A pragmatic, single-arm clinical trial of a dose modification algorithm for preventing cytopenia-related delays during FOLFOX chemotherapy

Abstract

Purpose:

Unplanned delays are common during FOLFOX chemotherapy. We evaluated a novel FOLFOX dose modification algorithm designed to reduce unplanned delays while maintaining dose intensity.

Methods:

We conducted a pragmatic, single-arm clinical trial to evaluate PAGODA, a proactive graduated dose modification algorithm for FOLFOX chemotherapy (NCT04526886). The algorithm prescribes chemotherapy dose reductions and delays based on absolute neutrophil count (ANC) and platelet count. Patients receiving FOLFOX-based chemotherapy were eligible. Participants received standard chemotherapy doses in cycle 1 (bolus 5-FU 400 mg/m2, oxaliplatin 85 mg/m2, and infusional 5-FU 2400 mg/m2/46 hours). The primary outcome was unplanned delay prior to cycle 6 (>18 days between cycles). We compared the incidence of unplanned delay against the historical proportion of 43%. Relative dose intensity (RDI) was a key secondary outcome.

Results:

There were 48 evaluable participants. The median age was 66, and 50% were female. The most common primary cancer sites were colorectal (n = 31) and gastroesophageal (n = 12). Sixteen of 48 subjects had any unplanned delay before completing cycle 6 (33%, 95% CI 0.22–0.47, p = 0.18 for comparison with the historical proportion) and 7 subjects had any cytopenia-related delay (15%, CI 0.07–0.27). Seven cycles were delivered without delay with an ANC of 750–999/microliter. The mean chemotherapy RDIs were: oxaliplatin, 86%; infusional 5-FU, 92%; and bolus 5-FU, 65%.

Conclusions:

The PAGODA dose modification algorithm was safe and was associated with a low rate of cytopenia-related delays. Further intervention refinement and testing may help to reduce unplanned delays and attendant time toxicity.

Background

Unplanned chemotherapy delays are a common and disruptive event during cancer treatment [1–3]. While such delays are sometimes necessitated by treatment-related side effects, many delays are triggered by asymptomatic changes in the neutrophil count and platelet count. Unplanned delays are disruptive for patients and their caregivers because they lead to additional travel, blood draws, clinic visits, and time away from home and work—a phenomenon increasingly recognized as time toxicity [4]. Unplanned delays are also inefficient for clinical teams, resulting in unused infusion capacity at the time of the delay and duplicative future clinic visits and laboratory testing. Lastly, delays that lead to reduced chemotherapy dose-intensity can have a negative effect on cancer treatment outcomes [5].

The modified FOLFOX chemotherapy regimen (mFOLFOX6; folinic acid, 5-fluorouracil and oxaliplatin) is a key component of treatment for colorectal and gastroesophageal cancers [6, 7]. Unplanned delays are particularly common among patients receiving FOLFOX [1, 8]; for example, a retrospective analysis of 214 patients with colorectal cancer found that 43% experienced one or more unplanned delays during the first six cycles of FOLFOX chemotherapy [2]. The majority of these delays were primarily attributable to neutropenia or thrombocytopenia, rather than symptomatic adverse effects. The justification for delaying chemotherapy treatment in the face of reduced blood counts is to reduce risk for infection or bleeding. However, blood count thresholds used for implementing a delay are not evidence-based. Clinical trial protocols of FOLFOX chemotherapy have historically mandated both a delay (usually 1–2 weeks) and a subsequent chemotherapy dose-reduction for patients with an absolute neutrophil count (ANC) of less than 1000–1500/microliter or a platelet count of less than 75,000/microliter [6]. This approach has been widely adopted in clinical practice. However, with nearly two decades of clinical experience with the mFOLFOX regimen, it is increasingly apparent that this reactive “delay and dose-reduce” approach is highly conservative and is associated with very low rates of fever and neutropenia (generally <5%) or clinically significant bleeding [1, 6, 9–11]. Outside of clinical trial protocols there are no widely accepted standards for chemotherapy dose and schedule modifications, and real-world clinical practice is variable and provider-dependent [12].

Given the negative effects of unplanned chemotherapy delays on patients, their families, and their clinical teams, we developed PAGODA, a proactive, graduated dose modification algorithm. PAGODA is a novel dose modification algorithm for FOLFOX chemotherapy designed to reduce the incidence of unplanned delays while maintaining chemotherapy dose intensity and safety of treatment. This manuscript describes a pilot study of PAGODA.

Methods

Study design and setting

We conducted a pragmatic, single-arm clinical trial to evaluate PAGODA (NCT04526886). This research was conducted in line with the principles of The Belmont Report of 1979. The study protocol was approved by the Dartmouth Hitchcock IRB. Participants were recruited at the Dartmouth Cancer Center from the Lebanon, NH, St. Johnsbury, VT, and Nashua, NH, clinic sites between September 14, 2020 and December 14, 2022.

Participants

Eligible patients included adults with gastrointestinal cancer who were planned to receive six or more treatments with a FOLFOX-based chemotherapy regimen, with or without concomitant monoclonal antibody therapy. Exclusion criteria included receipt of systemic chemotherapy (other than radiation-sensitizing chemotherapy) within the prior year, baseline absolute neutrophil count of <1500/microliter, baseline platelet count of <100,000/microliter, known dihydropyrimidine dehydrogenase (DPD) deficiency, uncorrected bleeding condition, or recent initiation (within 14 days) of therapeutic-dose anticoagulation. Eligible participants were enrolled in the study after completing day 1 of mFOLFOX6 chemotherapy using institutional standard dosing of chemotherapy drugs (bolus 5-FU, 400 mg/m²; leucovorin, 350 mg flat dose; oxaliplatin 85 mg/m²; infusional 5-FU, 2400 mg/m²/46 hours). All participants provided written informed consent to take part in the study.

Procedures

Following cycle 1, subsequent treatments with FOLFOX chemotherapy were scheduled in 14-day cycles. For each chemotherapy treatment in cycles 2–6, chemotherapy dose-reductions or delays were guided by: 1) neutrophil and platelet counts, and 2) other symptoms and adverse events. The complete blood count (CBC) was evaluated on day 1 of each planned chemotherapy cycle, or else one or two days prior to day 1. Dose modifications based on neutrophil and platelet counts were implemented by the treating oncology provider, based on the PAGODA algorithm. The algorithm was described in a 1.5-page document (see Online Resource 1) accessible to the study team via an internal website and within the protocol document. The algorithm prescribes graded chemotherapy dose reductions without delay for an absolute neutrophil count (ANC) of 750–1499/microliter or a platelet count of 50,000–99,000/microliter. For example, in a patient with ANC of 1200 on day 1 of cycle 2, the algorithm would prescribe treatment without delay with the following chemotherapy doses: bolus 5-FU, 200 mg/m²; oxaliplatin, 75 mg/m²; infusional 5-FU, 2400 mg/m²/46 hours. The algorithm prescribes a delay when ANC is less than 750/microliter or platelet count is less than 50,000/microliter on day 1 of cycles 2–6. Development of the dose modification algorithm was informed by prior research demonstrating that the incidence of febrile neutropenia remains low when FOLFOX chemotherapy is given at neutrophil counts similar to the threshold value of 750/microliter [10].

Chemotherapy dose modifications and delays for reasons other than neutropenia or thrombocytopenia were at the discretion of the treating provider, as were clinically-indicated dose modifications of concomitant monoclonal antibodies. Use of granulocyte colony stimulating factor (GCSF) was not permitted during protocol therapy, except in participants who had already incurred an unplanned delay (e.g., for ANC of <750/microliter). Use of GCSF after an initial unplanned delay was at the discretion of the treating oncologist.

Outcome measures

The primary study endpoint was the incidence of any unplanned chemotherapy treatment delay prior to delivery of cycle 6 of FOLFOX chemotherapy. Based on the standard cycle length for mFOLFOX of 14 days, a delay was defined as any interruption of therapy leading to a cycle length of >18 days. The 18-day interval allowed for minor schedule-related variations in treatment intervals. An unplanned treatment delay was defined as any treatment delay that was not premeditated as of day 3 of the preceding treatment cycle. Occurrence of any unplanned delay prior to completion of cycle 6 was assessed for each evaluable participant. Unplanned chemotherapy discontinuation was also counted as an unplanned delay. Participants were considered unevaluable for the primary endpoint if they discontinued chemotherapy prior to cycle 6 of FOLFOX due to elective surgery or disease progression, without having experienced an unplanned delay.

Secondary endpoints included chemotherapy relative dose intensity (RDI) and safety. RDI was defined as the ratio of the delivered dose intensity (dose delivered divided by days elapsed) and the standard dose intensity (standard dose divided by standard treatment interval) for cycles 1–6 [13]. RDI was assessed separately for each component of the FOLFOX chemotherapy regimen (oxaliplatin, infusional 5-FU, and bolus 5-FU). The safety endpoint was a composite measure of the occurrence of any of the following: febrile neutropenia (CTCAE v5.0 grade 3 or 4), major bleeding [14] with concurrent grade ≥3 thrombocytopenia (platelet count <50,000/microliter), any grade 4 neutropenia (ANC <500/microliter), and any grade 4 thrombocytopenia (platelet count <25,000/microliter).

Analysis

For the primary endpoint analysis we used a two-sided one-sample test of proportions to evaluate whether the observed proportion of patients with any unplanned delay differed from the estimated historical proportion of 43% [2]. The sample size was chosen to provide 80% power to detect a 50% reduction in the proportion of patients with unplanned delays, based on a one-sample test of proportions. This resulted in a planned accrual of 53 participants, in anticipation that at least 45 participants would be evaluable. Summary descriptive statistics were used for reporting of secondary outcomes and survey responses.

Participant survey

We conducted a survey of participants to better understand the patient experience and burdens of attending a scheduled chemotherapy treatment visit (see Online Resource 2). The purpose of the survey was to provide context for the opportunity cost associated with an additional provider visit at the time of an unplanned delay. Participants completed the survey on day 1 of cycle 3, 4, or 5.

Results

Fifty-two participants were enrolled over the 27-month enrollment period, and 48 were evaluable for the primary outcome of one or more unplanned chemotherapy delays prior to cycle 6 (see Figure 1). The median age among evaluable participants was 66 years (range, 41 to 81), and 50% were female. The most common primary cancer sites were colorectal (n=31; 65%) and gastroesophageal (n=12; 25%). Metastatic cancer was present in 67%, and 35% received a concurrent monoclonal antibody together with FOLFOX chemotherapy. Detailed clinical and demographic characteristics of participants are shown in Table 1.

Figure 1.

CONSORT diagram

^aParticipants were considered unevaluable if they discontinued chemotherapy prior to cycle 6 due to elective surgery or disease progression, without having experienced an unplanned delay.

Table 1.

Demographic and clinical characteristics of study participants, n = 48

Characteristic	n (%)
Age group
- <55 years	8 (17)
- 55–64 years	14 (29)
- 65–74 years	16 (33)
- ≥75 years	10 (21)
Sex
- Female	24 (50)
- Male	24 (50)
Primary cancer site
- Colorectal	31 (65)
- Gastroesophageal	12 (25)
- Other	5 (10)
Therapeutic intent
- Curative	21 (44)
- Palliative	27 (56)
Concurrent monoclonal antibody therapy
- Bevacizumab	7 (15)
- Trastuzumab	5 (10)
- Checkpoint inhibitor	5 (10)
- None	31 (65)

Primary outcome

Sixteen of 48 participants experienced an unplanned delay before completing cycle 6 of FOLFOX (33%, 95% CI 0.22–0.47, p = 0.18 for comparison with the estimated historical proportion of 43%). Of the 16 participants with unplanned delays, 15 had one delayed cycle (including three participants who discontinued FOLFOX chemotherapy before completing six treatments) and one had two delayed cycles. Cytopenia-related delays occurred in 7 participants and accounted for 7 of 17 unplanned delays (41%). Details regarding the per-cycle incidence of delays, unplanned delays, and cytopenia-related delays are shown in Table 2.

Table 2.

Incidence of delays in chemotherapy treatment, by cycle

Event	Cycle 2	Cycle 3	Cycle 4	Cycle 5	Cycle 6	Total
Evaluable cycles, n	48	47	46	46	45	232
Any delay, n (%)	4 (8.3)	6 (12.8)	4 (8.7)	0 (0)	4 (8.9)	18 (7.8)
Unplanned delay, n (%)	4 (8.3)	6 (12.8)	4 (8.7)	0 (0)	3 (6.7)	17 (7.3)
Cytopenia-related delay, n (%)	1 (2.1)	2 (4.3)	1 (2.2)	0 (0)	3 (6.7)	7 (3.0)

Secondary outcomes

Among the 45 evaluable participants who completed 6 cycles of FOLFOX chemotherapy, the mean (median) chemotherapy relative dose intensity (RDI) for each chemotherapy agent was: oxaliplatin, 86% (90%); infusional 5-FU, 92% (98%); and bolus 5-FU, 65% (67%). The mean interval from day 1 of cycle 1 to day 1 of cycle 6 was 74.9 days, with a range of 69 to 121 days.

The composite safety outcome (febrile neutropenia, thrombocytopenia with major bleeding, grade 3–4 neutropenia, and/or grade 3–4 thrombocytopenia) occurred in 3 patients (6%), all of whom experienced grade 4 neutropenia (only). Grade 4 neutropenia was deemed unrelated to the dose modification algorithm for two of these patients, including one participant with grade 4 neutropenia during cycle 1 of chemotherapy (after receiving standard initial chemotherapy doses) and one participant with grade 4 neutropenia after cycle 3, without having an ANC of less than 1500/microliter for any prior cycle (dose modification is not indicated in usual care for patients with a day 1 ANC of ≥1500/microliter). The third participant had transient grade 4 neutropenia related to the dose modification algorithm after cycle 5 of chemotherapy, without fever and without hospitalization.

Twenty-three participants (48%) had one or more algorithm-driven chemotherapy dose reductions, including 21 participants who received algorithm-driven dose reductions without an associated delay. Seven cycles of chemotherapy were delivered with dose reduction but without delay (and without GCSF) in 7 patients presenting with a day 1 ANC of 750–999/microliter. ANC was increased on the planned day 1 of the subsequent cycle for six of the seven patients; one participant had a decreased ANC on the planned day 1 of the subsequent cycle (grade 4 neutropenia), as described in the preceding paragraph. Detailed information about ANC response in the 7 participants who had a non-delayed chemotherapy treatment with day 1 ANC of 750–999/microliter is shown in Figure 2.

Figure 2.

Response of ANC in 7 participants receiving non-delayed chemotherapy with day 1 ANC of 750–999/microliter

ANC, absolute neutrophil count. Each line segment represents one non-delayed chemotherapy cycle with day 1 ANC of 750–999/microliter. The leftward point of each line segment indicates the ANC on day 1 of the cycle, and the rightward point indicates the ANC on the planned day 1 of the subsequent cycle.

Participant survey

We collected survey responses from 45 participants regarding burdens of attending a scheduled chemotherapy treatment visit. The median one-way travel distance to attend the treatment visit was 28 miles, and the median one-way travel time was 40 minutes. Most participants reported that they had traveled to that day’s clinic visit with a family member, friend, or caregiver (64%), and 43% reported current employment during cancer treatment. Five participants reported that they expected to have co-pays for their office visit and/or laboratory testing, with expected co-pay amounts of $50-$100. Four participants reported expected co-pays for chemotherapy treatment, with expected co-pay amounts of $20-$100. Additional information from the participant survey is shown in Table 3.

Table 3.

Survey responses relating to burdens of attending a chemotherapy treatment visit

	Number (%)	Respondents, n
One-way travel distance for treatment visit, median	28 miles	45
One-way travel time for treatment visit, median	40 minutes	45
Mode of travel to treatment visit		45
- Personal vehicle	36 (80)
- Vehicle of family member, friend, or caregiver	9 (20)
Did a family member, friend, or caregiver travel with you?		45
- Yes	29 (64)
- No	16 (36)
Did any of the people traveling with you adjust their work schedule to accompany you?		29a
- Yes	11 (40)
- No	18 (62)
Are you currently employed?		45
- Yes	19 (42)
- No	26 (58)
How many hours did you work for pay in the last 7 days? (median)	30 hours	18b
Did you modify your work schedule because of today’s treatment visit?		19b
- Yes	11 (58)
- No	8 (42)
Are you a non-employed caregiver for another person?		45
- Yes	4 (9)
- No	41 (91)

^aIncludes respondents who indicated that they traveled to the treatment visit with a family member, friend, or caregiver.

^aIncludes respondents who indicated that they were currently employed.

Discussion

We designed the PAGODA dose modification algorithm with the goal of preventing unplanned cytopenia-related chemotherapy delays during FOLFOX chemotherapy. These unplanned delays are common in both clinical trials and real-world care delivery, resulting in time toxicity (extra clinic visits and travel time), reduced chemotherapy dose intensity, and inefficient use of health care resources (provider, nurse, pharmacy, infusion, and administrative). In this pragmatic, single-center evaluation of the PAGODA algorithm, we demonstrated that the algorithm is safe and results in a low rate of cytopenia-related delays. Only 15% of the study participants (7 of 48) experienced cytopenia-related delays, compared with 34% of the participants in the retrospective study of Kogan and colleagues [2]. Nearly half of all patients had at least one algorithm-driven dose modification, and 7 patients received a non-delayed chemotherapy cycle with ANC in the range of 750–999/microliter. Chemotherapy dose intensity for oxaliplatin (mean RDI = 85%) and infusional 5-FU (92%) was similar to or higher than dose intensity described in other reports of FOLFOX chemotherapy [1, 13].

The PAGODA algorithm shares a common form with dose modification algorithms routinely used in clinical trials, but with several with notable differences. Unlike most clinical trial dose modification schemas, PAGODA follows an approach of “dose reduction without delay” for patients with moderate cytopenias, reserving the approach of “delay with subsequent dose reduction” for patients with more severe cytopenias. Also unlike most other dose modification schemas, PAGODA does not impose equivalent percentage dose reductions on each chemotherapy component of the FOLFOX regimen. Instead, the algorithm preferentially reduces the 5-FU bolus most aggressively—because this component of the FOLFOX regimen is a strong contributor to cytopenias, and because evidence suggests that the bolus may be non-essential for achieving desired treatment outcomes [15–17]. Our algorithm prioritizes preservation of the infusional 5-FU dose, maintaining the standard dose unless the ANC is less than 1000/microliter or the platelet count is less than 75,000/microliter. Importantly, the algorithm is only used after symptomatic toxicities are first considered and incorporated into the treatment plan by the cancer treatment team.

A different strategy for preventing unplanned chemotherapy delays is use of GCSF (e.g. pegfilgrastim). Long-acting GCSFs are costly; the average US sales price for a single administration of biosimilar pegfilgrastim at the time of this writing is approximately $1400 [18]. Depending on insurance and co-insurance provisions, a significant portion of this cost may be passed on to patients. GCSFs do not prevent thrombocytopenia and can also cause bothersome side effects, such as bone pain. For all of these reasons, routine use of primary prophylactic GCSF during FOLFOX is not recommended [19, 20]. Furthermore, the real-world effectiveness of GCSFs for preventing unplanned chemotherapy delays is uncertain. One randomized phase 2 study of patients receiving various biweekly chemotherapy regimens for colorectal cancer showed that prophylactic GCSF reduced the incidence of dose delays; however, the trial protocol required delay of chemotherapy until recovery of the ANC to at least 1500/microliter [21]. A retrospective chart review of 140 patients receiving adjuvant FOLFOX at a multi-site community practice found delays in 8.8% of cycles without GCSF support, compared with delays in 15.9% of cycles with GCSF support [12]. In our pilot trial, GCSF use was prohibited prior to the occurrence of an unplanned delay, and only 3 of 48 participants (6%) received GCSF during the study.

We view the PAGODA dose modification algorithm as a form of clinical decision support. Clinical decision support tools have a broad and growing footprint in cancer care delivery, even as studies examining their impact remain limited [22]. Like other clinical decision support tools, the dose modification algorithm provides a timely management recommendation relevant to a specific clinical decision, using relevant patient-specific characteristics (neutrophil count, platelet count, and absence of other symptoms requiring chemotherapy dose reduction or delay). In this way, the dose modification algorithm specifies a default approach for when and how to modify chemotherapy treatment in the context of moderate neutropenia or thrombocytopenia [23].

Another way of thinking about PAGODA is as a situationally specific heuristic, or decisional shortcut [24]. While many decisions in clinical oncology are evidence-based, most day-to-day decisions are not. Much of daily clinical practice is governed by the operational heuristics that clinicians develop in the course of their careers, through training and experience. A strength of this approach is its efficiency; however, the heuristics of individual oncologists are more intuitive than analytical. By adopting interventions like the PAGODA dose modification algorithm, oncologists can replace their general, intuitive operational heuristics with a data-informed decisional shortcut designed for a specific clinical scenario (how to incorporate neutrophil count and platelet count into chemotherapy dose modification), maintaining the desirable cognitive efficiency of heuristic decision making.

The development of the FOLFOX dose modification algorithm was motivated by our team’s desire to reduce time toxicity and inefficiency of care that are often consequences of unplanned chemotherapy delays. The results of the embedded patient experience survey highlight the burdens associated with attending a cancer treatment visit, including substantial travel for patients— and often for their family members, friends and caregivers as well. There is increasing attention to the importance of understanding and alleviating the time burden of healthcare for patients with cancer and other serious illnesses [4, 25], and some have advocated for “healthy days at home” as a patient-centered measure of care quality [26, 27]. Specific interventions to reduce time toxicity in cancer care are needed, and improving the patient-centeredness of frequent chemotherapy treatments presents a favorable target for intervention. Unplanned chemotherapy delays caused by marginal blood counts result in a substantial disruption to the lives of patients and their care partners, and such delays are not limited to patients with gastrointestinal cancer [28, 29].

We interpret the results of this pilot trial of the PAGODA dose modification algorithm as promising, showing a low rate of unplanned chemotherapy delays, with very few cytopenia-related delays. Critically, the dose modification algorithm was safe, without occurrence of fever and neutropenia or prolonged neutropenia. Further evaluation of PAGODA is warranted in order to confirm that this approach reduces unplanned delays in real-world care delivery settings.

Data Availability Statement.

The data that support the findings of this study are not openly available due to reasons of sensitivity. Deidentified data are available from the corresponding author upon reasonable request. Data files are deposited in controlled access data storage at the Dartmouth Hitchcock Medical Center.