Feasibility and Reliability of Home-based Spirometry Telemonitoring in Allogeneic Hematopoietic Cell Transplant Recipients

By: Ajay Sheshadri1, Amin Alousi2, Lara Bashoura1, Karen Stolar2, Shiva Baghaie1, Muhammad H. Arain1, Laila Noor1, Amulya Balagani1, Akash Jain1, David Blanco1, Abel Ortiz1, Susan K. Peterson3, Renee Langhals4, Michael Taylor4, Alex Stenzler4, Rohtesh S. Mehta2, Uday R. Popat2, Chitra Hosing2, Gabriela Rondon2, Fan Shen4, Liang Li4, Guang-Shing Cheng6, David E. Ost1, Richard E. Champlin2, Burton F. Dickey1

Morbidity and mortality from bronchiolitis obliterans syndrome (BOS) remain unacceptably high after allogeneic hematopoietic cell transplantation (A-HCT) (1). Prompt diagnosis may improve outcomes (2). Adherence to home-based spirometry (HS) in lung allograft recipients is high, supporting the feasibility of BOS surveillance (3-6). Low adherence to HS has been a barrier to implementation after A-HCT (7-10), possibly due to psychosocial burnout and fatigue (11, 12). The goal of this pilot study was to: 1) determine the feasibility and validity of HS real- time telemonitoring in A-HCT recipients; 2) determine factors associated with adherence to HS; and 3) determine the variability of HS among participants without acute illness.


We consented and enrolled adult A-HCT recipients at around 100 days post-transplantation between October 2016 and June 2018 at a single transplant center, excluding those who had pneumonia within 4 weeks of screening. The MD Anderson Institutional Review Board approved the study (2015-0990). Participants received a Bluetooth®-compatible home spirometer (GoSpiro, Monitored Therapeutics Inc., Dublin, OH), which instantaneously transmits data to a cloud-based portal for review by patients and clinicians. Participants were trained in HS immediately following clinic-based spirometry (CS) and instructed to perform three maneuvers/session up to thrice weekly for 9 months, with the goal of recording at least one high-quality session/week. Week 1 measurements were used for training. Baseline FEV1 was defined as mean FEV1 in weeks 2-3. We defined adherence as recording at least one session during a Sunday-Saturday period. One week of non-adherence resulted in phone call and/or email reminders. All measurements were assessed for technical acceptability (13). FEV1 measurements that declined >10% from baseline values generated a daily email alarm to the study team. Declines in FEV1 of 10-19% sustained over two consecutive weeks, or >20% declines in FEV1 at any time and confirmed within 24 hours, triggered a clinical evaluation.

Each participant was expected to have three CS measurements – enrollment, 3 months and 9 months. Patient data was censored upon death, cancer relapse, or loss of contact with the study team for over one month. Mean HS measurements in the week before and after CS were compared to CS measurements using Bland-Altman analyses with limits of agreement and 95% confidence intervals (CIs). We used generalized linear mixed models with random intercept and random slope of time to analyze adherence. Linear mixed models were fitted to repeated measurements of HS FEV1 with a fixed effect of time and a random intercept. Intra-class correlation (ICC) was calculated to assess the longitudinal HS reliability, and 95% CIs for ICC were calculated with bootstrap. In addition, we calculated the average of within-subject coefficients of variation (CoV) across all participants, with 95% bootstrap CI. All statistical analyses were performed in R Version 3.5.2.


Figure 1 shows our cohort selection (n=51). The median age of the final cohort was 55 years (interquartile range 41-64); 34 (66.7%) participants were female, and 47 (92.2%) participants were non-Hispanic white. The most common malignancy was acute myeloid leukemia (49%); other malignancies requiring A-HCT included acute lymphoblastic leukemia (n=4), chronic lymphocytic leukemia (n=4), chronic myeloid leukemia (n=4), lymphoma (n=6), multiple myeloma (n=3), and myelofibrosis (n=2). 39 (76.5%) participants experienced acute graft-versus-host disease (GVHD) after A-HCT. Most participants had normal predicted lung function at baseline, and none experienced chronic dyspnea or cough at the time of enrollment, though one participant with normal baseline lung function subsequently developed intermittent asthma (14). Asymptomatic mild (n=3) and moderate (n=1) airflow obstruction were noted in a few participants, and one participant had mild restriction attributed to weakness, which resolved after A-HCT. Participants performed a mean of 3.4±3.5 acceptable measurements/week. Weekly adherence to HS was 69%. 75% of all loops met ATS/ERS technical criteria, and 94% of patient-weeks with measurements had at least one technically acceptable measurement. The primary reasons for missed sessions were non-adherence (403 patient-weeks), technical issues (85 patient-weeks), and health-related issues, such as hospitalization or acute outpatient illness (56 patient-weeks). Acute GVHD was associated with lower adherence (OR 0.37, 95% CI 0.13-1.02, p=0.05). We did not identify an association between chronic GVHD and adherence in this cohort (OR 0.85, 95% CI 0.3-2.2, p=0.71). The probability of performing one session/week dropped by 6% per week during the study (p=0.008).

The ICC of repeated measurements, after adjusting for time effects, was 0.85 (95% CI 0.75-0.91), indicating that only 15% of measurement variability was due to within-patient variation. The within-subject CoV for longitudinal FEV1 was 8% (95% CI 6.6%-9.5%). Similarly, adjusted ICC for FVC was 0.86 (95% CI 0.78-0.91) and the CoV was 7.3% (6.2-8.4%). We performed Bland-Altman analyses comparing HS to CS (Figure 2). HS FEV1 measurements systematically underestimated CS FEV1 by 0.33 L (95% CI -0.45 L to -0.22 L). Similarly, HS FVC measurements systematically underestimated CS FVC by -0.29 L (95% CI -0.42 to -0.17 L, Figure 3).

In a representative month during which 23 participants were concurrently being monitored, we spent 7 hours in total to remind non-adherent participants or provide technical assistance. Fourteen participants experienced significant pulmonary declines during the study. Of these, eight were due to poor spirometric technique and were easily identified and corrected remotely. The remainder had clinically confirmed pulmonary impairment, but no participants developed BOS during the study period.