TOXview: a novel graphical presentation of cancer treatment toxicity profiles

Abstract

Background: Toxicity profiles play a crucial role in the choice between specific palliative chemotherapy regimens. To optimize the quality of life for cancer patients, patients should be adequately informed about potential toxicities before undergoing chemotherapy. Therefore, we constructed TOXviews, a novel graphical presentation and overview of toxicity profiles to improve information provision about adverse events. As an example, we analyzed first-line chemotherapy regimens for advanced esophagogastric cancer (AEGC).

Methods: We searched PubMed, EMBASE, CENTRAL, ASCO and ESMO for prospective phase II or III randomized controlled trials (RCTs) on palliative first-line systemic treatment for AEGC until February 2017. We extracted proportions of Common Terminology Criteria for Adverse Events grade 1–2 (mild) and 3–4 (severe) adverse events from each chemotherapy arm and pooled these by using single-arm meta-analysis. Toxicity profiles per chemotherapy regimen were visualized in bidirectional bar charts with pooled proportions plus 95% confidence intervals. For comparative analysis, chemotherapy regimens were grouped in singlets, doublets and triplets.

Results: We included 92 RCTs with a total of 16,963 patients. TOXviews for 3 fluoropyrimidine singlets, 5 cisplatin-containing doublets (C-doublets), 10 fluoropyrimidine non-cisplatin containing doublets (F-doublets), 4 anthracycline-containing triplets (A-triplets) and 5 taxane-containing triplets (T-triplets) were constructed. C-doublets, A-triplets and T-triplets all showed an increased incidence of grade 3–4 adverse events and clinically relevant grade 1–2 adverse events compared to F-doublets.

Conclusion: TOXview provides a new graphical presentation and overview of chemotherapy toxicities. TOXviews can be used to educate physicians about the incidences of AEs of systemic therapy and improve informed decision-making.

Introduction

In order to maintain the quality of life for patients with cancer, tolerability of cytotoxic treatment is crucial. This is especially true in the case of advanced disease when curation is not an option and treatment is aimed at improving survival while maintaining the quality of life. Especially when survival outcomes of certain treatment regimens are similar, differences in toxicity profiles between treatment regimens may guide treatment decisions.

Accurate communication with patients about adverse events that might occur in different treatment regimens is especially important to promote shared decision-making. In turn, a well-informed decision can result in a better quality of life [1,2]. Thus, comprehensive overviews of toxicity are urgently needed [3,4]. Several studies have attempted to construct a graphical presentation format (e.g., bar charts and stream plots) to aid information provision regarding toxicity profiles [5]. However, the currently available graphical methods to present toxicities are hard to interpret by physicians because of the lack of relevant data, such as number of participants, trials and confidence intervals in these presentations [6]. Furthermore, most illustrations of toxicity that are available represent the adverse events of a single study. To our knowledge, no pooled overview of adverse events profiles including multiple studies for chemotherapy is available yet.

In our research, we suggest a novel graphical approach of presenting toxicity of systemic treatment regimens: TOXview. As a clinical example, we studied first-line chemotherapy regimens for advanced esophagogastric adenocarcinoma (AEGC), one of the major causes of cancer-related mortality worldwide [7–9]. Currently, there is no world-wide consensus about the optimal first-line palliative chemotherapy for AECG and several regimens with similar survival outcomes are available thus making expected adverse events highly relevant for treatment decision making [10].

Methods

Literature search

We updated our previously performed search in the databases Pubmed, EMBASE and the Cochrane Central Register of Controlled Trials (CENTRAL) and the meeting abstracts from the American Society of Clinical Oncology (ASCO) and European Society for Medical Oncology (ESMO) for randomized controlled trials (RCTs) up to February 2017 [10]. Medical subject headings (MeSH) and text words for esophagogastric cancer and for each treatment option were used in the search as described previously [10] (Supplementary Table 1). Two independent reviewers (EtV and JJvK) were involved in screening the titles, abstracts and full-texts. Disagreements were discussed with a third arbiter (HvL) until consensus was reached.

Study selection and quality assessment

We selected RCTs that included at least one treatment arm with first-line palliative chemotherapy containing currently used compounds for patients with locally irresectable or metastatic esophagogastric cancer: fluoropyrimidines (i.e., 5-fluorouracil [5-FU], capecitabine or S-1), platinum (i.e., cisplatin or oxaliplatin), anthracycline (i.e., epirubicin), taxanes (i.e., docetaxel or paclitaxel) and irinotecan. In addition, any data on toxicity in the RCTs was scored according to the Common Terminology Criteria for Adverse Events (CTCAE) and was reported using the maximum grade method (i.e., only the highest occurring grade of an adverse event is registered per patient) [11].

Critical study appraisal was performed with the Cochrane Risk of bias tool (version 5.1.0) [12] and was done by two authors independently (EtV and LN). Items were scored as low, high or unknown risk of bias.

Data extraction and statistical analysis

From each separate chemotherapy arm in RCTs, AE incidence counts and the total number of patients in the toxicity population were extracted from the toxicity tables or from the text in the study reports. We divided the AEs into mild (grade 1 or 2), severe (grade 3 or 4) and fatal (grade 5). As the majority of the RCTs solely reported the AEs that occurred in ≥5% of the patients, we extracted a pre-specified set of common chemotherapy AEs, in order to standardize the extraction of AEs for all chemotherapy regimens and to keep the TOXviews comparable as CTCAE is composed of numerous categories and not every adverse event is present in each regimen due the maximum grade method of reporting.

Statistical analyses were performed with the metafor package [13] (version 2.0-0) in R (version 3.5.1) [14]. From the extracted AE incidence counts and toxicity populations, pooled proportions and 95% confidence intervals (95%CI) were calculated with a single-arm proportional random-effects meta-analysis following a logit transformation. To graphically present the toxicity profiles of each chemotherapy regimen in a TOXview, the pooled proportions of AE-incidences were summarized in bar charts, along with the 95%CI. Tables containing the number of events, total number of participants, number of trials and pooled estimates were constructed and shown alongside the bar chart. In accordance with a network meta-analysis reported previously, TOXviews for the following regimens were constructed: fluoropyrimidine-alone, cisplatin-containing doublets (C-doublets), non-cisplatin fluoropyrimidine-containing doublets (F-doublets), anthracyclin-containing triplets (A-triplets) and taxane-containing triplets (T-triplets) [10].

To validate the data presented in the TOXviews, we performed a comparative analysis of incidences of AEs between regimens based on our previous work [10] using the Wald test as overall test and additional two-sided post-hoc tests, with Holms correction for multiple comparisons. All tests had a significance level of p = .05. To display the comparisons in a graphical presentation, we constructed a bidirectional chart of TOXviews.

Results

Description of the included studies

Until February 2017, a total of 5765 unique titles were retrieved through the database search, of which 191 RCTs remained after screening the titles and abstracts. After full-text assessment, 108 were excluded and 83 were found eligible. In addition, 9 studies retrieved from the conference search were found eligible (Supplementary Figure 1). A total of 92 studies containing 16,963 patients were included [15–106]. No major differences in study design were observed (Supplementary Table 2). Forty-four (48%) studies were rated as having a low risk of bias (Supplementary Figure 2). Fourteen (15%) and 18 (20%) studies were rated as unclear risk of bias on only one item or two items, respectively. The other 16 studies (17%) were rated as unclear on three items or were reported as abstract only.

Presentation of TOXview of singlet, doublet, triplet cytotoxic regimens

It was possible to construct individual toxicity profiles for twenty-eight cytotoxic regimens and these were graphically represented in TOXviews (Supplementary material). As stated before, we constructed TOXviews for regimens including cytotoxic compounds that are used nowadays in the treating AEGC and which are potentially clinically relevant chemotherapy regimens according to previous work [10]. Two types of TOXviews were presented, using the toxicity profile of the doublet oxaliplatin-capecitabine as an example: one that consists of a table and a bar chart with confidence intervals and one that is displayed as a bidirectional chart with confidence intervals (Figure 1(A,B)). TOXviews for chemotherapy regimens were divided into different groups: cytotoxic single agent regimes, cytotoxic doublet regimens, cytotoxic triplet regimens (Figure 2(A–C)) which were used for further comparative analysis.

Figure 1. Example for TOXview. Toxicity profile of capecitabine (CAP)+oxaliplatin (Ox) (CAPOX) presented in two different ways. (A) Barchart and table CAPOX. (B) Bidirectional chart and table CAPOX. Grade 1–2 left directed, grade 3–4 right directed, a grade 5 adverse event (fatal) was represented as toxicity related death.

Figure 2. Overview of chemotherapy regimens used for TOXview. (A) Cytotoxic singlet regimens (B) Cytotoxic doublet regimens (C) Cytotoxic triplet regimens RCT: randomized controlled trials; 5-FU: 5-fluorouracil; CAP: capecitabine; C: cisplatin; D or DTX: docetaxel; I: irinotecan; PTX: paclitaxel; Ox: oxaliplatin; T: taxane; F: fluoropyrimidine; A: anthracycline; FAMTX: 5-FU + doxorubicin + methotrexate; FEMTX: 5-FU + epirubicin + methotrexate; ECX: epirubicin + cisplatin + capecitabine; ECF: epirubicin + cisplatin + 5-FU; FOXP: 5-FU + oxaliplatin + paclitaxel; FOXD: 5-FU + oxaliplatin + docetaxel; EOF: epirubicin + oxaliplatin + 5-FU; EOX: epirubicin + oxaliplatin + capecitabine.

Comparative analysis of TOXviews

C-doublets were compared to F-doublets as shown in the bidirectional plots and table (Figure 3; Table 1). To show the differences among the F-doublet regimens, C-doublets were also compared to separate F-doublets (Supplementary Figure 3). Next, F-doublets were compared against T-triplets and A-triplets (Figure 4; Table 2). In general, F-doublets showed more grade 1–2 but less grade 3–4 differences in incidences of adverse events compared to C-doublets. Compared to F-doublets, A-triplets and T-triplets showed more grade 3–4 adverse events and toxicity-related-deaths. When A and T-triplets were compared, T-triplets showed more adverse events than A-triplets.

Figure 3. Bidirectional comparison plot for C-doublets versus F-doublets. C: Cisplatin. F: Fluoropyrimidine.

Figure 4. Bidirectional comparison plot for A-triplets versus T-triplets versus F-doublets. A: anthracycline; T: Taxane; F: Fluoropyrimidine.

Table 1. Statistically significant differences between incidences of chemotherapy regimens (p < .05) C-doublets versus F-doublets.

	C-doublets vs F-doublet		C-doublets vs FI		C-doublets vs FOx		C-doublets vs FT
Grade	F > C	C > F	FI > C	C > FI	FOX > C	C > FOX	FT > C	C > FT
1-2	Neutropenia	Trombocytopenia	Neutropenia	Trombocytopenia	Neutropenia	Nausea	Leukopenia	Fatigue
	Leukopenia	Vomiting	Leukopenia	Neuropathy	Leukopenia	Vomiting	Hand foot syndrome	Anemia
	Anorexia		Anemia		Trombocytopenia		Stomatitis	Trombocytopenia
	Stomatitis		Diarrhea		Anorexia			Diarrhea
	Alopecia		Nausea		Neuropathy			Vomiting
	Mucositis		Fatigue		Fatigue			Neuropathy
	Fatigue		Alopecia		Stomatitis
3-4	Diarrhea	Neutropenia	Diarrhea	Trombocytopenia	Trombocytopenia	Neutropenia	Anorexia	Neutropenia
	Anorexia	Leukopenia		Neutropenia	Neuropathy	Leukopenia	Neuropathy	Trombocytopenia
	Neuropathy	Febrile neutropenia		Leukopenia		Febrile neutropenia	Stomatitis	Nausea
	Stomatitis	Fatigue		Nausea/vomiting		Nausea	Mucositis	Vomiting
	Mucositis	Nausea		Fatigue		Vomiting		Fatigue
		Vomiting				Fatigue

F: Fluoropyrimidine; C: Cisplatin; FI: Fluoropyrimidine and irinotecan; FOX: Fluoropyrimidine and oxaliplatin; FT: fluoropyrimidine and taxane; >: Higher incidence than.

Table 2. Statistically significant differences between incidences of chemotherapy regimens (p < .05). F-doublets versus A-triplets versus T-triplets.

	A-triplets vs F-doublets		T-triplets vs F-doublets		T-triplets vs A-Triplets
Grade	A > F	F > A	T > F	F > T	T > A	A > T
1–2	Anemia	Trombocytopenia	Anemia	Neutropenia	Anemia	Neutropenia
	Diarrhea	Nausea	Diarrhea	Leukopenia	Trombocytopenia	Leukopenia
	Nausea/vomiting	Anorexia	Nausea/vomiting	Anorexia	Diarrhea	Nausea/vomiting
	Neuropathy		Neuropathy	Thrombocytopenia	Nausea	Hand-foot syndrome
	Stomatitis		Fatigue		Anorexia
	Alopecia		Alopecia		Fatigue
	Hand-foot syndrome		Mucositis		Alopecia
3–4	Neutropenia	Anemia	Neutropenia	Mucositis	Neutropenia
	Febrile neutropenia	Trombocytopenia	Leukopenia		Leukopenia
	Nausea	Diarrhea	Febrile neutropenia		Febrile neutropenia
	Vomiting	Mucositis	Vomiting		Nausea/vomiting
	Fatigue		Neuropathy		Neuropathy
	Toxicity related death		Fatigue
			Toxicity related death

A: Anthracycline; F: Fluoropyrimidine; T: Taxane; >: Higher incidence than.

Heterogeneity

High heterogeneity numbers were detected for a majority of single-arm meta-analyses. In the individual TOXviews in which the portion of heterogeneity was statistically significant for any given AE according to the Q-test, the I² of these individual TOXviews ranged from 47% to 99%. In the comparative analysis, the statistically significant heterogeneity in I² ranged from 54% to 97%.

Discussion

This study introduces a new graphical method to present toxicity of systemic therapies by using single arm pooled proportions meta-analysis. With TOXview, we constructed a method where we can illustrate the effect of an intervention and its confidence interval. With single arm pooled proportions meta-analysis, we combined adverse events while taking into account each confidence interval of the adverse event in a specific study. We thereby could estimate the pooled proportion with its CI with a random effects model. Moreover, by adding information about the events, the total number of participants and the number of trials, it is possible to interpret the confidence interval of the pooled proportion.

Up to now, there was no uniformly utilized graphical presentation format of AE to use during a consultation. With the lack of an accurate toxicity overview, it can be difficult for physicians to provide realistic risks of adverse events while discussing systemic treatment options. A graphical presentation, such as TOXview, could be a convenient tool to improve accurate communication by physicians about toxicity and to improve understanding of patients about the expected toxicity of specific treatment regimens [3,4]. With TOXview, it is possible for both the physicians and patient to see how the toxicities are distributed per chemotherapy regimen. As the incidence and confidence intervals are given, physicians can make better estimations of the potential tolerability for an individual patient and can provide detailed information about the nature of adverse events that patients can expect with a certain regimen. We have used palliative first-line chemotherapy for AEGC as an example. According to our previous work [10], C-doublets were less effective and had increased toxicity compared to F-doublets whereas A-triplets and taxane + cisplatin + fluoropyrimidine (TCF) were not more effective and had increased toxicity over F-doublets. Fluoropyrimidine + Oxaliplatin + Taxane (FOxT) was found more effective but had increased toxicity compared to F-doublets. Also from the analysis of the TOXviews, we can conclude that F-doublets are the preferable chemotherapy regimen based on its low incidence of high-grade toxicities. Compared to F-doublets, C-doublets, A-triplets and T-triplets had higher incidences of severe adverse events. This shows that the data used in the TOXviews collected from single arms in different RCTs is in concordance with the results of our previous work based on between-arm comparisons of toxicities [10].

Thanarajasingnam et al. [4] published a new graphical method as well, based on individual patient data and incorporated time in the toxicity presentation. We are aware that our presentation of toxicity does not provide any information to compare toxicity over time between regimens, even though this may be an important feature to share with patients when choosing a chemotherapy regimen. Unfortunately, in most clinical studies, the maximum-grade method is used to report data about toxicity and time to event is usually not incorporated [11]. To make a better estimation of the incidences of toxicities, analysis of individual patient data would preferably incorporate time to event. This also opens the opportunity to study multiple events per patient over time.

TOXview also has some limitations. Firstly, high heterogeneity numbers were observed in the TOXviews, indicating substantial variation between studies. This was partly expected because data of single arms from RCTs were used. Moreover, different dosages of chemotherapy and the intensity of cycles could explain the heterogeneity numbers in our data. Due to high heterogeneity, all analyses were performed with a random-effects model. In addition, as described before by Sivendran et al. [11], possible causes of the high heterogeneity could be unawareness of systematically reporting adverse events, inadequate compliance to protocols and selective reporting. It is difficult for patients to notice all aspects of their adverse event and report it back to their physicians as not every physician score the same for each adverse event [107]. It has been shown that documentation of adverse events in daily out-patient clinical practice is difficult for several reasons, such as unclear adverse events scoring terms, few experiences of scoring toxicity and the difficulty of the small difference between toxicity related adverse events or disease-related events [11,107,108]. Up to now, the most accurate documentation of toxicities is done in clinical trials where a standard protocol is followed to administrate adverse events systematically. This can also contribute to a different estimation of the number of adverse events. Besides the heterogeneity caused by possible factors described as before, TOXviews have the limitation that confidence intervals are dependent on the number of studies and patients. In Figure 1, we chose to show capecitabine + oxaliplatin (CAPOX), because this regimen is frequently used in clinical practice. This TOXview contains a maximum of four studies, thereby contributing to a wide CI. Other frequently used regimens that were based on many more studies and therefore, show less wide confidence intervals.

Another limitation of this study is that most studies only reported AEs in the study report when these reached a threshold of e.g., 5% of the total group [11]. This could imply that some adverse events are underreported in the TOXviews even though they were present. To counter this problem, we used a prespecified set of common AEs of chemotherapy; as such the most important AEs could be compared between the regimens. In addition, very few studies explicitly mention that the adverse events are possibly related to treatment, which could also result in underreported adverse events [41,64]. For future research, using individual patient data or creating an international adverse event report system e.g. PRO-CTCAE™ [109] could create better estimation of the incidences in the TOXview and adverse events in general.

Conclusion

TOXview is the first attempt to present a pooled overview of systemic treatment toxicities from multiple studies. We advocate TOXview as a new tool in the clinic to promote shared decision making of physicians with their patients. Particularly regarding AEGC, for which no formally golden standard first-line palliative chemotherapy exists, TOXview could provide better guidance in communicating expected toxicities and may promote shared decision-making between patient and physician.

Disclosure statement

H.W.M.v.L. served as a consultant for Celgene, Lilly, and Nordic. H.W.M.v.L: received funding outside the submitted work from Bayer, BMS, Celgene, Lilly, Merck Serono, MSD, Nordic, Philips and Roche. M.G.H.v.O received funding outside the submitted work from Bayer, Lilly, Merck Serono and Roche; all other authors declare that they have no potential conflicts of interest.