Real-world Clinical Outcomes of First-Line Ibrutinib or Chemoimmunotherapy in Patients with Chronic Lymphocytic Leukemia by Risk Status

This was a retrospective, multisite, medical chart review study of treatment patterns and clinical outcomes in patients with CLL, stratified by genetic risk status, who received either ibrutinib or CIT as the first-line treatment. The study was approved by the New England Institutional Review Board, a WIRB Copernicus Group Company (#120190010).

High-risk patients were those with confirmed presence of at least one of the following genetic abnormalities: del(17p), del(11q), unmutated IGHV, TP53 mutations, and complex karyotype (defined as having at least three chromosomal abnormalities) [18]. Non-high-risk patients included those with confirmed absence of all aforementioned abnormalities (excluding complex karyotype). As complex karyotype is not routinely tested in clinical practice, if this information was not available for any patients, they were categorized as non-high risk if they had none of the del(17p), del(11q), unmutated IGHV, or TP53 mutations.

Data were collected from March 2019 to July 2019, after the approval by the New England Institutional Review Board, a WIRB Copernicus Group Company (#12019010). Oncologists and/or hematologists throughout the USA were identified in a national physician database and invited to participate. Physicians who had treated or managed at least 10 patients with CLL in the past 12 months were randomly sampled from the database to represent different geographic areas and practice settings. These physicians were invited for participation if they were not currently acting as a consultant or clinical investigator for a company involved in marketing or producing of oncology drugs. Participating physicians received honoraria.

Patients with CLL who had been treated by the participating physicians with either ibrutinib or CIT as first-line therapy were identified, and relevant data were extracted to an electronic or paper case report form. The collected data were de-identified and included no protected health information. The extracted data encompassed the period from a patient’s diagnosis of CLL through the patient’s most recent visit or the end of the data collection period (December 31, 2018), whichever occurred first. If a patient died before December 31, 2018, the patient was censored.

The patients included were adults 18 years of age or older, had a confirmed diagnosis of CLL, received first-line treatment for CLL with either single-agent ibrutinib or a CIT regimen consistent with the treatment guidelines at the time [19], and had an index date between February 1, 2014 and December 31, 2016 (the index period). The index date was defined as the initiation of first-line treatment for CLL. The index period was chosen to encompass at least 2 years of data accumulation after the initial US approval of ibrutinib. The CIT regimens were any combinations of an immunotherapy agent among alemtuzumab, obinutuzumab, ofatumumab, and rituximab and one or more chemotherapy agents among bendamustine, chlorambucil, cyclophosphamide, cytarabine, doxorubicin, fludarabine, oxaliplatin, pentostatin, vincristine, and cladribine.

Patients were excluded from the chart review if they were enrolled in a cancer clinical trial after the date of CLL diagnosis or if their medical history was incomplete from diagnosis through the most recent visit (or death). Complete medical history included available CLL diagnosis details, first-line and subsequent treatments, procedures, tests, and follow-up information.

The outcomes evaluated were treatment patterns and the time to next treatment (TTNT), summarized using descriptive statistics for each of the following patient groups: (1) high-risk treated with ibrutinib, (2) non-high-risk treated with ibrutinib, (3) high-risk treated with CIT, and (4) non-high-risk treated with CIT.

An assessment was conducted before treatment group comparisons to confirm that a balance could be achieved between the groups for confounders. This balance, termed “equipoise,” is a method frequently used to assess patient similarity and the feasibility of conducting unbiased comparative research between two naturalistic groups [20]. Clinical equipoise was quantified using a preference score, derived from a propensity score, to assess potential overlap of key clinical variables between the ibrutinib and CIT groups. The variables included demographics (age at CLL diagnosis, gender, year of index date, race), clinical characteristics (including family history of CLL, body mass index at diagnosis, Rai stage at diagnosis, comorbidities, Charlson Comorbidity Index score), and cytogenetic/molecular testing status (TP53 mutation, del(17p), del(11q), unmutated IGHV). If more than half of the patients in each group had preference scores between 0.3 and 0.7, the two groups would be considered sufficiently similar for comparative effectiveness research [20]. The equipoise assessment confirmed that comparative analysis was suitable between the high-risk ibrutinib and CIT patient groups. The non-high-risk groups did not reach the acceptable variation between ibrutinib and CIT and were considered not sufficiently similar.

In real-world studies, the treatment chosen for each patient is usually not entirely random, and patient characteristics may be potential confounders in head-to-head comparisons. To minimize the effect of confounders in the comparison between the high-risk ibrutinib and CIT groups, we used inverse probability of treatment weighting (IPTW) [21] to balance key characteristics. First, a logistic regression model was developed to generate a propensity score to reduce the selection bias on the estimates of the treatment effect. Subsequently, the propensity score weight was calculated as the inverse of the propensity score. Each patient was assigned a weighting coefficient, which was the inverse of the propensity score for the ibrutinib group and the inverse of (1 − propensity score) for the CIT group. As a result, the sample size in each weighted treatment group, generated from the IPTW process, was different from the original dataset, even though the same patients contributed to the dataset. Once the IPTW dataset was generated, standardized differences in means (SDM) were used to evaluate the balance in the key characteristics between the weighted high-risk ibrutinib and CIT groups (see Table S1 in the electronic supplementary material for details). When the SDM was less than 10%, a good balance between the groups for a given baseline characteristic was confirmed. P values from propensity score-weighted t tests were also generated to further evaluate the balance in baseline characteristics.

The outcome TTNT was used as a surrogate for progression-free survival and was defined as the duration from the date of first-line treatment initiation (either CIT or ibrutinib) to the start date of the second-line treatment for CLL. TTNT was compared between the following paired groups: (1) weighted high-risk ibrutinib versus weighted high-risk CIT, (2) high-risk versus non-high-risk ibrutinib, and (3) high-risk versus non-high-risk CIT. The comparisons between high-risk and non-high-risk patients for either treatment did not involve weighted analysis, because high-risk and non-high-risk patients were prescribed the same medication of interest (CIT or ibrutinib).

Patients who did not initiate a second-line treatment by the data cutoff date (December 31, 2018) were censored at the date of their last follow-up visit or date of death, whichever was later. TTNT was evaluated using Kaplan–Meier curves, and median durations with 95% confidence intervals (CIs) are reported. Paired comparisons in TTNT between the previously specified groups were conducted using Cox proportional hazards regression analyses. Log-rank tests were used to assess statistical significance in the comparisons between groups. The Kaplan–Meier curves were truncated at the time point when the number of patients at risk (patients who had not reached second-line treatment) dropped below 10% in either of the compared groups [22]. Median follow-up duration was calculated using the reverse Kaplan–Meier method [23].