Criteria for identifying acute respiratory events based on FEV1 decline in home spirometry for lung transplant patients

Abstract

Background

Lung transplantation is a critical treatment for end-stage lung diseases, but long-term survival is challenged by graft rejection and infection. The detection of adverse respiratory events depends on home spirometry, which can exhibit greater fluctuations than laboratory tests and may not provide timely alerts. Our LT-FollowUp system offers an internet-based platform for daily FEV1 monitoring. This paper explores whether a new algorithm using LT-FollowUp data can detect the clinically significant FEV1 declines that predict adverse respiratory events.

Methods

A retrospective cohort study of lung transplant patients from the University of Tokyo Hospital was conducted using LT-FollowUp. The accuracy of the algorithm was evaluated using a nested case-crossover study comparing FEV1 declines before acute respiratory events with control periods, and a nested case-time-control study comparing cases with matched controls to adjust for time trends and bias.

Results

Of the 95 patients included in this study, 21 experienced acute respiratory events. The odds ratios derived from conditional logistic regression in the nested case-crossover study and the conditional logistic regression in the nested case-time-control study are 5.42 × 10⁵ and 1, respectively. There is a clear association between abnormal FEV1 decline and acute respiratory events. No clear time trend is observed.

Conclusion

The proposed algorithm using LT-FollowUp data shows promise for the real-time detection of respiratory events in lung transplant patients, potentially facilitating early interventions that may prevent chronic lung allograft dysfunction. Further validation in larger, multi-centre studies is needed to confirm these findings and enhance clinical utility.

Introduction

Lung transplantation is a crucial therapeutic option for end-stage lung diseases. However, the long-term survival of grafts and patients is limited by adverse events such as rejection and infection. Traditionally, the detection and diagnosis of these events rely on laboratory pulmonary function tests. One key measure is the forced expiratory volume in one second (FEV1), with a baseline value calculated from the two highest post-operative measurements taken at least three weeks apart. A decline in FEV1 of over 10% from this baseline indicates the need for further examination to detect potential abnormalities, with the aim of preventing progression to chronic lung allograft dysfunction (CLAD) [1]. However, these tests are infrequent, potentially delaying the detection of adverse events.

To improve monitoring, daily FEV1 measurements via home spirometry are widely used. The reproducibility of these measurements allows for the early detection of graft abnormalities [2]. However, reliance on handwritten patient charts limits real-time data sharing with physicians, reducing the effectiveness of home spirometry. The LT-FollowUp system we have developed is an internet-based platform that enables the daily input of weight, vital signs, pulmonary function measurements (FEV1 and FEV6), and immunosuppressant doses. These data, input via smartphones, tablets, or computers, are sent to a cloud server, providing real-time feedback to patients and medical personnel. A pilot study demonstrated the system’s viability and patient satisfaction as an alternative to handwritten charts [3,4,5].

Home spirometry results can exhibit greater fluctuations than laboratory tests conducted by professionals, potentially rendering single-time recordings less reliable. Thus, new criteria may be needed for the detection of adverse events when using LT-FollowUp. This study assesses whether a new algorithm leveraging LT-FollowUp can detect clinically meaningful FEV1 declines that predict adverse respiratory events, providing real-time alerts to medical personnel and patients. We evaluate the algorithm’s accuracy by examining its correlation with adverse effects detected by clinical diagnostic tests.

Methods

This study followed the 2015 Standard for Reporting Diagnostic Accuracy (STARD) statement, except for points which were not compatible with the original statement [6].

Development of algorithm to detect abnormality in FEV1

The algorithm development involved discussions among clinical experts on the general trends and temporal changes in FEV1 for patients with and without clinically diagnosed pulmonary events. The focus was on measurement fluctuations in stable patients and the likelihood of true pulmonary events when measurements deviated from typical fluctuations (Fig. 1). Once developed, the algorithm’s accuracy was examined.

Fig. 1

Screenshots of LT-FollowUp. a Login screen. b Input screen. c Data list screen. The data are displayed as a list. d Graph screen. Input values are visualised graphically. e Example of a patient with relatively large daily fluctuations in FEV1 values

Settings and participants for examining the accuracy of the newly developed algorithm

All post-lung transplant patients followed up at the University of Tokyo Hospital by June 2023 were considered for inclusion. Typically, patients are discharged 1–3 months post-transplantation and begin using LT-FollowUp before discharge, measuring pulmonary function and other metrics twice daily. The inclusion criteria were as follows: (1) the patient had undergone lung transplantation at the University of Tokyo Hospital, (2) the patient had a filling rate of at least 30% in the morning or evening (the filling rate is the number of days with LT-FollowUp inputs divided by the number of days from the first entry date to the last entry date). The filling rate must exceed 30% both 0–90 days and 91–180 days ahead of the acute respiratory event. The following patients were excluded from the study: (1) patients with missing data during the week prior to the onset of the acute respiratory event, (2) patients who suffered acute respiratory events within three months of their initial use of LT-FollowUp. Follow-up was conducted from October 2020 to December 2023.

Definition of acute respiratory events diagnosed by established clinical diagnostic tests

We retrospectively reviewed bronchoscopy and patient charts to identify acute respiratory events post-discharge. The conditions included were pneumonia, all types of rejection or rejection-related events, and anastomotic stenosis including stent obstruction. Pneumothorax was excluded as it typically presents with early peripheral symptoms such as chest pain, cough, and dyspnoea, which do not usually result in a significant reduction in FEV1 as measured by home spirometry. COVID- 19 infection was excluded because of hospitalisation policies in Japan. Patients were classified as having acute events based on findings during regular or emergency visits requiring bronchoscopy or hospitalisation, which included (i) abnormal opacities on chest radiographs or CT scans and/or (ii) clinically significant FEV1 declines confirmed by a pulmonary function test at the hospital. Patients who did not fall into the above-mentioned categories were defined as having no respiratory event. Patients who underwent pulmonary resection for malignancy after lung transplantation were censored at the time of intervention. To avoid the influence of FEV1 declines associated with previous events, only the first event after the introduction of LT-FollowUp was considered for analysis.

Detection of abnormality in FEV1 recorded by LT-FollowUp using the newly developed algorithm

In developing the algorithm, clinical experts discussed temporal changes in the test values of several typical patients, focusing on measurement fluctuations in stable patients and the likelihood of actual pulmonary events when a measurement exceeded the fluctuations. We hypothesised the following conditions: (i) among stable patients, the FEV1 value remains within two standard deviations of the mean μ, (ii) the variance of measurement fluctuations is stable in each patient, and (iii) acute severe respiratory events cause FEV1 to decline more than two standard deviations below the mean (i.e. FEV1 < μ − 2σ).

The abnormal threshold for potential respiratory events was set to measurements of FEV1 < μ − 2σ at least three times within a week. This was because the likelihood of a naturally low measurement of FEV1 < μ − 2σ is approximately 15% (7 C1 × 0.025 × 0.9756), and the likelihood of two such measurements within a week is approximately 1.2% (7 C2 × 0.0252 × 0.9755). Three such measurements within a week could be expected to occur in less than 1% of cases (7 C3 × 0.0253 × 0.9754).

Even for stable patients, pulmonary function can change over time post-transplant. Thus, determining the mean value was crucial. For this purpose, we defined a significant decline in FEV1 through the following steps: (a) the mean μ and variance σ² were calculated from the previous three months, excluding the past 14 days to mitigate the influence of recent acute decline in FEV1, and (b) an FEV1 decline was judged to be significant if the input was less than μ − 2σ.

Abnormal FEV1 declines were examined for each patient and recorded with the patient ID, date, and timing (morning/evening). 95 patients included in the study had a filling rate of at least 30%, but in the case of fewer than seven FEV1 inputs per week (i.e. missing data), the expected number required to signal an FEV1 decline was adjusted using the following formula:

$$\frac{Number\ of\ FEV1\ declines\ in\ a\ week}{Number\ of\ inputs\ of\ FEV1\ entered\ in\ a\ week}\times 7 \left(times/week\right)$$

This individualized approach was selected because home spirometry measurements exhibit high inter-individual variability, and a fixed decline threshold (e.g., 10%) may not provide consistent sensitivity across all patients. By using a deviation-based approach, we aimed to accommodate patient-specific trends and fluctuations in pulmonary function.

Study design to examine the algorithm’s accuracy

We conducted a nested case-crossover study by retrospectively reviewing an ongoing, prospective cohort of post-lung transplant patients. This study compared the occurrence of abnormal FEV1 declines detected by the algorithm in the week before each patient’s first hospitalisation for an acute respiratory event (case window) with alerts from the week 90 days before hospitalisation (control window). Additionally, we conducted a nested case-time-control study to account for potential time trends. For this, patients with acute respiratory events (case patients) were matched 1:3 based on post-operative follow-up duration and procedure type (single/bilateral lung transplant) to control patients without post-operative acute respiratory events. The relationship between the case and control windows for case and control patients is shown in Fig. 2. We did not calculate the sample size because the main objective of this study was rather exploratory.

Fig. 2

Overview of the analytical plan used in our study

Statistical analysis

The case-crossover analysis assessed the association between abnormal FEV1 declines and acute respiratory events. Exposure was defined as the presence or absence of abnormal FEV1 declines during specified time windows, with outcomes being acute events detected by established clinical tests. Conditional logistic regression quantified the association between exposure and outcomes, providing odds ratios and 95% confidence intervals. This approach accounts for within-subject correlation, offering a robust assessment of the temporal relationship between alerts and events. Analyses were conducted separately for morning and evening timings, with abnormal FEV1 declines recorded if the criteria were met once in either timing. Additionally, a nested case-time-control study controlled for time trends. In this analysis, patients with acute respiratory events (case patients) were matched on post-operative follow-up period and type of procedure (single/bilateral lung transplantation) at a ratio of 1:3 with control patients who had suffered no post-operative acute respiratory events. The odds ratios were calculated at the same timing as a matched case within the control patients and compared with those of cases to assess the time trend (Fig. 2). All statistical analyses were performed using R software (R v4.3.3; www.r-project.org). For the conditional logistic analysis, we used the clogit function from the survival package.

Results

Figure 3 shows a flowchart depicting the inclusion and exclusion criteria for this study. Of the 163 patients screened, 68 were excluded because their lung transplant had been performed at other hospitals (7 patients), they had suffered acute respiratory events within three months of starting LT-FollowUp (6 patients), their filling rate was less than 30% (52 patients), or they had failed to fill in data one week before the event (3 patients). Among the 95 included patients, 21 experienced acute respiratory events, including seven cases of acute rejection, seven cases of pneumonia, four cases of CLAD, two cases of effusion and one case of stent occlusion.

Fig. 3

Flowchart for the selection of study participants

The patients’ characteristics are described in Table 1. There is no significant difference in sex ratios between case patients (6/15) and control patients (34/40). The average age is comparable between case and control patients, 44.9 and 45.4, respectively. The most prevalent primary disease for both groups is idiopathic interstitial pneumonia, and brain-death donor is the most common donor type, with no difference in surgery type. Although the follow-up period does not differ significantly, acute respiratory events are more likely in case patients within the first post-operative year.

Table 1 Demographic data of the study participants

First, a case-crossover analysis using conditional logistic regression quantified the association between exposure (presence or absence of abnormal FEV1 decline during specified time windows) and outcome (acute respiratory events). The analysis showed an odds ratio of 5.42 × 10⁵ (Table 2), suggesting a strong association between acute respiratory events and FEV1 decline detected through the newly developed algorithm.

Table 2 Odds ratios and analysis in each group

Next, a nested case-time-control study accounted for potential time trends. In this analysis, 21 case patients were matched with 63 control patients based on post-operative follow-up period and procedure type (single or bilateral lung transplantation). Upon analysis of case and control windows using conditional logistic regression in control patients, an odds ratio of 1 indicated no trend over time (Table 2).

Discussion

Main results

This study evaluated an algorithm based on FEV1 decline from home spirometry and analysed the association between potential abnormalities detected by the algorithm and clinically assessed acute respiratory events. The comparison between case and control windows revealed a clear association between abnormal FEV1 declines detected by the algorithm and the occurrence of acute respiratory events. When case patients were matched with control patients regarding post-operative follow-up period and procedure type, the odds ratio obtained by conditional logistic regression was 1, indicating no apparent time trend during the study period.

Traditionally, repeated medical test measurements have been used to monitor patient changes during follow-up. In post-lung-transplantation patients, laboratory spirometry is periodically measured to detect CLAD, diagnosed as a decline of more than 20% in FEV1 from the average of the two best values taken at least three weeks apart. CLAD is a chronic allograft dysfunction that is considered irreversible, with no effective treatments [1]. However, acute pulmonary complications, such as graft rejection or infection, could trigger CLAD [1, 7], making early detection indispensable [8, 9].

As the CLAD statement recommends, a drop in FEV1 of more than 10% from the baseline necessitates further examination to detect potential acute respiratory events. However, periodic laboratory spirometry, conducted during regular outpatient visits, may not offer sufficiently early detection of acute FEV1 decline. Home spirometry addresses this gap by measuring FEV1 daily, and LT-FollowUp allows data in real time. Unlike laboratory spirometry conducted by trained medical personnel, however, home spirometry is measured by the patients themselves, and its reliability can vary from patient to patient. To mitigate this variation, a new algorithm was developed to detect FEV1 decline by standardising the mean and standard deviation measurements and simultaneously returning an alert. By using statistical estimates instead of absolute values, uniform comparisons can be made among patients with different degrees of measurement variability.

In evaluating the accuracy of our algorithm, it is important to assess the risk of bias and applicability. The Quality Assessment of Diagnostic Accuracy Studies tool was proposed in 2003 and updated in 2011 [10]. This tool suggests evaluating study quality in four domains: patient selection, index tests, reference standards, and flow and timing. Each domain contains several questions regarding the degree of bias risk and applicability. This procedure enables a straightforward comparison of the accuracy of newly proposed tests against the gold standard. In our study, several points are incompatible with this evaluation process. STARD 2015 recommends at least one measure of accuracy, such as sensitivity, specificity, or predictive values, for test tool comparison [6]. Because there is no gold standard test for detecting FEV1 decline in home spirometry, it is not possible to calculate the sensitivity and specificity. Additionally, our algorithm is updated daily, meaning the test is performed daily to check FEV1 decline, which complicates the simple comparison of accuracy.

To address these issues, we employed case-crossover and case-time-control designs to evaluate the association between alerts and acute respiratory events and compare them between case and control time windows for the same patients as self-controls. During our study, only one case had a false positive alert, leading to statistical calculation convergence issues in conditional logistic regression. However, alerts were present in 11 of 21 patients, and the association between alerts and acute respiratory events was clear using conditional logistic regression. Our exploratory results are encouraging for the early detection of acute respiratory events using home spirometry. We believe this new alert system could contribute to earlier detection of post-operative acute respiratory events.

Furthermore, the physician's response to a decline in FEV1 detected by the LT-FollowUp system is critical. When the algorithm detects an abnormal drop in FEV1, an alert is issued to both the patient and the healthcare professional. Upon receiving this alert, the patient is first assessed for any new symptoms, such as dyspnoea, cough or fever. If no symptoms are present, the patient is instructed to repeat home spirometry and if an abnormal decline is identified, the patient is instructed to contact our hospital for further assessment. If the patient does not contact our hospital, the doctor contacts the patient to check on the patient's condition. As a result of the medical assessment, the physician decides whether additional diagnostic tests, such as bronchoscopy, imaging or blood tests, are necessary. If the patient is deemed to require treatment, early therapeutic intervention is initiated.

Limitations and future work

There are several limitations to this study. First, our assumption that the mean of the measured FEV1 values changes little in stable patients does not always hold. Pulmonary function improves gradually in the first year post-transplantation [11], and the mean value during this period may not reliably detect any decline in FEV1. To address this issue, we must investigate how long pulmonary function improves post-operatively before reaching a plateau. This study only assessed the first event per patient for the above reason. Once we can implement this trend line in statistical calculations, the algorithm could be applied beyond the first event. We are already addressing these issues and planning further validation studies.

Second, although system adherence was good in terms of filling rates compared with other studies [12], it was not 100%. As a result, a criterion based solely on the number of FEV1 declines per week may miss patient abnormalities. We used an adjustment equation to account for weekly input rates in this study, but previous studies reported much lower filling rates. In addition, adherence to remote lung function monitoring tends to decrease over time [5, 13]. Patient education is therefore essential to improve adherence and ensure effective algorithm use and future studies should investigate interventions such as personalized reminders and telehealth support to enhance compliance with home spirometry monitoring.

Third, the number of participants included in this study was small, and the study period was not long. Japan's Organ Transplant Bill came into force in 1997, but they did not reach 100 cases annually until 2022 [14], limiting the possibility of single-centre studies. The relatively small number of acute respiratory events (n = 21) may limit the statistical power and generalizability of our findings, but our results may be broadly applicable to lung transplant populations. To increase the sample size and ensure long-term observation, we have already conducted a multi-centre study involving all Japanese lung transplantation institutions. This expanded study will allow validation of the algorithm in a larger and more diverse cohort (e.g. procedure, disease, post-operative period), thereby addressing concerns about external validity and enhancing clinical applicability.

Fourth, the LT-FollowUp system includes real-time feedback mechanisms, the absence of direct supervision may have introduced variability in measurement technique. To mitigate this issue, we employed a Z-score-based statistical estimate derived from standard error, rather than relying on absolute FEV1 thresholds, to reduce the impact of individual measurement variability. Future studies should evaluate the impact of structured in-person or remote training sessions on home spirometry accuracy and patient adherence.

Despite these limitations, we have demonstrated that the proposed algorithm can detect acute respiratory events. Combined with LT-FollowUp, this algorithm leverages recent medical data to improve transplantation patient prognosis by detecting acute events as early as possible.

Conclusion

This study introduced a novel algorithm for detecting declines in FEV1 using home spirometry data from LT-FollowUp. Through a nested case-crossover study, we established an association between our alert system and the occurrence of acute respiratory events in post-lung-transplantation patients. This real-time monitoring approach shows promise for the early detection of acute respiratory events, potentially enabling timely interventions which would prevent progression to CLAD. Our findings are encouraging, although further validation in larger, multi-centre studies is necessary to enhance the algorithm’s clinical utility and reliability.