2.1 Patient enrollment

Patients with SCA, age 12 years and older, who had been on HU for more than 3 months, were eligible. Exclusion criteria included chronic transfusions, transfusion within 8 weeks of enrollment, hospitalization for vaso-occlusive crisis or acute chest syndrome within 30 days of enrollment, alanine aminotransferase four times the upper limit of normal, physical inability to exercise, and unavailability for timed study activities. To ensure adherence to study drug, patients with good HU adherence, evidenced by Hgb F level and hemogram, were offered participation.

Written IRB-approved informed consent was obtained. Participants age 12–17 provided written assent. Patients were compensated with gift cards for CPET completion and mileage reimbursement for travel to exercise lab.

2.2 Intervention and assessments

One month after enrollment, participants had baseline labs drawn and performed CPET#1. The following day, voxelotor 1500 mg daily was started. After 1 month of voxelotor, monitoring clinic visit and routine labs were done. After 2 months of voxelotor, routine and study labs were drawn, and CPET#2 was conducted. The final study visit was 1 month following CPET#2, when quality-of-life assessments were obtained. Weekly phone calls were made to monitor adverse events and answer questions.

2.3 Biochemical markers

Blood samples were obtained before each CPET. Routine labs included complete blood count, reticulocyte count, comprehensive metabolic panel, lactate dehydrogenase, indirect and direct bilirubin, and Hgb F. Blood samples were shipped on ice overnight to the UCSF RBC lab for study labs, including oxygen dissociation (Hemox Analyzer, TCS Scientific Corp), and percentage of cells containing Hgb F (% F cells) by flow cytometry (Fortessa, Becton Dickinson).

2.4 Exercise testing

CPET was performed using a motorized treadmill (Trackmaster Model TMX428, Full Vision, Inc.), and a modified Bruce Protocol, which includes 2-minute stages that increase in incline and speed incrementally to induce peak physical effort. Participants were asked to exercise for 8−12 minutes, but could terminate any time they exhibited signs of inability. Breath-by-breath gas exchange data were collected and analyzed using a VMax Encore 29C metabolic cart (CareFusion Corp, San Diego, CA, USA). A 12-lead electrocardiogram (ECG) and pulse oximetry were used to monitor participants during exercise. Peak heart rate (peak HR), peak blood pressure, peak oxygen consumption (peak VO₂), oxygen pulse (O₂ pulse), VE/VCO₂ slope to the respiratory compensation point (minute ventilation/carbon dioxide production; a marker of ventilatory efficiency), VO₂/HR slope, ventilatory anaerobic threshold (VAT; when the shift from aerobic to anaerobic metabolism begins), percent predicted VO₂ at VAT, and time exercised were measured. A maximal exercise response is defined as a respiratory exchange ratio > 1.1, meaning excess CO₂ was produced from O₂ consumed for exercise, which is evidence of participant maximal effort. Prediction calculations included Cooper's equation for height or weight and Wasserman's equation; all were adjusted for treadmill.

2.5 Adherence monitoring

Participants were asked to bring back all voxelotor bottles to each clinic visit. Pill counts were used to calculate percent adherence. Hemoglobin electrophoresis detected voxelotor-modified Hgb. The distinct shift in the oxygen dissociation curve (ODC) curve at low PO₂ as the result of voxelotor further confirmed adherence.

2.6 HRQoL

Health-related quality of life (HRQoL) was assessed by patients and clinicians via the Patient Global Impression of Change survey (PGIC) and the Clinical Global Impression of Change-Improvement survey (CGIC), respectively. The surveys rated patient overall improvement after voxelotor on a seven-point scale.

2.7 Statistical analysis

Stata/IC version 16 (StataCorp, College Station, TX, USA) was used for statistical analysis. All study parameters were subjected to descriptive statistics. Student paired t-test was used to compare CPET#1 and CPET#2 peak VO₂. p-Values less than or equal to .05 were considered statistically significant. Power could not be calculated for this pilot study of only 10 subjects.

3.1 Study participants

Between October 9, 2020 and June 9, 2021, 14 patients were enrolled. Four subjects were withdrawn due to one screen failure, one adverse event on study drug with fever and chills, one hypersensitivity reaction with fever, hives, swelling, requiring steroid therapy, and one did not keep HU dose constant. Ten participants completed the study, including five females and five males aged 12−24, with average age 16.2 ± 3.9 (Table 1).

All participants had Hgb SS with stable Hgb F levels on HU, which was continued without dose change throughout the study. All voxelotor pill bottles were returned, and pill counts confirmed adherence at 84%–100% (average 98%).

3.2 Hematology and biochemical markers

All lab studies were collected at designated times and showed expected changes after voxelotor, confirming that the drug worked (Table 1, Figure 1A). Mean Hgb change was +1.6 ± 1.3 g/dL, range +0.4 to +4.9 g/dL. Mean reticulocyte change was −2.0% ± 2.5%, range −7.4% to +2%. Mean total bilirubin change was −0.4 ± 0.4 mg/dL, range −0.9 to +0.3 mg/dL.

Accurate percent Hgb F quantitation was challenging after voxelotor due to drug modification causing variant hemoglobin species. Percent Hgb F was more reliably obtained from capillary zone electrophoresis (Labcorp), and changed by −0.4% to −16%, with an average decrease of −7.1% ± 5.5%. Hgb F % dropped significantly for all but two subjects, including decreasing by approximately half in four subjects (Table 1). Flow cytometry showed 71.4%–93.4% F cells at baseline. After voxelotor, % F cells decreased in all but one subject, ranging from −8.1% to −24.1%, with average change of −11.3% ± 7.0% (Table 1).

3.3 Oxygen dissociation curves

All participants demonstrated a leftward shift of P₅₀, ranging from −23 to −6.2 mmHg, with mean and significant shift of −11 mmHg, from 33 to 22 mmHg (p < .0001) (Figure 2). For reference, P₅₀ is 35 mmHg for Hgb S, 27 mmHg for Hgb A, and 20 mmHg for Hgb F. In contrast to the classical sigmoidal shape, all voxelotor-treated ODCs had a steeper slope at lower partial pressure of oxygen, reflecting proportionally less oxygen off-loading in more oxygen-deprived environments.

3.4 CPET

All subjects completed CPET#1 and CPET#2. All subjects attained respiratory quotient (VCO₂/VO₂) ≥ 1.1, confirming maximal exercise effort. Subject 10 showed ST depression on ECG during CPET#1, thought to be due to anemia, as his Hgb was 7 g/dL. He underwent cardiac workup that was ultimately negative, including echocardiogram and CT angiogram. This delayed his CPET#2 until 7 months later, at which time his Hgb rose by 4.9 to 11.9 g/dL, and no ECG changes occurred during CPET#2. No other adverse events occurred during exercise testing.

Reported reproducibility of CPET ranges from 2% to 10%; this exercise lab accepts ±5% for intra-individual variability, and a difference of more than 5% in either direction is considered significant. Oxygen consumption was measured as peak VO₂ in L/min, in mL/kg/min, and as % predicted (Table 2). Peak VO₂ changes (in L/min) before and after voxelotor ranged from −0.38 to +0.32 L/min. To account for sex, age, and other variables, peak VO₂ was also compared as % predicted, and individual changes ranged from −12.8% to +11.3%. Subject 1 was the only participant with significant positive change of +11.3% in peak VO₂. Four subjects showed insignificant change (Subjects 7, 8, 10, 11). Five subjects showed significant negative change (Subjects 2, 5, 6, 9, 12) (Table 2, Figure 1B).

Peak VO₂ calculated in mL/kg/min showed similar results. Change in peak VO₂ expressed in L/min or in mL/kg/min both showed only one subject demonstrated significant improvement (Subject 1), and two subjects (Subjects 6, 9) showed significant reduction. The remaining seven subjects showed either minimal or no significant change in either direction (Table 2, Figure 1C).

Changes in individuals’ anaerobic threshold (AT), O₂ pulse, VE/VCO₂ slope, and time exercised were not consistent and did not correlate with changes in peak VO₂. Individual changes in hematologic parameters also did not correlate with changes in peak VO₂ or other CPET measurements.

3.5 PGIC/CGIC

On the seven-point questionnaires evaluating activity limitations, symptoms, emotions, and overall quality of life, PGIC and CGIC were concordantly positive for seven of 10 participants, and discordant for three of 10 participants reporting no change or almost the same, while clinicians reported improvement (Table 3). Clinicians reported basing their CGIC response mostly on Hgb change after voxelotor.

4.1 Voxelotor and O₂ consumption in study patients

Exercise tolerance is increasingly relevant in sickle cell disease with improving medical care and more SCA patients striving for lifestyles similar to their peers. Higher hemoglobin holds promise for improved exercise capacity. Our hypothesis that voxelotor, the primary clinical outcome for which is higher hemoglobin, could increase peak VO₂ on CPET for adolescents with SCA, was not supported by data from this pilot study. Only one of 10 participants demonstrated significant positive change in peak VO₂ after voxelotor, while two of 10 participants performed significantly worse. Notably, the participant with the largest hemoglobin increase (Subject 10) from 7 to 11.9 g/dL while on voxelotor for 7 months between CPET#1 and CPET#2, only demonstrated +6% change in peak VO₂ and +2.5% change in % predicted peak VO₂, which were not significant given ±5% cutoff for intra-individual variability (Table 2, Figure 1B). Although only one example, this participant demonstrated that even impressive correction of severe anemia with voxelotor for longer time did not result in higher peak VO₂.

Study subjects demonstrated lower-than-normal O₂ pulse (Table 2). O₂ pulse is typically considered a surrogate for stroke volume, but based on the Fick equation, also includes O₂ extraction, which is affected by hemoglobin and O₂ saturations. No patient had significant O₂ desaturation, and hemoglobin increased in all patients on voxelotor, suggesting impairment in either oxygen extraction or stroke volume as reasons for low O₂ pulse. Peak HR was higher than 85% predicted for all patients, confirming no chronotropic incompetence. Cardiac physiology can be abnormal in SCA due to diastolic dysfunction, abnormal muscle function, and decreased pumping action.^{24, 25} Cardiac assessment was not included in this pilot study, but was normal for one subject when done outside of the study, which showed that lack of peak VO₂ improvement in this subject (Subject 10, with hemoglobin rise of 4 g/dL) was not due to cardiac dysfunction. Abnormal vasculature can affect pulmonary and systemic vascular resistance, but VE/VCO₂ was normal in all patients, suggesting no significant ventilation to perfusion (V/Q) mismatch or intrapulmonary shunting (patients did not have to excessively ventilate to remove CO₂). This leaves limited oxygen extraction due to high-affinity voxelotor-modified hemoglobin as a plausible reason for low O₂ pulse. In support of this concept, in a detailed study of voxelotor in normal volunteers showing oxygen saturation improvement during incremental exercise, oxygen extraction and consumption also did not change. The researchers suggested that differential O₂ off-loading offsets the advantage of higher O₂ saturation, leading ultimately to no change or decrease in peak VO₂.²⁶ Interestingly, GBT reported normal oxygen delivery in exercise testing of normal volunteers taking voxelotor, as assessed by resting and peak exercise heart rate, but no VO₂ data were presented.²⁷

Additionally, when assessing VO₂ at AT, which represents the highest VO₂ (aerobic capacity) before converting to anaerobic metabolic pathways, many patients had a lower value on the subsequent CPET, indicating less O₂ was extracted. This includes Subject 1 who had a higher peak VO₂ on the second test. VO₂ at AT is typically approximately 50%–60% of the predicted peak VO₂, but more than 40% is considered normal. Four of 10 patients on the second CPET had values lower than 40%, while none had values less than 40% on the first CPET (Table 2). This is an unusual finding and one could speculate whether voxelotor affected O₂ extraction.

4.2 Voxelotor-modified Hgb and oxygen dissociation curve

Voxelotor directly increases oxy-hemoglobin, which should increase oxygen-carrying capacity. Indeed, detailed exercise studies with pulmonary gas exchange and arterial blood gas measurements by Stewart et al. showed higher arterial oxygen saturation during exercise after voxelotor, consistent with left-shifted ODC.²⁶ However, peak VO₂ decreased after voxelotor, which the authors suggested may be due to altered oxygen off-loading by voxelotor-modified Hgb. Using a sickling assay and structural modeling, Henry et al. concluded that potential oxygen delivery to tissues gained from reduced sickling is offset by less oxygen off-loading.²⁸ Here, oxygen consumption did not improve in 9 of 10 SCA patients after voxelotor, while ODC assumed a hyperbolic shape and shifted leftward (Figure 2A), with a significant drop in average P₅₀ of −11 mmHg (Figure 2B). Thus, live exercise data in SCA patients reported here are consistent with the model of decreased oxygen off-loading, although other explanations, such as cardiac limitations, cannot be excluded.

The usual sigmoidal shape of ODC ensures more oxygen off-loading from hemoglobin as oxygen tension decreases and tissue oxygen demand increases. Instead, ODC of hemoglobin containing voxelotor-modified species has the opposite hyperbolic shape, showing less proportional oxygen off-loading, that is, less oxygen delivery, when tissues are more hypoxic (Figure 2A). The need to balance voxelotor's ability to prevent sickling by increasing oxygen affinity against voxelotor's negative effect on oxygen delivery was ardently articulated in a commentary by Quinn and Ware,²⁹ and the recent NEJM review by Bunn.³⁰ High-affinity Hgb F also shifts ODC to the left, but the sigmoidal shape is maintained, as is shown in our patients with high Hgb F prior to starting voxelotor (Figure 2A). Although we have no data on in vivo fractional oxygen extraction or tissue oxygen status in the study subjects, the remarkably altered ODC of voxelotor-modified hemoglobin, low O₂ pulse, and lower VO₂ at AT are consistent with limited oxygen off-loading.

4.3 Peak VO₂ in SCA patients on hydroxyurea

Peak VO₂ at baseline for the participants ranged from 53.4% to 92.9% of predicted, with an average of 75.6%, which is roughly consistent with published reports that CPET performance is approximately 30% lower in patients with SCA compared to control populations (Table 2, Figure 1C).^{11, 12, 16} This is surprising in that HU increases Hgb and Hgb F, and prevents sickling, which might be expected to positively affect exercise tolerance. All study participants were long-term HU users, and some were athletically active, yet their peak VO₂ overall fell in the range of untreated sickle cell patients. Baseline average P₅₀ at 33 mmHg is to the right of P₅₀ for Hgb A, which is 27 mmHg, thus oxygen off-loading is not a problem in patients with high Hgb F. An ancillary study from the Multicenter Study of Hydroxyurea tested 10 subjects on HU and 14 placebo subjects by cycle ergometry at 6, 12, and 18 months and found more increase in peak muscle power and more decrease in peak heart rate response to workload among HU subjects, reflecting improvements in anaerobic muscle performance and aerobic cardiovascular efficiency, suggesting improvement in exercise capacity associated with HU.³¹ Other publications reporting VO₂ have not separated out HU-treated versus untreated SCA patients, perhaps because HU-treated patients did not stand out on exercise testing.

4.4 Improving exercise capacity in SCA patients

Studies that showed improvement in exercise capacity with higher Hgb in SCA patients have increased Hgb by transfusion with Hgb A blood.^{20, 21} In the exchange transfusion experiment attempting to keep Hgb the same (but actually Hgb increased by 1.4 g/dL), the effect of Hgb A was directly queried and results suggested improved exercise capacity was associated with Hgb A.²¹ In contrast, in a study of children on long-term exchange transfusion maintaining Hgb S approximately 30%, average peak VO₂ was still less than 80% compared to healthy controls, but the transfused patients also had significantly lower Hgb (9.3 ± 0.5 g/dL) than healthy controls.³²

Several studies documented improved exercise performance in SCA patients after moderate exercise training.^{1, 5, 7-9} Anecdotally, Subject 1 in this study, the only one who demonstrated increase in peak VO₂, reported engaging in regular exercise after starting voxelotor, but this was not prospectively or objectively documented. At least in one study, muscle biopsy demonstrated altered skeletal microvasculature in SCA patients limiting oxygen extraction, which improved after a regular exercise program, along with improved peak VO₂.^{6, 7} It is possible that improved Hgb along with guided exercise training could lead to sustained improvement in exercise capacity.

4.5 Voxelotor effect on Hgb F

A notable finding of voxelotor is its effect on Hgb F. Voxelotor treatment decreased % F cells, but due to higher RBC number after voxelotor, the absolute number of F cells remained largely unchanged. However, % Hgb F, the absolute amount of Hgb F, as well as the amount of Hgb F per F cell decreased in many subjects after voxelotor (Table 1 and Table S1). In addition, voxelotor modification of Hgb F would also alter ODC of Hgb F, raising the same concern that impairment of oxygen delivery from voxelotor-modified Hgb F counters the benefit of Hgb F in inhibiting Hgb S polymerization. Given that Hgb F thus far is the singularly most effective molecule in protecting against deleterious complications of sickle hemoglobin, potential interference with Hgb F level and function is another consideration in using voxelotor in combination with HU.

4.6 Limitations

Limitations of this study include small sample size, lack of cardiac function assessments, and lack of spirometry at baseline or post exercise to determine potential pulmonary limitation to exercise.