboplatin chemotherapy (Gréen et al., 2016;Svedberg et al., 2020;Björn et al., 2020a;Björn et al., 2020b).Three of them were whole-exome sequence studies concentrating of gemcitabine/carboplatin-induced grade 3-4 thrombocytopenia (Gréen et al., 2016;Björn et al., 2020a), leukopenia (Svedberg et al., 2020) and neutropenia (Gréen et al., 2016;Svedberg et al., 2020).The fourth study performed a GWAS in a subset of 96 patients (Björn et al., 2020b).These studies identified and validated several genetic variations, genes and hematopoiesisrelated pathways to be potential significance and created weighted genetic risk score (wGRS) prediction models for predicting the risk of chemotherapy-induced hematological toxicity.There were GWAS concentrating on NSCLC (Cao et al., 2016), cervical cancer (Huang et al., 2015), and unclassified carcinomas (Low et al., 2013).A list of novel genetic variants such as rs13014982 at 2q24.3 and rs9909179 at 17p12 were identified (Cao et al., 2016) and the artificial neural networks model based on the multiple risk factors were constructed (Huang et al., 2015).Yin et al. developed a strategy to establish a predicted model of toxicity integrating both genetic and clinical factors using DM techniques (Yin et al., 2016).
It's also worth mentioning that GSTP1 rs1695 (A313G, Ile105Val) show significant association in six studies.Among them, four studies show protective role of GSTP1 rs1695 against platinum-induced hematological toxicities, including a comprehensive analysis that targeted resequencing of 100 notable pharmacokinetics-related genes in which GSTP1 rs1695 showed the smallest p-value (p = 0.00034) (Yoshihama et al., 2018).Positive associations of GSTP1 rs1695 with increased risk of platinuminduced hematological toxicity was found in two meta-analysis (Lv et al., 2018;Kim et al., 2022), while these two meta-analysis studies face the limitations of insufficient data availability.Alteration of DNA repair ability might play an important role in the development of platinum-induced hematological toxicity.Genetic variants in the candidate NER genes may affect the repair function and are most promising in predicting platinum-related hematological toxicity, since it is the main pathway responsible of repairing the bulky DNA intra-strand adducts generated by platinum agents (Hilary et al., 2024).ERCC1 rs11615 (C118T, Asn118Asn) and ERCC1 rs3212986 (C8092A) are two common variant that affect ERCC1 (key enzyme in NER pathway) mRNA expression or mRNA stability (Chen et al., 2000;Yu et al., 2000).ERCC1 rs11615 present consistent results with increased risk of grade 3-4 neutropenia or anemia in four studies, and ERCC1 rs3212986 show decreased risk of grade 3-4 hematologic toxicity in two studies.These two SNPs may be important molecular biomarkers for predicting platinum-induced hematological toxicities.XRCC1 rs25487 (G23885A, Arg399Gln) show positive association in six studies, emphasizing the potential contribution of BER pathway in oxidative stress in platinum-induced hematological toxicity, since cisplatin can exert cytotoxic effects through the generation of ROS (Zheng et al., 2020).
Apart from that, lots of genetic variants located in genes not directly linked to drug exposure are investigated.The GWAS have identified a handful of candidate genetic variants associated with platinum-based hematological toxicities and novel biologic pathways of potential impact.But these studies still face the challenge of statistically underpowered and stringent threshold of multiple testing.Additional validation in multiple independent sample sets or functional analyses are required to further elucidate the gene-phenotype relationship (Low et al., 2014).

Quality and inconsistency among studies
Genetic association studies require a large number of patients to provide adequate power, as a rare variant with large effect, or common variant with modest effect is more probable in genetic epidemiology (Campbell et al., 1995).While the majority of the studies in our systematic review did not indicate the sample size calculation in their statistical analysis, and the sample size in most studies in our systematic review are much smaller than that would be implied, which may be underpowered to detect a statistically significant relationship (Jorgensen and Williamson, 2008).
One of the challenges in pharmacogenomics is the ethnic background of the study population.The prevalence of toxicity varies according to the ethnic background (O'Donnell and Dolan, 2009;Li and Meyre, 2013).For example, higher rates of toxicities have been observed in east Asian populations compared to European and North American populations (Watanabe et al., 2003).In addition, allele frequency of genetic variants vary depending on the ethnic background or even the geographical location.When cases and controls are drawn from multiple ethnic or geographic groups, population stratification exists, which may put the study at risk of confounding and can lead to false positive associations (Jorgensen and Williamson, 2008).Population stratification was seldom assessed in most genetic association studies, which has been cited as a major reason for lack of replication.
Different treatment protocols are recommended according to cancer types.The composition, proportion and cycles of chemotherapy regimen can influence the incidence and degree of hematological toxicity.Furthermore, treatment protocols may also differ with regard to the use of concomitant supportive treatments.The time and dose of granulocyte colonystimulating factor administered may be various across institutions, which are seldom mentioned in the method part and may bring confounds in genetic association studies.Moreover, there are overlap in metabolic pathway between platinum and other antineoplastic drugs, which can alter the pharmacogenetic effects of polymorphisms.For example, The ABC transporter ABCB1 and ABCC2 is responsible for the efflux of many commonly used antineoplastic drugs that usually used in combination with platinum agents, including taxanes (Marsh, 2006;Lambrechts et al., 2015).If Pt-DNA lesions are not repaired, the DNA lesion triggers activation of the apoptosis pathway, an essential step for the effectiveness of platinum-based chemotherapeutics for killing tumor cells, which is also the apoptosis pathway of many other cytotoxic drugs (Fulda and Debatin, 2006).Apart from the role of participating in DNA synthesis, MTHFR encode key enzymes for the metabolism of 5-FU (Lee et al., 2013) and RRM1 was also the primary target for gemcitabine (Yuan et al., 2015).
Genetic association studies are heavily reliant on the phenotype, but it may be difficult to establish the true phenotype.Although myelosuppression is quantitative, the degree of myelosuppression could be missed based on the frequency of measurement.Apart from that, differences in the toxicity assessment criterion and endpoints may hamper reproducibility of previous findings.Some studies analyzed the total hematological toxicity (Wang et al., 2008;Kim et al., 2009;Gu et al., 2012), while other studies analyzed the detail hematological toxicity (leukopenia, neutropenia, thrombocytopenia or anemia) (Isla et al., 2004;Tibaldi et al., 2008;Xu et al., 2012).The detail hematological toxicity may be more accurate and clinically relevant, but its low incidence requires cohort study with large sample size, which are unattainable in the current researches.Furthermore, methodological flaws were observed in the imprecise dichotomization of patients with mild toxicity and severe toxicity.The majority use the occurrence of grade 3-4 as endpoints, other studies use grade 2-4 (Isla et al., 2004;Erčulj et al., 2012;Huang et al., 2015;Jia et al., 2016;Liblab et al., 2020) or grade 1-4 (Chen et al., 2010;Erčulj et al., 2012;Deng et al., 2015;Walia et al., 2022).Medical interventions of dose reduction and/or treatment discontinuation were taken when grade 3-4 hematological toxicity arises.Therefore, the toxicity endpoints of grade 2-4 or grade 1-4 may not be clinically relevant (Supplementary Table S4).
There is statistical heterogeneity in the data analysis (Supplementary Table S5).Some studies just used chi-square test or fisher's exact test to estimate the difference of genotype distribution in cases and controls (Kimcurran et al., 2011;Kalikaki et al., 2015).Others use logistic regression to make comparisons between groups and generate odds ratios and 95% confidence intervals.Logistic regression is more appropriate as it can provide a quantitative measure of the relationship between the groups, allow adjustment for confounding factors, and detect gene-gene or gene-environment interactions.Some studies made a clear statement about mode of inheritance assumed for analysis, and used more than one assumption (Peng et al., 2014;Bushra et al., 2020;Ferracini et al., 2021;Walia et al., 2022), while other studies only compared the three categories of genotype frequencies (homozygous wild type, heterozygous, homozygous variant) between cases and controls (Isla et al., 2004;Han et al., 2007;Kim et al., 2009;Lee et al., 2013;Liblab et al., 2020).30 of the included studies performed the correction for multiple comparisons and only 10 studies performed validation of the results or by splitting the cohort for a primary and an exploratory analysis (Marsh et al., 2007;Qian et al., 2012;Cao et al., 2016;Gréen et al., 2016;Jia et al., 2016;Yin et al., 2016;Zheng et al., 2017;Björn et al., 2020a;Björn et al., 2020b;Svedberg et al., 2020).

Limitations
This review has several limitations, which mainly reflects the status of current genetic association studies.In the search process, many studies had to be excluded because hematological toxicity was not clearly described or there is no toxicity grading criterion.What's more, we observed that the most studies did not provide toxicity statistics data (no p-value) or just provide insufficient toxicity data (for example, some studies only mentioned correlation in the results section without providing the original statistical data), which may raise some doubts of data authenticity.Also, some studies did not control exclusion criteria of prior history of chemotherapy and/or radiation or a considerable proportion of the patients receive chemotherapy combined with radiation or other non-platinum chemotherapy regimens, which may increase the likelihood of cumulative hematological toxicity.There is much heterogeneity between incorporated studies, thus we were unable to perform a quantitative comparison and meta-analysis.

Implications for clinical practice and research
Future studies should focus on the following aspects.Firstly, genetic association study inevitably faces the concerns of bias and confounding, and are susceptible to inappropriate conclusions, therefore it calls for careful planning of study design to improving quality of methodology (Saito et al., 2006).Secondly, the majority genetic polymorphisms identified in the eligible publications were repeated in only one or two studies (Hilary et al., 2024).SNPs screened out as potential factors in susceptibility to hematological toxicity in our systematic review require well-planned, methodologically robust studies to validate them.Thirdly, functional predictions of significant genetic variants needed to be confirmed and validated in vitro or in vivo work (Felipe Antonio de Oliveira et al., 2022).Fourthly, more studies should emphasize on the hematopoiesis-related pathways identified in a whole-exome sequenced study of 215 NSCLC patients treated with a single treatment (Björn et al., 2020a).Moreover, clinical trial with large sample size to perform subgroup analysis should be conducted to allow proper stratified analysis.Finally, owing to the complexity of mechanisms of platinum action, a single SNP alone may have low effect to platinum response, thus supporting a polygenetic effect in platinum-induced hematological toxicity (Daniel et al., 2023).Future direction should be establishing appropriate statistical methods with capacity to integrate multiple genetic, phenotypic, epidemiological and clinical variables effects.

Conclusion
To summarize, this systematic review has successfully reported and evaluated studies on genetic associations of platinum-based hematological toxicities.Review of these studies identified several genetic variants that potentially affect the risk of platinum-induced hematological toxicity.Many methodological issues exist that may affect reproducibility of results and lead to inconsistency, including insufficient sample size, population stratification, various treatment schedule, heterogeneity in the assessment of hematological toxicity and statistics.Well-designed studies with sufficient samples sizes and standardization of phenotypes are warranted to address the limitations of the current studies and to ensure the robust findings that can be more effective to be used in personalized therapeutics.

TABLE 1
Overview of pharmacogenetic studies on platinum-induced hematological toxicity.

TABLE 1 (
Continued) Overview of pharmacogenetic studies on platinum-induced hematological toxicity.

TABLE 1 (
Continued) Overview of pharmacogenetic studies on platinum-induced hematological toxicity.

TABLE 1 (
Continued) Overview of pharmacogenetic studies on platinum-induced hematological toxicity.

TABLE 1 (
Continued) Overview of pharmacogenetic studies on platinum-induced hematological toxicity.

TABLE 1 (
Continued) Overview of pharmacogenetic studies on platinum-induced hematological toxicity.

TABLE 1 (
Continued) Overview of pharmacogenetic studies on platinum-induced hematological toxicity.

TABLE 2
Genetic polymorphisms investigated more than twice for association with platinum-induced hematological toxicity.

TABLE 2 (
Continued) Genetic polymorphisms investigated more than twice for association with platinum-induced hematological toxicity.

TABLE 2 (
Continued) Genetic polymorphisms investigated more than twice for association with platinum-induced hematological toxicity.

TABLE 3
Summary of positive associations in genetic polymorphisms that investigated more than twice.

TABLE 3 (
Continued) Summary of positive associations in genetic polymorphisms that investigated more than twice.

TABLE 3 (
Continued) Summary of positive associations in genetic polymorphisms that investigated more than twice.

TABLE 3 (
Continued) Summary of positive associations in genetic polymorphisms that investigated more than twice.

TABLE 3 (
Continued) Summary of positive associations in genetic polymorphisms that investigated more than twice.