Introduction

Individualized care is a key strategic focus in precision medicine with relevance across almost all facets of care. In particular, methods of enhancing anti-cancer treatment efficacy have advanced rapidly, enabling patients to receive treatments specifically tailored to their genetic traits and tumor microenvironment [1, 2]. Unfortunately, the importance of personalized supportive cancer care is less well recognized, with limited appreciation for the complex interactions that account for the significant heterogeneity seen in toxicity profiles of various anti-cancer agents.

Mucositis is a ubiquitous complication of almost all cancer therapies, with treatment-specific manifestations reflecting core pharmacological and biophysical actions of various agents and treatments [3]. Common to all therapies is the high degree of heterogeneity in the clinical manifestations of mucositis, with patients presenting along a spectrum—from mild mucosal irritation to severe, ulcerative lesions—despite equivalent tumor burden, demographic profile, and treatment specifications [4].

In their mild forms, both oral (OM) and gastrointestinal mucositis (GI-M) can be adequately managed with existing supportive care measures [5,6,7,8,9,10]. When severe (NCI CTCAE III–V), OM and GI-M are catalysts for potentially lethal complications including infection, renal insufficiency, and graft versus host disease [11,12,13]. In the setting of pelvic radiation, acute GI-M is by far the largest predictor of late GI injury [14, 15]. When neutropenia is controlled, severe mucositis is the single largest factor defining the maximal tolerated dose of anti-cancer therapy and a significant driver of dose reductions, interruptions/delays, and treatment discontinuation [16]. Severe mucositis requires intensive in-patient supportive care such as the provision of intravenous fluids and electrolytes, parenteral nutrition, and opioid analgesics and thus greatly impacts quality of life for weeks to months [17]. Early intervention is considered critical in managing the clinical burden of severe mucositis; however, it remains unclear who is at risk of developing severe symptoms. As such, there is a clear and currently unmet need to identify patients at risk of severe mucositis to enable methods of risk minimization and the provision of targeted supportive care.

To date, methods of predicting mucositis risk have focused on the individual and their treatment [2]. These approaches are limited as they are largely unmodifiable and highly dichotomous. Similarly, of the studies that have begun to illustrate the genetic nature of mucositis risk [18,19,20], the translation of these findings are plagued by substantial inconsistencies and disparities between the proportion of people that carry distinct mutations and those that develop severe mucositis. Given our understanding of mucositis pathobiology has become increasingly sophisticated [3], new methods of risk prediction must acknowledge the variety of factors that determine mucositis risk and our siloed approach to personalized supportive care must be revised.

This systematic review, coordinated by the Multinational Association for Supportive Care in Cancer/International Society of Oral Oncology (MASCC/ISOO) Mucositis Study Group, represents an in-depth evaluation of the factors that contribute to mucositis risk. We aim to provide a clear overview of the factors identified to contribute to mucositis risk, discuss new approaches to risk prediction, and provide a practical framework for future studies aiming to develop the next generation of risk prediction tools.

Methods

Search strategy, study selection, and data retrieval

PubMed and Web of Sciences were searched for papers published using keywords listed in Table S1. Papers were screened for eligibility (HRW/JB) based on title and abstract. Clinical studies conducted between January 2000 and July 2019, in which the objective was the identification of (1) predictors of OM/GI-M or (2) associations between baseline characteristics and mucositis outcomes, were included. Reviews, preclinical research, strictly interventional studies, and case studies were excluded. Full details of the search strategy can be found in the supplementary material.

Eligible papers were delegated to independent reviewers who (in a single-fashion) extracted data using a standard electronic template to define study design, cancer type and stage, treatment specifications, patient demographics, mucositis outcome measure (OM or GI-M) and assessment scale/study endpoint, blinding, participant recruitment (consecutive/non-consecutive), statistical analyses, and key findings related to mucositis prediction. Irrelevant papers were excluded before all data were synthesized by study leads (HRW/JMB).

Studies were stratified based on the following: (1) patient cohort (children (0–18 years), adults), (2) treatment modality (high-dose chemotherapy/total body irradiation for hematopoietic stem cell transplantation, chemotherapy, radiotherapy, chemoradiation), (2) mucositis type (OM, GI-M), and (3) category of risk predictor (genetic, individual, tumor, treatment, comorbidities/medication use, and miscellaneous). Information to assist with the interpretation of results was additionally extracted including oncology cohort, tumor type, treatment specifications, sample size, drop-out rate, statistical approach(es), and mucositis assessment scale. Data were synthesized and organized into tables indicating evidence for each mucositis risk predictor. Factors were assessed for their association of OM and GI-M and assigned level 1 evidence when ≤ 3 high-quality studies identified an association with mucositis risk, level 2 evidence when 3–7 studies were identified, and level 3 evidence when there were > 7 studies. Inclusion of systematic review/meta-analysis data elevated the level of evidence. Importantly, these tables are intended to provide a simple and high-level overview of the factors identified to influence mucositis risk. They are intended to guide future research, not clinical practice.

Results

A total of 11,018 papers were identified from our initial search strategy, 197 eligible for inclusion (Fig. 1). After a full-text review, further 85 papers were excluded largely due to mucositis not being a primary endpoint, no attempt of risk prediction, or case reports, leaving 113 papers for final inclusion. Studies included in this review were prospective observational, retrospective observational, and secondary analyses of previously conducted randomized control studies. Most studies were conducted in adult patients, with only ten predicting mucositis in children.

Fig. 1
figure 1

PRISMA flow chart

Full size image

The majority of studies evaluated a single risk prediction category (89/113) with cancer regimen-related factors the most common category. Studies evaluating genetic predictors were also common, and closely followed by those assessing factors related to the individual. The remaining studies evaluated the impact of tumor-related factors, biological parameters, medication and supportive care use, comorbidities, hygiene, and the microbiota on mucositis outcomes.

Mucositis assessment

Almost all studies evaluated OM or GI-M in isolation, with only 24 of the 113 studies included evaluating both. There was a significant diversity in type and quality of scales used to define mucositis presence and severity, with both clinical and patient-reported outcomes used. OM was assessed using the National Cancer Institute’s Common Terminology Criteria for Adverse Events (NCI CTCAE; N = 34, including studies that assessed OM and GI-M concurrently), Radiation Oncology Research Group (RTOG; N = 23), the World Health Organization OM assessment scale (WHO; N = 13), the Head and Neck Radiotherapy Questionnaire (HNRQ; N = 1), the Oral Mucositis Assessment Scale (OMAS; N = 1), the Oral Mucositis Index (OMI; N = 4), the Oral Assessment Guide (OAG; N = 1), and the Bearman’s Oral Mucositis Scale (N = 1). GI-M was assessed primarily using the NCI CTCAE (N = 40, including studies that assessed OM and GI-M concurrently); however, other methods were also used including RTOG (N = 4), WHO (N = 3), the Inflammatory Bowel Disease Questionnaire (IBDQ; N = 1), endoscopic evaluation (N = 1), the bowel problem scale (N = 1) [21], and intestinal wall thickness (N = 1) [22]. In N = 8 cases, mucositis severity was reported but the assessment scale used was not specified.

Treatment specifications define mucositis risk

The impact of treatment specifications, particularly dosimetric parameters, on OM and GI-M risk was evaluated in a variety of cohorts undergoing chemotherapy, radiotherapy, and chemoradiotherapy for a variety of solid and hematological tumors (summarized in Table 1). In all cases, “risk” is defined as an increased risk of mucositis unless specified. Clear dose-response relationships were identified for both OM and GI-M, with increasing doses of melphalan [23, 24], methotrexate (MTX) [25], and radiation [26,27,28,29,30] predictive of OM, GI-M, and late rectal morbidity (radiation only). Unsurprisingly, the volume of bowel exposed to radiation was a strong predictor of acute and late GI toxicity reported across several studies [21, 31,32,33,34,35,36,37,38,39,40,41].

Table 1 Treatment-related factors associated with increased mucositis risk
Full size table

Treatment modality was also identified as an important mucositis predictor, with concurrent chemotherapy (in particular, cisplatin, and cetuximab) [42, 43] increasing the risk of OM and conventional 2D/3D radiotherapy associated with more severe mucositis compared with intensity-modulated radiotherapy (IMRT) [41, 44, 45]. One study also reported higher risk of OM in only daily fractionated radiotherapy compared with altered fractionated schedules [46]. Fully ablative/myeloablative conditioning (FAC/MAC) regimens were consistently reported as predictors of mucositis in people with hematological malignancies, particularly when containing etoposide [47], high-dose melphalan [23, 48, 49], busulfan [50], doxorubicin [22], or total body irradiation [51]. Of interest, three studies also reported the potential effect of circadian rhythm on mucositis risk, with radiotherapy delivered in the evening associated with a higher degree of OM [52] and associated weight loss [53] in patients with HNC. This is in contrast to GI-M in women with cervical cancer, with overall diarrhea and grade III–IV diarrhea both significantly increased when radiation was delivered in the morning compared with the evening [54].

Genetic predictors of mucositis risk

Pharmacogenetic variants

Polymorphism affecting a variety of gene clusters was evaluated for their ability to predict mucositis risk (summarized in Table 2). Mutations in drug-metabolizing pathways have been widely described for their influence on mucositis risk including methylenetetrahydrofolate reductase (MTHFR), cytochrome P450 (CYP), dihydropyrimidine dehydrogenase (DYPD), thymidylate synthase (TSMS), ATP-binding cassette (ABC) transporters, and glucoronosyltransferase 1A (UGTA1) (for detailed description of these genes and their pathways, please see [55]). Umbrella reviews support strong predictive power for UGT1A1*6 and *28 for irinotecan-induced GI-M, MTHFR C677 for MTX-induced OM and GI-M, and DPYD- and TYMS-related SNPs for mucositis caused by fluoropyridimine and platinum-based therapies [56,57,58]. The impact of MTHFR mutations on MTX-induced OM was also confirmed in two large cohorts of allo-SCT recipients with the C677 TT genotype associated with OM risk [51, 59], as well as a cohort of pediatric patients undergoing chemotherapy for ALL [60].

Table 2 Genetic factors associated with increased mucositis risk
Full size table

Cell signaling variants

Genes related to cellular growth and DNA repair mechanisms were also evaluated for their impact on mucositis risk (Table 2). The NBN rs1805794 CC genotype, high RPM1 gene expression, MDM2 (309T>G), and RB1 rs2227311 were all associated with a high OM risk in HNC patients [55, 61,62,63]. TGFBR25P and RAD51 G315C were associated with increased risk of GI-M in patients undergoing chemoradiotherapy for rectal and anal malignancies [64, 65]. The largest study in this cohort also reported strong GI-M prediction based on the VEGFR2 H472Q Q/Q rs1870377 genotype [66]; however, this was not consistent across both cohorts included in the study. The ERCC1 rs3212986 AA genotype was associated with a greater risk of late-onset GI-M in women with cervical cancer undergoing chemoradiotherapy [67]. Systematic review showed XRCC1 rs25487 was strongly predictive of OM and GI-M risk in HSCT recipients [68]; however, this was not upheld in a smaller cohort of HNC patients [69]. XRCC1 polymorphic variants were also associated with OM risk in HNC patients treated with chemoradiation [70].

Immunogenetic variants

The third category of genetic factors evaluated for their impact on OM and GI-M were those related to inflammatory/immune pathways (Table 2). Immunogenomic studies report variants in pattern recognition receptor pathways, with TLR2/TNFA associated with heightened GI-M risk in people with gastric/colorectal cancers undergoing 5-FU-based chemotherapy [71]. TNF-related variants were also identified as predictors of GI-M, with the 1031 TT genotype (and IL-1B-511 TC/TT) and baseline TNF mRNA expression associated with increased GI-M in patients undergoing chemoradiotherapy for GI [72] and esophageal cancers [73], respectively. Similarly, TNFRSR1A-610T>G, TNFA-1211T>C (CC genotype) and GHLR-2531C>T were strongly associated with OM risk in HNC patients undergoing radiotherapy [74,75,76]. Radiation-induced esophagitis was also linked with mutations in a variety of inflammatory pathways, including PSTG, TNF, IL6, IL4, and IL10 [43].

Uncategorized genetic predictors of OM risk also identified included the GT genotype in EDN1 rs1800541 [77], ZNF24 rs11081899-A [78], APEH c.1521G>C (rs4855883 CC genotype) [75], and miR-1206 rs2114358 (homozygous GG) [79].

Demographic, lifestyle and performance indicators linked with mucositis risk

Evidence from included studies for demographic and lifestyle predictors of mucositis was variable (summarized in Table 3). A strong and consistent outcome was female sex, which was linked to more severe chemotherapy-induced OM and GI-M across several studies [23, 41, 80,81,82,83]. Nonetheless, negative findings were also reported, with particularly robust evidence supporting no correlation between sex and OM in a study of 381 multiple myeloma patients [48]. Similarly, contradictory evidence supported male sex as a predictor of OM in HNC patients undergoing chemoradiotherapy [42].

Table 3 Patient-related factors associated with increased mucositis risk
Full size table

Age was commonly reported as a predictor of mucositis; however, there was no uniformity with both decreasing and increasing age reported as predictive of mucositis severity [25, 41, 84,85,86,87]. Similarly, the evidence was unclear regarding the impact of BMI on mucositis risk, with evidence indicating that low BMI [23, 88] and a BMI > 25 [51] were predictive of OM risk. In contrast, while the evidence may be limited, there are consistent reports detailing baseline performance status, with low KPS [34] and ECOG performance status [24] predictive of OM and GI-M risk. Smoking was also identified as a predictor of OM in HNC patients and transplant recipients [69, 88, 89]. In contrast, one study also reported that non-smokers had a 2.7-fold increase in the risk of grade II–IV OM [90].

Several biological/laboratory parameters (assessed prior to therapy) were linked with mucositis risk (summarized in Table 3), including low colonic thymidylate synthase (TS) expression [91], ALT (>16Ul) [81], low urea [81], neutropenia [22], lymphopenia [42], low platelet counts [42], low hemoglobin levels [42], renal dysfunction [12], and serum creatinine [48]; however, the evidence was sparse and limited to a small number of stand-alone studies.

Comorbidities, medication use and previous therapy

There are limited data on the impact of comorbidities on mucositis outcomes (summarized in Table 3), with only one study reporting inflammatory bowel disease (IBD) as a predictor of late-onset bowel injury caused by radiotherapy [28]. Similarly, the number of bowel movements at baseline was reported to predict GI-M risk [92]. Herpes simplex virus (HPV) was also reported to increase the risk of OM, with HPV+ patients reported to have a 6.8-fold increased risk of grade III–IV OM following chemoradiotherapy [90].

Few studies investigated the impact of medication use. One study reported multivitamin use prior to therapy as a predictor of lower OM scores [51]. Another study reported that antibiotic therapy for > 10 days in the month preceding treatment was a predictor of GI-M (neutropenic colitis) in children with ALL [22]. This was also supported in a cohort of adult patients undergoing conditioning chemotherapy for HSCT, with antibiotic use and GvHD prophylaxis with MTX predictive of OM [88].

Exposure to previous radiotherapy was reported as a predictor of OM in auto-SCT recipients [47], supporting reports of residual DNA damage as an independent predictor of OM risk [93]. Similarly, OM in previous chemotherapy treatment was reported to predict OM in HN patients undergoing adjunctive radiotherapy [85].

Tumor-related variables linked with mucositis risk 

Both tumor location [42, 84], size [92], and stage [69] were reported as predictors of OM and GI-M, respectively, in stand-alone studies (summarized in Table 4). A single study also reported a diagnosis of NHL (versus other hematological malignancies) as a predictor of OM [47]. In children, germinal tumors (versus non-germinal) and hematological malignancies versus CNS tumors (specifically Hodgkin’s lymphoma) were reported to increase the risk of HSCT-induced OM [50, 94].

Table 4 Tumor characteristics and diagnostic features associated with increased mucositis risk
Full size table

Areas requiring further investigation

While there were a large number of positive associations identified in our systematic review, we also identified negative studies or those that required further validation. For example, two studies reported no significant association between GI-M and MTHFR 677T mutations in patients undergoing chemoradiotherapy for rectal/anal cancer [95, 96]. Similarly, two large studies (N = 322, N = 113) failed to identify any significant association between MTX transporter mutations and TYMS SNPs for OM risk in children with ALL [19, 97] despite strong evidence supporting its role in adult cohorts [57]. While a systematic review of over 6500 patients supported a strong association between XRCC1 (rs25487) and OM and GI-M risk in HSCT recipients [68], another reported no significant association in XRCC1 mutations and OM risk in HNC patients [69]. Similarly, Lunberg et al. [98] and Goutham et al. [99] reported no significant effect of TGFB on OM despite a positive association identified for GI-M. Several studies, including a large prospective trial in 381 participants [100], also failed to identify any demographic influence on mucositis risk, including age, sex, BMI, and race [24, 48, 101,100,103].

Discussion

This paper is the first to collect and synthesize factors identified to contribute to mucositis risk and symptom severity. The purpose of which is to provide a foundation for exploration of promising research pathways to optimize risk stratification tools. The panel reviewed 197 publications, with 113 eligible for assessment. The study inclusion period of 2000–2019 was chosen to reflect the rapidly changing face of cancer treatment and toxicity management with the advent of new technologies and supportive agents. Despite this, mucositis continues to be a major challenge in the delivery of effective care. It is critical that research efforts and resources are dedicated to identifying patients at risk of severe mucositis to mitigate the constellation of clinical, psychosocial, and economic burdens with which it is associated, and potentially prevent long-term sequelae.

The most common factors investigated for their influence on mucositis were treatment-related factors, with dosimetric parameters evaluated across several oncology cohorts. Unsurprisingly, the strongest level of evidence was achieved for the volume of target organ (gastrointestinal tract and oral cavity, to lesser extent) exposed to irradiation [31, 32, 35, 36, 38,39,40, 100]. Similarly, increasing doses of melphalan [23, 24] and methotrexate [25], myeloablative conditioning therapy [88, 104, 105], and conventional RT (versus IMRT) [45] were all robustly demonstrated to increase an individual’s risk of mucositis. However, these parameters provide little benefit in terms of risk as they are largely unmodifiable, and the concept that increasing exposure of the alimentary mucosa to cytotoxic therapy increases mucositis risk reflects what is already well understood from a pathobiological perspective. It is also important to note that while female sex was consistently linked with increased mucositis risk, this may not be a direct effect and is likely mediated through altered dosimetric parameters. Of the studies included that identified female sex as a risk factor, dosimetric parameters were not considered, and in one case, females were reported to receive higher doses due to the way in which chemotherapeutic drugs are administered unadjusted for sex-dependent anthropometry [24].

Multiparameter risk prediction modeling was uncommon in the studies included for review, with most hypothesis-driven focusing on a single parameter in a strictly defined cohort, and a compartmentalized/siloed viewpoint when it comes to treatment outcomes. While this provides focus to many studies, single-candidate gene studies fail to address the complexities of mucosal injury and are limited by the overriding risk of false negatives, given the large number of participants required to obtain sufficient power [106]. Similarly, they are impacted with greater ease by confounding factors such as cohort heterogeneity, and gene-drug dose interactions need to be considered. As such, it is unsurprising to see such inconsistency in the literature. Pathway analyses and genome-wide association studies (GWAS) have the potential to offer greater insight both pathobiologically and predictively [4]; however, they are rare, presumably due to the high logistical demand of patient recruitment, biospecimen collection, and mucositis assessment, as well as cost. It was for this reason that we decided to include both association and risk prediction studies to provide a comprehensive overview of the current landscape regarding mucositis prediction/risk factors. So, while association studies provide a solid foundational knowledge of the factor that are likely to govern risk, they fail to provide tangible benefits in the provision of personalized supportive care. Future studies must therefore aim to prospectively develop multiplexed predictive strategies that are developed in training cohorts and validated in an independent manner. There has already been promising early work identifying clustered predictors and algorithms that predict high-risk patients and their response to intervention [107,106,109]. Moving forward, coordinated efforts, appropriate infrastructure, and consistent/comprehensive data collection across multiple locations with access to biobanks would drastically improve our ability to identify meaningful risk predictors, which will lay the foundation for personalized supportive care in which toxicity, efficacy, and cost can be prioritized based on the individual’s needs (Fig. 2).

Fig. 2
figure 2

Conceptual framework for implementing personalized cancer therapy that optimizes efficacy, toxicity, and economic outcomes based on comprehensive risk evaluation and individual needs of the patient. Multiplexed risk evaluation tools should provide risk stratification for all toxicities/side effects, providing the option to preferentially avoid toxicities considered most impactful for the individual while optimizing efficacy and reducing cost. Proactive and tailored supportive care can be directed to high-risk patients, while low-risk patients can avoid unnecessary interventions and be treated in an out-patient setting with confidence. Clear risk evaluation also provides patients with greater clarity regarding treatment outcomes which can be used to guide decision-making particularly in end-of-life care when quality of life may be prioritized over quantity

Full size image

Critical to such efforts is the recognition of non-genomic risk predictors, which are likely to offer great predictive power when combined with treatment and genomic factors. This has been adopted in gerioncology, with a risk stratification tool developed using geriatric assessment variables; laboratory tests; and patient, tumor and treatment variables [110, 111]. While authors used a multiplexed approach, the variables included were primarily related to the performance status of the individual and treatment specifications, and thus could be improved by inclusion of modifiable biological and genetic parameters. An emerging area of personalized risk prediction, with relevance to geriatric oncology and the broader field, is the microbiota. While this has primarily focused on immunotherapy-induced colitis, evidence suggests pre-therapy microbial composition dictates mucositis caused by both chemotherapy and radiotherapy, with early microbial changes predictive of ulcerative OM development [112, 113]. While an area of interest is increasing due to the modifiable nature of the microbiota, this is a relatively new area of risk prediction, and thus, few studies were included in this systematic review. Given the profound influence the microbiota has on both local mucosal and immune function, the influence of the gastrointestinal and oral microbiota is likely to be significant. Future biobanking efforts to evaluate pre-therapy microbial predictors and clinically feasible methods of characterizing the microbiota (e.g., FAIMS) are therefore warranted.

With the development of multiparameter risk prediction models and appropriate recognition of systems medicine approaches to analysis, risk prediction serves to not only identify high-risk patients but also guide the development of next-generation interventions. These are likely to be unique to specific oncology cohorts and will almost certainly differ between adults and children. Disappointingly, in the 113 papers included in our analysis, only 10 addressed the topic of risk prediction in children. Developing risk prediction tools for children undergoing cancer therapy is of particular relevance as mucositis (and its associated manifestations) not only poses an acute impact on their quality of life but is also increasingly recognized for its influence on late effects in survivors of childhood cancer [114, 115].

Risk prediction in children is also critical as many factors recognized for their influence on mucositis development in adults are likely to be poorly extrapolated to children (e.g., comorbidities, smoking status, medication use). Similarly, microbial communities are well recognized to differ between adults and children [116], hindering efforts to predict pediatric mucositis based on microbial enterotypes identified in adults. A growing body of evidence also suggests that mucosal barrier injury and microbial injury are catalysts for other late-onset conditions (seen in both adults and children), including cardiac toxicity [117], graft versus host disease [13], fatigue [118], and neurocognitive impairment [119]. As such, a new frontier in risk prediction will be the interaction between acute and chronic toxicities, and the ability to mitigate chronic morbidity via the personalized prevention of acute injury.

In addition to its influence on late-onset toxicities, mucositis is well established to cluster with other significant toxicities, including fatigue, pain, and blood stream infection, suggesting common molecular pathways [118, 120,119,122]. It is therefore intuitive to suggest that symptom clusters also have common, or at least complementary, risk predictors. This has already been highlighted with the microbiota now recognized to predict multiple toxicities. Similarly, overlap exists in predictors of mucosal and hematological toxicities, reiterating the importance of targeting ubiquitous mechanisms of injury and inflammation with relevance to the broader field of regimen-related toxicities. These approaches will likely be based on multiplexed approaches, with weight given to certain parameters with particular relevance to organ-specific toxicities.

Despite a large number of positive studies included in this review, we also identified several negative studies that require further investigation or validation. While the number of negative studies is likely to be higher than those identified, biased by the pressure to produce positive findings, these studies also emphasize the challenges faced in conducting rigorous risk prediction studies. We identified a number of studies that had very low patient numbers, used non-validated assessment scales, and poorly defined patient populations that were undermined by high heterogeneity. Future studies must be appropriately powered and be guided by validated principles for mucositis assessment. The increased utilization and appropriate use of biobanking and shared data may help overcome the difficulties of recruitment to predictive supportive care studies and increase power. The involvement of bioinformatic and statistical experts to develop uniform methods of risk prediction analysis in the field of supportive care is also critical to success.

While this paper represents an important step in laying a foundation of knowledge regarding mucositis risk, it is not without its flaws. Firstly, we cast a large net to capture the breadth and variability of studies addressing mucositis risk prediction. This resulted in a large number of studies or varying relevance that were identified in our initial search strategy. With no uniform terminology relating to mucositis, particularly GI-M (i.e. diarrhea, gut toxicity, colitis), there is also the possibility that papers were not identified in our search strategy when they were not indexed using our defined terminology criteria. This highlights the need to standardize reporting of mucositis in clinical trials. Similarly, variable assessment scales and outcome measures prevented meta-analyses being performed and severely compromised our ability to draw parallels between studies.

In conclusion, mucositis risk prediction is becoming increasingly recognized for its role in the provision of individualized cancer care. Despite identification of distinct factors associated with mucositis risk, we remain limited in our ability to identify high-risk patients with current evidence undermined by inconsistencies, siloed approaches, and unpowered studies. For significant advances in our ability to provide personalized supportive care, large and coordinated research efforts with innovative data sharing and input by statistical support personnel are required. These approaches should focus on multiplexed approaches that address related toxicities to provide a comprehensive risk evaluation of an individual with specific weighting reflecting their unique oncological scenario and personal requirements.

Take home messages and call to action

  1. 1.

    Personalizing supportive cancer care is critical and requires a solid foundational understanding of the factors that govern individual risk.

  2. 2.

    Multiparameter risk prediction efforts with appropriate statistical approaches must be prioritized over restrictive, hypothesis-driven studies.

  3. 3.

    Risk prediction strategies identified in adults are not easily translated to children, and as such, specific risk stratification methods must be identified for pediatric cohorts.

  4. 4.

    Oral and gastrointestinal mucositis must be approached more holistically with risk prediction methods developed in parallel given the commonalities in pathobiology.

  5. 5.

    Validated assessment scales and outcome measures should be prioritized and used consistently.

  6. 6.

    Coordinated, international biobanking of patient biospecimens coupled with comprehensive data collection related to patient demographics, treatment specifications, and outcomes are desperately required to enable more robust risk prediction efforts with greater power and clinical influence.

  7. 7.

    Future efforts in risk prediction should focus on multiplexed approaches that transcend a single toxicity and aim to predict clusters of related symptoms