Non-parametric methods for analysing recurrent complications of varying severity

Statistical methods for the analysis of adverse event data in randomised controlled trials: a scoping review and taxonomy

BACKGROUND

Statistical methods for the analysis of harm outcomes in randomised controlled trials (RCTs) are rarely used, and there is a reliance on simple approaches to display information such as in frequency tables. We aimed to identify whether any statistical methods had been specifically developed to analyse prespecified secondary harm outcomes and non-specific emerging adverse events (AEs).

METHODS

A scoping review was undertaken to identify articles that proposed original methods or the original application of existing methods for the analysis of AEs that aimed to detect potential adverse drug reactions (ADRs) in phase II-IV parallel controlled group trials. Methods where harm outcomes were the (co)-primary outcome were excluded. Information was extracted on methodological characteristics such as: whether the method required the event to be prespecified or could be used to screen emerging events; and whether it was applied to individual events or the overall AE profile. Each statistical method was appraised and a taxonomy was developed for classification.

RESULTS

Forty-four eligible articles proposing 73 individual methods were included. A taxonomy was developed and articles were categorised as: visual summary methods (8 articles proposing 20 methods); hypothesis testing methods (11 articles proposing 16 methods); estimation methods (15 articles proposing 24 methods); or methods that provide decision-making probabilities (10 articles proposing 13 methods). Methods were further classified according to whether they required a prespecified event (9 articles proposing 12 methods), or could be applied to emerging events (35 articles proposing 61 methods); and if they were (group) sequential methods (10 articles proposing 12 methods) or methods to perform final/one analyses (34 articles proposing 61 methods).

CONCLUSIONS

This review highlighted that a broad range of methods exist for AE analysis. Immediate implementation of some of these could lead to improved inference for AE data in RCTs. For example, a well-designed graphic can be an effective means to communicate complex AE data and methods appropriate for counts, time-to-event data and that avoid dichotomising continuous outcomes can improve efficiencies in analysis. Previous research has shown that adoption of such methods in the scientific press is limited and that strategies to support change are needed.

TRIAL REGISTRATION

PROSPERO registration: https://www.crd.york.ac.uk/prospero/display_record.php?RecordID=97442.

Modernizing adverse events analysis in oncology clinical trials using alternative approaches: rationale and design of the MOTIVATE trial

In oncology clinical research, the analysis and reporting of adverse events is of major interest. A consistent depiction of the safety profile of a new treatment is as crucial in establishing how to use it as its antitumor activity. The advent of new therapeutics has led to major changes in the management of patients and targeted therapies or immune checkpoint inhibitors are administered continuously for months or even years. However, the classical methods of adverse events analysis are no longer adequate to properly assess their safety profile. Indeed, the worst grade method and time-to-event analysis cannot capture the duration or the evolution of adverse events induced by extended treatment durations. Many authors have highlighted this issue and argue that the analysis of safety data from clinical trials should be modernized by considering the dimension of time and the recurrent nature of adverse events. This paper aims to illustrate the limitations of current methods and discusses the value of alternative approaches such as the prevalence function, Q-TWiST, the ToxT and the recurrent event approaches. The rationale and design of the MOTIVATE trial, which aims to model the evolution of toxicities over time using the prevalence function in patients treated by immunotherapy, is also presented ( ClinicalTrials.gov Identifier: NCT03447483; Date of registration: 27 February 2018).

The importance of jointly analyzing treatment administration and toxicity associated with targeted therapies: a case study of regorafenib in soft tissue sarcoma patients

Background

Different methods have been proposed to analyze adverse events (AEs) associated with targeted therapies. While these AEs lead to dose adjustments for many patients, conventional reporting methods do not take drug administration into consideration. This paper underlines the importance of jointly reporting AEs and drug administration using prevalence, and proposes a complementary approach to reporting.

Patients and methods

The prevalence method estimates the probability of progression-free patients being in a particular health state (state 1: AEs with full dose; state 2: AEs with reduced dose; state 3: no AEs with reduced dose) at different time points. To take into account the impact of dose adjustments on efficacy, the weighted prevalence method can be used by assigning utility weights to the different health states. The benefit of these methods was illustrated using data from a phase II trial of regorafenib.

Results

Only 4.6% of progression-free patients developed mucositis/stomatitis (grade ≥2) at 3 months. The prevalence of patients not experiencing this AE but whose dose was reduced or treatment interrupted was 58.1%. The weighted prevalence of the regorafenib toxicity profile and dose reduction was higher in the control arm.

Conclusion

This case study confirms the importance of jointly analyzing AEs and drug administration. The weighted prevalence approach is an average score that incorporates the dimension of drug administration into AE assessment. This can be helpful for regulatory agencies as well as for clinicians to evaluate the benefit-risk ratio of therapies in their treatment choice.

Clinical trial

NCT01900743.

Independent component analysis for rectal bleeding prediction following prostate cancer radiotherapy

BACKGROUND AND PURPOSE

To evaluate the benefit of independent component analysis (ICA)-based models for predicting rectal bleeding (RB) following prostate cancer radiotherapy.

MATERIALS AND METHODS

A total of 593 irradiated prostate cancer patients were prospectively analyzed for Grade ≥2 RB. ICA was used to extract two informative subspaces (presenting RB or not) from the rectal DVHs, enabling a set of new pICA parameters to be estimated. These DVH-based parameters, along with others from the principal component analysis (PCA) and functional PCA, were compared to "standard" features (patient/treatment characteristics and DVH bins) using the Cox proportional hazards model for RB prediction. The whole cohort was divided into: (i) training (N = 339) for ICA-based subspace identification and Cox regression model identification and (ii) validation (N = 254) for RB prediction capability evaluation using the C-index and the area under the receiving operating curve (AUC), by comparing predicted and observed toxicity probabilities.

RESULTS

In the training cohort, multivariate Cox analysis retained pICA and PC as significant parameters of RB with 0.65 C-index. For the validation cohort, the C-index increased from 0.64 when pICA was not included in the Cox model to 0.78 when including pICA parameters. When pICA was not included, the AUC for 3-, 5-, and 8-year RB prediction were 0.68, 0.66, and 0.64, respectively. When included, the AUC increased to 0.83, 0.80, and 0.78, respectively.

CONCLUSION

Among the many various extracted or calculated features, ICA parameters improved RB prediction following prostate cancer radiotherapy.

How to report toxicity associated with targeted therapies?

BACKGROUND

In the era of personalized medicine, molecularly targeted therapies (MTT) have modified the outcome of some cancer types. The price of tumor control needs to be balanced with toxicity since these new therapies are administered continuously for several months or sometimes for several years. For cytotoxic drugs, the incidence of adverse event (AE) was traditionally reported as frequency and intensity. This simple measure is not sufficient to capture the recurrent nature and duration of AE. This paper presents two methods to better describe the toxicity burden across the time: prevalence and Q-TWiST.

PATIENTS AND METHODS

Limitation of worst-grade method and advantages of prevalence and Q-TWiST in the analysis of toxicity were illustrated using data from a phase II trial and a hypothetically simulated clinical trial.

RESULTS

Prevalence integrates the recurrent nature of AE. Using prevalence, it is possible to obtain a time profile of AE. Q-TWiST method evaluates the weighted time spent in each health state and also considers the recurrent nature of side-effects in order to assess the 'risk-benefit' ratio of a treatment. When interpreting Q-TWiST results, it is necessary to take into account overall survival and progression-free survival and to define a clinically relevant difference according to the setting.

CONCLUSION

The two methods presented here capture different effects. They are helpful for physicians in their treatment choice (balance benefit risk), to counsel patients and to optimize supportive care. In order to ensure consistency and provide critical information required for medical decision-making, it is important to encourage the use of alternative statistical methods in the analysis of toxicities associated with MTT.

CLINICAL TRIAL

NCT00541008.

Analysis of survival adjusted for quality of life using the Q-TWiST function: Interface in R

OBJECTIVE

The analysis of treatment effects in clinical trials usually focus on efficacy and safety in separate descriptive statistical analyses. The Q-TWiST (Quality adjusted Time Without Symptoms and Toxicity) method has been proposed by Gelber in the 90s to enable a statistical comparison between two groups with a graphical representation by incorporating benefit and risk into a single analysis. Although the method has been programmed in SAS, it is rarely used. The availability of the method in the freely software environment system like R would greatly enhanced the accessibility by researchers. The objective of this paper is to present a program for Q-TWiST analyses within R software environment.

METHODS

The qtwist function was developed in order to estimate and compare Q-TWiST for two groups. Two individual patient data files are required used for input: one for visits and one for follow-up. Q-TWiST is obtained as a sum of time spent in three health states: period in toxicity (TOX), period without relapse and toxicity (TWiST) and period in relapse (REL), weighted by associated utility scores restricted to median overall survival for example. The bootstrap method is used for testing statistical significance. Threshold analysis and gain functions allow a group comparison for different utility values.

RESULTS

Input data is checked for consistency. Descriptive statistics and mean durations for each health state are provided, allowing statistical comparisons. Graphical results are presented in a PDF file. The use of the function is illustrated with data from a simulated data set and a randomized clinical trial.

CONCLUSIONS

qtwist is an easy to use R function, allowing a quality adjusted survival analysis with the Q-TWiST method.

Assessment of health status over time by Prevalence and Weighted Prevalence functions: Interface in R

The importance of evaluating complications and toxicity during and following treatment has been stressed in many publications. In most studies, these endpoints are presented descriptively and summarized by numbers and percentages but descriptive methods are rarely sufficient to evaluate treatment-related complications. Pepe and Lancar developed Prevalence and Weighted Prevalence functions which take into account the duration and the severity of complication unlike conventional methods of survival analysis or competing risks which are limited to the time to first event. The purpose of this paper is to describe features and use of two R functions, main.preval.func and main.wpreval.func, which were designed for the analysis of survival adjusted for quality of life. These functions compute descriptive statistics, survival and competing risks analysis and especially Prevalence and Weighted Prevalence estimations with confidence intervals and associated test statistics. The use of these functions is illustrated by several examples.

Treating prostate cancer with radiotherapy

This Review focuses on the adverse effects of radical radiotherapy for localized prostate cancer. The adverse effects are described in the context of alternative treatment modalities. First, we consider the methodological issues that make comparison between the different treatment modalities problematic. Such issues include differences in baseline levels of urinary, bowel and sexual dysfunction, the importance of using patient-reported outcomes and the distinction between actuarial and prevalence rates of treatment-related toxic effects. Second, we describe the pattern of adverse effects that occur over time after radiotherapy. Here, we focus on evidence for a beneficial effect of radiotherapy on some urinary symptoms, and the controversy regarding the risk of secondary malignancy. Third, predictors of radiation toxicity are discussed. Accurate prediction of radiotherapy toxicity would be an invaluable tool for treatment individualization. It is noteworthy that the data on the adverse effects of prostate radiotherapy necessarily relate to treatment as it was delivered in the past. It is likely that recent technical advances, such as intensity modulation and image guidance, will further improve the toxicity profile of prostate radiotherapy.

Randomized trial of a hypnosis intervention for treatment of hot flashes among breast cancer survivors

PURPOSE

Hot flashes are a significant problem for many breast cancer survivors. Hot flashes can cause discomfort, disrupted sleep, anxiety, and decreased quality of life. A well-tolerated and effective mind-body treatment for hot flashes would be of great value. On the basis of previous case studies, this study was developed to evaluate the effect of a hypnosis intervention for hot flashes.

PATIENTS AND METHODS

Sixty female breast cancer survivors with hot flashes were randomly assigned to receive hypnosis intervention (five weekly sessions) or no treatment. Eligible patients had to have a history of primary breast cancer without evidence of detectable disease and 14 or more weekly hot flashes for at least 1 month. The major outcome measure was a bivariate construct that represented hot flash frequency and hot flash score, which was analyzed by a classic sums and differences comparison. Secondary outcome measures were self-reports of interference of hot flashes on daily activities.

RESULTS

Fifty-one randomly assigned women completed the study. By the end of the treatment period, hot flash scores (frequency x average severity) decreased 68% from baseline to end point in the hypnosis arm (P < .001). Significant improvements in self-reported anxiety, depression, interference of hot flashes on daily activities, and sleep were observed for patients who received the hypnosis intervention (P < .005) in comparison to the no treatment control group.

CONCLUSION

Hypnosis appears to reduce perceived hot flashes in breast cancer survivors and may have additional benefits such as reduced anxiety and depression, and improved sleep.

Delayed rectal and urinary symptomatology in patients treated for prostate cancer by radiotherapy with or without short term neo-adjuvant androgen deprivation

BACKGROUND AND PURPOSE

To identify contributing factors to delayed rectal and urinary symptoms in a randomised trial comparing different durations of maximal androgen deprivation (MAD), given prior to radiotherapy, for locally advanced prostate cancer.

PATIENTS AND METHODS

Between 1996 and 2000, 818 patients with stages T2b,c, 3 and 4 prostate cancer were entered into a trial comparing 0, 3 and 6 months of MAD prior to and during radiotherapy. Their delayed normal tissue effects were recorded by their treating doctors using standardised scales and by the patients using a self-assessment questionnaire regularly. Time to occurrence and prevalence data were analysed.

RESULTS

Rectal and urinary symptom levels were observed to vary markedly over time in at least 80% of patients, with some indicating lasting resolution of symptoms. Prevalence rates were found to be substantially lower than actuarial probability rates. Baseline symptom levels and greatest acute symptom levels were both very powerful predictors. Obstructive lower urinary tract symptoms were noted to improve during the first 4 years after radiotherapy in approximately 60% of cases in each treatment arm. However, the treatment arm itself was not shown to influence these improvements in other univariate or multivariate analyses. MAD was shown to reduce both time to occurrence and prevalence of delayed proctopathic symptoms, but this effect was confirmed statistically in the 3 month treatment arm only. Multivariate models indicated that higher levels of haemoglobin prior to any treatment may in some way protect against delayed proctopathic symptoms.

CONCLUSIONS

Prevalence data provide more clinically meaningful estimates of risk of delayed effects in normal tissues where assessment relies substantially on reported symptom levels. In these tissues consideration of the impact of baseline symptom levels and pathologies, and greatest acute symptom levels in analyses of delayed effects appears mandatory. Obstructive lower urinary symptoms improve over several years in the majority of patients treated for locally advanced prostate cancer by radiotherapy. Future research could address whether rectal toxicity is affected by initial haemoglobin levels and declines in it due to MAD.

Methodologic lessons learned from hot flash studies

PURPOSE

In the course of conducting a series of prospective clinical trials devoted to defining new treatment opportunities for hot flashes in cancer survivors, considerable experience has been acquired with related methodologic issues. This article has been written in response to many queries regarding this methodology.

PATIENTS AND METHODS

A series of seven different clinical trials that involved 968 patients was used for this work. Reliable and valid definitions of hot flash intensity were developed from patient-reported descriptions. Concomitant validity and reliability assessment of patient-completed diaries was undertaken to compare hot flash data with toxicity and quality-of-life (QOL) end points and to examine consistency across patient groups using variability analysis and correlation procedures. Parametric data from this meta-analysis was used to examine relative power considerations for the design of phase II and phase III clinical trials.

RESULTS

Daily diaries used in these studies exhibited consistency and reliability and had few missing data. Hot flash frequency and hot flash score (frequency multiplied by average severity) variables produced almost identical end point results. For phase III placebo-controlled studies, 50 patients per treatment arm seem appropriate to provide sufficient power specifications to detect a clinically meaningful change in hot flash activity. For phase II trials, 25 patients per trial seem to provide reasonable estimates of eventual hot flash efficacy to screen potential agents for more definitive testing.

CONCLUSION

Given the data gained from these experiences, we can plan and carry out more efficient trials to identify efficacious agents for the reduction of hot flash activity.

Impact on treatment outcome and late effects of customized treatment planning in cervix carcinomas: baseline results to compare new strategies

PURPOSE

The aim of the study was to determine the predictive factors of complications, to evaluate the impact of customized treatment planning on late normal tissue effects per stage, and to report disease-free survival (DFS) and local control (LC) rates.

METHODS AND MATERIALS

From 1970 to 1994, 642 patients were treated with radiotherapy alone for carcinoma of the intact uterine cervix. According to the International Federation of Gynecology and Obstetrics (FIGO) substaging, 34% were Stage I, 39% Stage II, and 27% Stage III. The analysis was divided into three periods: 1970-1978 (use of standard prescriptions),1979-1984 (implementation of individual adjustments), 1985-1994 (systematic individual adjustments). Five-year DFS, LC, and complications rates were calculated using the Kaplan-Meier method. Predictive factors of complications were determined by univariate analysis using frequency tables and nonparametric t-tests. Multivariate analysis consisted of a polychotomous stepwise regression.

RESULTS

The comparison of the three time periods showed a significant reduction of the external radiation dose (dose above 40 Gy in 47% of patients before 1979 vs. 36% after 1984), of the use of parametrial boost (55% vs. 39%), of the use of vaginal cylinder (28% vs. 11.5%), and of the HWT volume (combined intracavitary and external irradiation) (842 cc vs. 503 cc on average). The total sequelae/complications rate, all toxicity grades, all stages, all organs was 51%. Five-year actuarial rate per toxicity grade was: G1, 42%; G2, 23.5%; G3, 10%; G4, 3%. The three main predictive factors for rectal and bladder sequelae/complications (all toxicity grades) taking into account time period were: the increase of external radiation dose, the high dose rate at reference points, and the whole vagina brachytherapy. No G4 occurred in the third period. The rate of G3 complications dropped from 16% to 6% over time: from 5% during the first period to 0% during the third period in Stage I, from 8% to 6% in Stage II, and from 23% to 12% in Stage III. G3 currently describes a variety of clinical situations with a different impact on quality of life which justifies further refinements of definitions of late effects. In our experience the severity of G3 markedly decreased: less than one-third of G3 had a real impact on quality of life in the last period compared to more than two-thirds in the first period. Meanwhile, 5-year LC rates remained stable in Stages I and II, 91% and 85% respectively. Conversely they fell from 75% to 55% in Stage III, thus raising the problem of underdosage and/or more accurate staging with time.

CONCLUSIONS

Customized treatment planning eradicated lethal complications and provided a significant decrease of G3 in all stages while maintaining high cure rates in early stages. Dose reduction should be considered with caution in Stage III.

Measures of treatment effectiveness on tumour response and survival: a multi-state model approach

This paper proposes the application of a counting process approach in the analysis of treatment effect on tumour response and survival. It relies on the definition of two transient states between which individuals may move over time, that is, response and non-response, and of one absorbing state, death. Three models are discussed according to the underlying Markov and duration dependence assumptions, as well as a marginal modelling of repeated transitions between the two transient states. We illustrate the proposed methods with data from a randomized clinical trial assessing the effect of fludarabine in patients with advanced chronic lymphocytic leukaemia.

[Complications of treatments of invasive cancers of the uterine cervix with intact uterus: results and prevention]

The prospective record of acute and late toxicity after treatment of cervix carcinomas is a part of the description of treatment outcome as well as local control and survival. Due to the large number of scales and glossaries used, the comparison of the results from one study to another is often difficult. The French-Italian syllabus seems to be the most reliable scale, providing implementation of quality of life assessments. The main predictive factor of complications, which is not related to the treatment type, is the previous history of abdominal or pelvic surgery. The incidence and severity of complications occurring after surgery are related to the surgical procedure and to the amount of peri-uterine tissues removed. The increase in dose and volume of external irradiation and brachytherapy and the increase in dose rate of the low-dose rate brachytherapy are responsible for the radiotherapeutic morbidity. The significant decrease of severe complication rates during the last 15 years was obtained by the implementation of individual adjustments in treatment planning. The treatment strategies of early bulky and advanced carcinomas are changing. Concurrent radiotherapy and chemotherapy is becoming a standard, but its late toxicity needs to be documented by a longer follow-up. The optimisation of radiation therapy should remain a reference to evaluate the outcome and morbidity of the new combined strategies: the addition of chemotherapy will never compensate for less than optimal radiotherapy/brachytherapy planning.