The impact of obstructive sleep apnea treatment on microvascular complications in patients with type 2 diabetes: a feasibility randomized controlled trial

Abstract

Study Objectives:

Obstructive sleep apnea (OSA) is associated with an increased risk of diabetes-related complications. Hence, it is plausible that continuous positive airway pressure (CPAP) could have a favorable impact on these complications. We assessed the feasibility of conducting a randomized control trial in patients with type 2 diabetes and OSA over 2 years.

Methods:

We conducted an open-label multicenter feasibility randomized control trial of CPAP vs no CPAP in patients with type 2 diabetes and OSA. Patients with resting oxygen saturation < 90%, central apnea index > 15 events/h, or Epworth Sleepiness Scale ≥ 11 were excluded. OSA was diagnosed using a multichannel portable device (ApneaLink Air, ResMed). The primary outcome measures were related to feasibility and the secondary outcomes were changes in various clinical and biochemical parameters related to diabetes outcomes.

Results:

Eighty-three (40 CPAP vs 43 no CPAP) patients were randomly assigned, with a median (interquartile range) follow-up of 645 (545, 861) days. CPAP compliance was inadequate, with a median usage of approximately 3.5 hours/night. Early CPAP use predicted longer-term compliance. The adjusted analysis showed a possible favorable association between being randomly assigned to CPAP and several diabetes-related end points (chronic kidney disease, neuropathy, and quality of life).

Conclusions:

It was feasible to recruit, randomly assign, and achieve a high follow-up rate over 2 years in patients with OSA and type 2 diabetes. CPAP compliance might improve by a run-in period before randomization. A full randomized control trial is necessary to assess the observed favorable association between CPAP and chronic kidney disease , neuropathy, and quality of life in patients with type 2 diabetes.

Clinical Trial Registration: Registry: ISRCTN; Name: The impact of sleep disorders in patients with type 2 diabetes; URL: https://www.isrctn.com/ISRCTN12361838; Identifier: ISRCTN12361838.

Citation:

Makhdom EA, Maher A, Ottridge R, et al. The impact of obstructive sleep apnea treatment on microvascular complications in patients with type 2 diabetes: a feasibility randomized controlled trial. J Clin Sleep Med. 2024;20(6):947–957.

BRIEF SUMMARY

Current Knowledge/Study Rationale: Obstructive sleep apnea and type 2 diabetes have a bidirectional relationship; patients with obstructive sleep apnea are at high risk of developing type 2 diabetes, and vice versa. Obstructive sleep apnea is linked to an increased risk of advanced retinopathy and renal function decline. Several randomized controlled trials of continuous positive airway pressure (CPAP), the gold-standard treatment of obstructive sleep apnea, in patients with type 2 diabetes did not assess the CPAP’s impact on diabetes-related hard end points due to their focus on glycemic control and were relatively short.

Study Impact: It is feasible to recruit and randomize over a period of 2 years. CPAP compliance may improve by the run-in period before randomization. Conducting a complete randomized controlled trial is necessary to assess the observed favorable relation between CPAP and complications of type 2 diabetes. CPAP has potential benefits in reducing diabetes-related complications.

INTRODUCTION

Obstructive sleep apnea (OSA) is characterized by frequent partial or complete obstructions of the upper airway during sleep, resulting in cyclical episodes of hypoxemia, sleep fragmentation, changes in heart rate and blood pressure, and increased intrathoracic pressure.¹ These episodes lead to various pathophysiological consequences, such as intermittent hypoxia, sympathetic activation, systemic inflammation, oxidative stress, and changes in the endocrine system.² OSA has consequently been linked to hypertension, cardiovascular disease, hyperlipidemia, type 2 diabetes (T2D), road traffic accidents, and impaired quality of life.³ OSA is very common, affecting 12.5–83.8% of men and 3.7–70.8% of women, depending on the country and study population; approximately 936 million individuals aged 30–90 years have OSA with apnea-hypopnea index (AHI) $\ge$ 5 events/h worldwide.⁴

T2D is very common and its global prevalence is increasing.⁵ The World Health Organization estimated that the total number of individuals living with diabetes had increased to 537 million adults in 2021.⁶ The burden of T2D on patients, the health care system, and wider society is high, with an increased risk of cardiovascular disease and microvascular complications such as neuropathy, nephropathy, retinopathy, and foot ulcer.^7,8 Despite improvements in the management of T2D and the attainment of treatment targets, the burden of diabetes complications remains significant.^9,10 Therefore, there is a need to identify new, innovative treatment strategies to reduce the impact of T2D further.

The close link between OSA and T2D is unsurprising, considering that obesity and advanced age are common risk factors.^5,11 We have previously demonstrated that patients with OSA are at increased risk of developing T2D, and vice versa.¹² We have also shown that patients with T2D and OSA are at increased risk of developing cardiovascular disease and microvascular complications compared with patients with T2D without OSA¹³ and that in patients with T2D OSA is associated with an increased risk of developing advanced retinopathy and renal function decline.^14,15 This is likely because both OSA and T2D display a similar impact on vascular disease risk factors, including inflammation, oxidative stress, and endothelial dysfunction, driving the development of these complications.¹⁶ Hence, it is plausible that treating OSA in patients with T2D could potentially reduce the burden of diabetes-related complications.

Continuous positive airway pressure (CPAP) is the gold standard for OSA treatment; however, compliance is challenging.¹⁷ Several randomized control trials of CPAP in patients with T2D have been conducted. These trials primarily focused on glycemic control and were relatively short. Therefore, they could not evaluate the impact of CPAP on other important diabetes-related hard end points.^18,19 Our overall aim is to conduct a randomized controlled trial (RCT) assessing the impact of CPAP on microvascular complications in patients with T2D. Hence, we first conducted an RCT of CPAP vs no CPAP in people with T2D with OSA to determine the feasibility of this approach and assist in designing a future definitive trial.

METHODS

The protocol of this feasibility RCT has been published previously.²⁰ Here, we provide an overall summary.

Objectives

The primary objectives are consistent with the feasibility design and related to the feasibility aspects of the RCT. These include the following:

1. Assess the willingness of participants to be randomly assigned.

2. Assess the willingness of clinicians to recruit participants.

3. Assess follow-up rates and adherence/compliance rates.

4. Provide data to inform the sample size for a substantive trial.

5. Optimize the choice of outcome measures for a substantive trial.

The secondary objectives focused on exploring CPAP’s impact on microvascular complications and clinical and biochemical parameters in patients with T2D (refer to the supplemental material).

Study design

We conducted a multicenter feasibility RCT in England. The RCT was embedded in an observational study examining the associations between sleep disorders and T2D-related outcomes over 2 years. The observational study recruited patients with T2D regardless of their OSA status but included testing for OSA. In this article, we report on the RCT.

The RCT was an open-label, randomized, controlled, parallel-arm clinical trial of patients with T2D and OSA. Participants were randomly assigned in a 1:1 ratio to CPAP or no CPAP in addition to receiving routine care for 2 years, with assessments performed at baseline and study end.²⁰ All patients were contacted via phone every 6 months till the study concluded. Patients randomly assigned to CPAP received extra contact at 2 and 4 weeks and further ad hoc contact. The follow-up duration and assessments were affected by the coronavirus disease 2019 pandemic (see the supplemental material).

This RCT was carried out in accordance with the Research Governance Framework for Health and Social Care, the relevant UK Statutory Instruments (including the Data Protection Act 2018 and the Human Tissue Act 2008), and Good Clinical Practice principles.²⁰ The study was registered in the ISRCTN registry (https://www.isrctn.com/ISRCTN12361838, Registered April 4, 2018, Protocol version: v5.0 02.12.19). The study was approved by National Research Ethics Committee West Midlands – The Black Country, reference 18/WM/0070. All study participants consented through a 2-stage process, one for the observational study and the other for the feasibility of RCT.

Settings and participants

Participants were recruited from diabetes clinics in 13 different National Health Service Trusts throughout England (Appendix S1 ^{(372.4KB, pdf)} in the supplemental material). Recruitment occurred between July 2018 and February 2020 and stopped at this date due to the coronavirus disease 2019 pandemic. Participants for the feasibility RCT were recruited from the observational study. A full list of the inclusion/exclusion criteria has been published previously and can be found in Table 1. They were selected to ensure patient safety while allowing ease of recruitment and kept relatively liberal to aid recruitment. Patients with evidence of excessive daytime sleepiness were excluded from randomization because they would have received CPAP treatment in real life. Not receiving a CPAP would affect their driving license.

Table 1.

Summary of inclusion and exclusion criteria.

Inclusion Criteria	Exclusion Criteria
Observational Cohort Study
Are ≥ 18 years old Diagnosis of T2D eGFR ≥ 15 ml/min/1.73 m2 in the last 12 months	History of T1D Known OSA Active malignancy CKD from reasons other than diabetes Receiving chemotherapy, immunosuppressant drugs, or home oxygen treatment History of recurrent hospital admissions due to infective exacerbation of a respiratory condition Received contrast imaging within the last 2 months Pregnancy Intending to undergo bariatric surgery during the study duration Unable to comply with the study protocol Unable to give informed consent Professional drivers, operators of heavy machinery, and/or working at high altitude History of falling asleep while driving within last 2 years
Feasibility RCT
Adults who were willing to be randomly assigned to CPAP vs no CPAP Has Epworth Sleepiness Scale < 11 Has an apnea-hypopnea index ≥ 10 events/h	Participants who have resting oxygen saturation < 90% or Participants who have episodes of central sleep apnea at > 15/h

CKD = chronic kidney disease, CPAP = continuous positive airway pressure, eGFR = estimated glomerular filtration rate, OSA = obstructive sleep apnea, RCT = randomized controlled trial, T1D = type 1 diabetes, T2D = type 2 diabetes.

OSA diagnosis

OSA was assessed based on a single overnight home-based polygraphy using a portable device (ApneaLink Air; ResMed UK Ltd., Didcot, Oxfordshire). The ApneaLink Air device comes with the AirView diagnostics cloud-based system (AV). Sleep studies were scored based on the American Academy of Sleep Medicine guidelines. Hypopnea was defined as $\ge$ 4% oxygen desaturation with $\ge$ 30% reduction in nasal airflow signal.²¹ Apnea was defined as a ≥ 90% reduction in airflow for a period of ≥ 10 seconds.²¹ An apnea-hypopnea index (AHI) ≥ of 5 events/h was consistent with the diagnosis of OSA.²² The severity of OSA was assessed based on the AHI and oxygen desaturation index based on 4% oxygen desaturation. OSA was classified as mild, moderate, or severe based on AHI ≥ 5 but < 15, ≥ 15 but < 30, and ≥ 30 events/h, respectively.²³

Randomization

A minimization algorithm within the computerized randomization system was used to ensure balance in the treatment allocation over the following variables: ethnicity (White Europeans, others), sex (female or male), and severity of OSA (AHI < 15 or ≥ 15 events/h).

CPAP initiation and compliance

In the CPAP arm, each participant was provided with a CPAP machine (ResMed Airsense 10 Autoset). Compliance was monitored remotely via a secured website, AirView (AutoRamp, ResMed); AV is compliant with the national and international security policies and practices, including NHS IG Toolkit: Ref: 8J317 (2015-2016 – Score 90%), ISO 27001 Certificate (IDS Host), and the Data Protection Act 2018 and the EU General Data Protection Regulation 2018. All patients were given an appropriate mask, connecting hose, a heated humidifier, and all the necessary accessories to operate the CPAP equipment for the trial.

The CPAP device was delivered and initiated by the sleep physiologist (M.N.) at the patient’s convenient place. The sleep physiologist discussed the OSA diagnosis and CPAP benefits, explained technical aspects and general usage of the device, provided information and potential solutions to the expected side effects of CPAP, and contacted the patients by telephone 2–3 times during the first week of CPAP for any troubleshooting. Patients were encouraged to contact the sleep physiologist for troubleshooting at any time and were contacted when compliance dropped.

Sleep data were monitored once to twice weekly. Adequate compliance in this trial was defined as an average use of CPAP $>$ 4 hours/night on 70% of nights.²⁰

Outcomes measures

The primary outcomes aimed to assess the feasibility of running a substantive RCT using the following criteria:

1. Recruiting the proposed sample size within the planned time frames.

2. Meeting the proposed time frames regarding interpreting the sleep assessments and initiating patients on treatment (within 8 weeks from registration or 2 weeks from randomization).

3. Achieving a follow-up rate ≥ 80% for randomly assigned patients.

4. Achieving a CPAP usage ≥ 4 hours/night on ≥ 70% of nights in ≥ 80% of patients randomly assigned to CPAP treatment.

5. Generating a mean and standard deviation regarding the predicted response to the intervention to allow future sample size calculations.

Secondary outcomes aimed to explore the impact of CPAP on diabetes-related end points (refer to the supplemental material). All secondary outcomes were measured at baseline and the 2-year follow-up visit.

Statistical analysis

Because this was a feasibility study, no formal sample size calculations were undertaken. However, around 500 patients who fulfill the main inclusion criteria were expected to be enrolled and screened for OSA, with approximately 140 of them randomly assigned for the RCT based on OSA prevalence and severity that we reported in the previous study.²⁴ An a priori statistical analysis plan was agreed upon to provide point estimates and 95% confidence intervals for 2-sided tests for all outcome measures. All outcomes were analyzed using the intention-to-treat method, where participants were analyzed in the groups to which they were randomly assigned, regardless of protocol noncompliance. Outcomes were measured at the 2-year follow-up time point and were adjusted for minimization variables and baseline value as fixed effects. If participants were missing follow-up values, they were not included in the analysis, but their baseline values were still included in the summary statistics. A log-binomial model with a log link was used for binary outcomes to generate adjusted relative risks and 95% confidence intervals at the follow-up time point. Continuous outcomes were analyzed using a linear regression model to generate mean differences and 95% confidence intervals. Appropriate summary statistics are reported for each outcome (eg, proportions [percent], mean [standard deviation], median [interquartile range]). Sensitivity analyses were limited to estimated glomerular filtration rate and consisted of a per-protocol analysis where those who were nonadherent to their allocated intervention were excluded. No subgroup analyses were planned for this feasibility study.

The impact of the coronavirus disease 2019 pandemic

The impact of the coronavirus disease 2019 pandemic is discussed in the supplemental material.

RESULTS

Of 229 patients recruited to the observational cohort study, 83 participants fulfilled the eligibility criteria for the RCT. They were randomly assigned to CPAP (n = 40) or no CPAP (n = 43) in addition to the continuation of routine care over a period of 2 years. The median (interquartile range) follow-up period was 645 (545, 861) days. A total of 12 patients withdrew from the CPAP arm and none withdrew from the no-CPAP arm. Out of the 12, 9 withdrew from treatment but agreed to continue with follow-up data collection, 2 withdrew from treatment and follow-up, and 1 provided no withdrawal information. Only the participants who withdrew from treatment and follow-up were included as withdrawals in Figure 1, because follow-up was still expected from other withdrawals. There was 1 protocol deviation reported in the no-CPAP arm: a patient with an Epworth Sleepiness Scale score ≥ 11 was randomly assigned in error.

Figure 1. Flow diagram of the study.

The 24-month follow-up visit represents the study-end visit and may have occurred before 24 months in some participants.

Baseline characteristics (Table 2)

Table 2.

Patient characteristics at baseline as total and categorized based on randomization to CPAP.

Baseline Characteristics	CPAP (n = 40)	No CPAP (n = 43)	Total (n = 83)
Age (years)	60.5 (11.9), 40	64.5 (9.5), 42	62.5 (10.9),82
Sex (male, %)	25 (62.5%)	34 (79.1%)	59 (71.1%)
Ethnicity (White, %)	35 (87.5%)	39 (90.8%)	74 (89.1%)
Diabetes duration (years)	10.6 (7.2), 39	13.7 (8.4), 43	12.2 (7.9), 82
Never smoked, n (%)	18 (45.0%)	16 (37.2%)	34 (41.0%)
BMI (kg/m2)	34.4 (6.5), 37	36.5 (7.7), 39	35.4 (7.2), 76
BMI classes, n (%)
< 25 kg/m2	2 (5.4%)	2 (5.1%)	4 (5.3%)
25 to < 30 kg/m2	9 (24.3%)	4 (10.3%)	13 (17.1%)
30 to < 35 kg/m2	10 (27.0%)	13 (33.3%)	23 (30.3%)
≥ 35 kg/m2	16 (43.2%)	20 (51.3%)	36 (47.4%)
Blood pressure (BP)
Systolic BP (mmHg)	130.9 (13.5), 39	137.9 (14.4), 42	134.5 (14.3), 81
Diastolic BP (mmHg)	77.1 (8.5), 38	77.7 (10.7), 42	77.4 (9.7), 80
Diabetes treatments
Oral glucose-lowering agents	36 (92.3%)	37 (90.2%)	73 (91.3%)
Insulin use	15 (37.5%)	25 (58.1%)	40 (48.2%)
GLP-1 use	7 (18.4%)	5 (12.8%)	12 (15.6%)
Lipid-lowering agents, n (%)	30 (75.0%)	31 (73.8%)	61 (74.4%)
Antihypertensives, n (%)	13 (33.3%)	21 (52.5%)	34 (43.0%)
Comorbidities
Stroke	0 (-)	2 (5.6%)	2 (3.0%)
Myocardial infraction	3 (9.7%)	8 (21.6%)	11 (16.2%)
AHI, median (IQR)	17.7 (13.7, 32.4), 40	24.4 (19.2, 41.1), 38	21.4 (14.9, 36.6)
Epworth Sleepiness Scale score	6.0 [3.0, 7.5], 40	6.0 [4.0, 8.0], 42	6.0 [3.0, 8.0], 82

Data are presented as % (n), median [IQR], n or mean (standard deviation), n. In the CPAP arm there were 3 missing BMI, 1 missing systolic BP, 2 missing diastolic BP, and 1 missing diabetes duration value. In the no-CPAP arm there were 4 missing BMI, 1 missing systolic BP, 1 missing diastolic BP, and 1 missing age value. BMI = body mass index, CPAP = continuous positive airway pressure, GLP-1 = glucagon-like peptide 1, IQR = interquartile range.

The study population was mostly middle-aged men of White European ethnicity. Most of the study population had obesity (77.7%). The use of blood pressure–lowering and lipid-lowering medications and insulin was high (43%, 74.4%, and 48.2%, respectively). A history of stroke was present in 3% of the study population (0% CPAP vs 5.6% [n = 2] no CPAP), and 16.2% had myocardial infarction (9.7% [n = 3] in CPAP vs 21.6% [n = 8] in no CPAP). The baseline medications in Table 2 show a high prevalence of antihypertensive medications in both CPAP and no-CPAP arms (76.3% vs 88.1%). The prevalence of AHI ≥ 15 events/h was 79.1% (n = 34) in the no-CPAP arm and 67.5% (n = 27) in the CPAP arm, and 10 $\le$ AHI $<$ 15 events/h prevalence was 32.5% (n = 13) in the CPAP arm vs 20.9% (n = 9) in the no-CPAP arm.

Feasibility outcomes (Table 3)

Table 3.

Summary of achievement of feasibility outcomes.

	CPAP (n = 40)	No CPAP (n = 43)	Total (n = 83)
Recruit the proposed sample size within the planned time frames
Proportion recruited of target	57.1%	61.4%	59.2%
Meet the proposed timeframes regarding interpreting the sleep assessments and initiating patients on treatment
CPAP initiations completed within 8 weeks of registration or 2 weeks of randomization	12/32a (37.5%)
Achieve a follow-up rate ≥ 80% for randomized patientsb
Follow-up rate	33/38c (86.8%)	33/43 (76.7%)	66/81c (81.5%)
Achieve a CPAP usage ≥ 4 hours/night on ≥ 70% of nights in ≥ 80% patients randomized to CPAP treatment
CPAP usage ≥ 4 hours/night on ≥ 70% of nights	8/40 (20.0%)
Generate a mean and standard deviation regarding predicted response to the intervention:
Baseline eGFR: Mean (standard deviation, n)	86.1 (26.4, 40)	74.9 (21.9, 41)	79.9 (24.8, 81)
Follow-up eGFR: Mean (standard deviation, n)	93.4 (30.9, 21)	75.5 (31.4, 28)	83.2 (32.1, 49)

^aEight of 40 participants did not start the CPAP after randomization: 5 withdrew before being issued a device, and 3 had no CPAP initiation date. ^bParticipants are considered to have follow-up data if they have attended their 24-month follow-up visit or have follow-up CPAP data recorded. ^cTwo participants withdrew from the follow-up data being collected. CPAP = continuous positive airway pressure, eGFR = estimated glomerular filtration rate.

Three out of the 5 feasibility criteria were not met. Only about 60% of the study population was recruited within the planned time frames. The planned recruitment period was from April 2018–October 2019, with a target of opening 10 sites by September 2018. The first site was opened in July 2018, and it took until January 2019 to reach 10 sites, with the first patient recruited in August 2018. The delay in obtaining research and development approvals from various local sites significantly contributed to the overall delay in recruitment. Further delays were caused by medical engineering taking longer than expected to sign-off equipment at various sites. It took 213 days, on average, from the first formal approach to a site to the first recruitment. In addition, the goal was to recruit 40 registered patients a month from all the research centers. The actual recruitment did not exceed 25 per month in specific centers, resulting in an increase in participating centers from 10 to 13. Our target of initiating CPAP within 8 weeks of registration or 2 weeks of randomization was achieved in 37.5% of participants due to multiple factors, including delays in the supplies of the CPAP equipment and a lack of patients’ availability within the planned time.

Only 8 out of 40 patients achieved our criteria for CPAP compliance. As detailed above, out of the 40 patients, 12 patients withdrew from CPAP treatment, and a further 2 patients did not use the CPAP. When examined, the remaining patients’ median CPAP usage per night was more than 3.5 hours (Table 4). This suggests significant CPAP usage in a proportion of the study population despite failing the strict compliance criteria of 4 hours on 70% of the nights. There are likely several factors contributing to the high number of withdrawals and CPAP nonusers (see Discussion). Our data show that the withdrawals happened in the first 10 months. Having a run-in period before randomization might have identified those individuals unlikely to be able to adhere to CPAP. In our trial, we found that CPAP compliance at 2 weeks predicted CPAP compliance at 1 year using an adherence threshold of 70% and 50% [χ² (1, n = 25) = 4.1, P = .043) and χ² (1, n = 25) = 17.6, P < .01, respectively] in Table S1 ^{(372.4KB, pdf)} in the supplemental material. The high follow-up rate was achieved likely due to the flexibility in collecting data remotely, including the CPAP usage.

Table 4.

Summary of CPAP usage at multiple time points during the study in patients who used CPAP.

Time Point	2 Weeks (n = 25)a	4 Weeks (n = 26)	1 Year (n = 26)	2 Years (n = 5)
No. of days used	11.0 [5.0, 13.0]	20.0 [7.0, 27.0]	260.0 [11.0, 346.0]	592.0 [360.0, 653.0]
Days ≥ 4 hours per night	7.0 [3.0, 10.0]	14.0 [1.0, 22.0]	169.5 [3.0, 267.0]	505 [257.0, 559.0]
Average time used per night (hours:minutes)	3:49 [1:29, 4:30]	3:32 [0:49, 4:45]	3:40 [0.06, 4:45]	4:58 [4:48, 10:13]

Data are presented as median [interquartile range]. ^aOf the 40 participants allocated to CPAP, 12 participants withdrew from using CPAP and 2 did not use CPAP. One participant is missing 2-week data only. Five participants had 2-year CPAP data available. CPAP = continuous positive airway pressure.

Secondary outcomes

CPAP and clinical and biochemical variables related to T2D

The impact of CPAP on various clinical and biochemical variables related to diabetes and obesity is summarized in Table S2 ^{(372.4KB, pdf)} . Analysis was adjusted for age, sex, ethnicity, OSA severity, and baseline value of the outcome of interest. All data available at both time points are reported, but only participants who had both baseline and follow-up values were included in each analysis. Blood pressure and adiposity variables were more favorable in patients randomly assigned to CPAP. Overall, the analysis showed no significant association between being randomly assigned to CPAP and these clinical and metabolic parameters.

CPAP and renal outcomes

Patients who were randomly assigned to CPAP had higher estimated glomerular filtration rates and lower urinary albumin/creatinine ratio at baseline. By the study’s end, estimated glomerular filtration rate and urinary albumin/creatinine ratio numerically improved in the CPAP arm but not in the no-CPAP arm (Table 5).

Table 5.

Summary of the relationship between CPAP and renal outcomes.

	Baseline		Study End		Adjusted Mean Difference (95% CI) or Relative Risk (95% CI)
	CPAP (n)	No CPAP (n)	CPAP (n)	No CPAP (n)
eGFR (ml/min/1.73 m2)	86.1 (26.4, 40)	73.9 (21.9, 41)	93.4 (30.9, 21)	75.5 (31.4, 28)	4.3a (−12.0, 20.6)
eGFR < 60 (ml/min/1.73 m2)	7 (17.5%)	12 (29.3%)	4 (19.1%)	9 (31.0%)	1.4a (0.2, 12.3)
Urinary ACR (mg/mmol)	11.4 (26.2, 33)	41.6 (83.3, 38)	2.0 (2.8, 18)	38.3 (96.5, 22)	−21.1b (−56.4, 14.2)
Albuminuria status
Abnormal, n (%)	8 (24.2%)	15 (39.5%)	2 (11.1%)	9 (38.1%)	0.33b (0.03, 3.9)
Microalbuminuria	6	5	2	4
Macroalbuminuria	2	10	0	5

Data are presented as mean (SD, n) or N (%). CPAP arm: eGFR 2 values missing at the end-study visit; albuminuria 7 values missing at baseline and 5 missing at study end. No-CPAP arm: eGFR 2 values missing at baseline and 4 missing at the study-end visit. Albuminuria 5 values were missing at baseline and 11 were missing at the study’s end. ^aAdjusted for ethnicity, sex, obstructive sleep apnea severity, age, systolic blood pressure, diastolic blood pressure, hemoglobin A1c (mmol/mol), and baseline value. Only participants with both baseline and follow-up values are included in this analysis. ^bAdjusted for ethnicity, sex, obstructive sleep apnea severity, age, systolic blood pressure, and baseline value. Only participants with both baseline and follow-up values are included in the analysis. A relative risk < 1 or a mean difference < 0 favors CPAP. Analysis was performed using mean difference and relative risk as appropriate. A relative risk < 1 favors CPAP. The mean difference was calculated as CPAP − no CPAP. ACR = urinary albumin/creatinine ratio, CI = confidence interval, CPAP = continuous positive airway pressure, eGFR = estimated glomerular filtration rate.

CPAP and quality of life

The CPAP group had a higher physical component score at baseline than the no-CPAP group. By the study’s end, this score was maintained in the CPAP arm and worsened in the no-CPAP arm.

The mental component scores at baseline and study end were similar in patients randomly assigned to CPAP or no CPAP and improved in both arms (Table S4 ^{(372.4KB, pdf)} ).

CPAP and retinopathy

It was impossible to analyze this outcome due to the small number of patients for whom we managed to obtain retinal screening results, especially for the follow-up time point. Data are summarized in Table S5 ^{(372.4KB, pdf)} .

CPAP and peripheral neuropathy

The Michigan Neuropathy Screening Instrument (MNSI):

An MNSI score on the examination component (MNSIe) > 2 or the questionnaire component (MNSIq) ≥ 7 was considered consistent with the diabetic peripheral neuropathy.^25,26 Being randomly assigned to CPAP was associated with a possible numerically favorable impact on peripheral neuropathy compared to no CPAP, whether defined based on MNSIe, MNSIq, or total score (Table 6). The individual components of the MNSIe are reported in Table S6 ^{(372.4KB, pdf)} .

Table 6.

Michigan Neuropathy Screening Instrument (MNSI) scores.

	Baseline		Follow-Up		Adjusted Relative Riska (95% CI)
	CPAP, n (%)	No CPAP, n (%)	CPAP, n (%)	No CPAP, n (%)
MNSI questionnaire
Abnormal	5 (12.5%)	12 (27.9%)	0 (-)	7 (26.9%)	0.78 (0.4, 1.6)
Normal	35 (87.5%)	31 (72.1%)	18 (100.0%)	19 (73.1%)
Missing	0	0	6	9
MNSI examination
Abnormal	18 (47.4%)	20 (50.0%)	7 (36.8%)	7 (50.0%)	0.83 (0.3, 2.7)
Normal	20 (52.6%)	20 (50.0%)	12 (63.2%)	7 (50.0%)
Missing	2	3	5	21
MNSI overall score
Abnormal	19 (50.0%)	22 (55.0%)	7 (41.2%)	12 (80.0%)	0.51 (0.1, 2.3)
Normal	19 (50.0%)	18 (45.0%)	10 (58.8%)	3 (20.0%)
Missing	2	3	7	20

The MNSI questionnaire has a range from 0–13, where 0 is the best and 13 is the worst. The MNSI examination has a range from 0–8, where 0 is best and 8 is worst. For the MNSI questionnaire score, “normal” is categorized as a score < 7 and “abnormal” is categorized as a score ≥ 7. For the MNSI examination score, “normal” is categorized as a score ≤ 2 and “abnormal” is categorized as a score > 2. MNSI’s overall score is “abnormal” if the participant has an “abnormal” score on the questionnaire and/or the examination and “normal” otherwise. ^aAdjusted for ethnicity, sex, obstructive sleep apnea severity, age, and baseline value. A relative risk < 1 favors CPAP. Only participants with both baseline and follow-up values are included in this analysis. CI = confidence interval, CPAP = continuous positive airway pressure.

There was no evidence of a difference between CPAP and no CPAP except in regard to the vibration perception at the great toe.

Neuropad and biothesiometry:

The Neuropad is based on a cobalt II compound that turns pink after being exposed to dermal foot perspiration for 10 minutes in the plantar foot areas of the foot. The change in the color is recorded as none, partial, or complete.²⁷ The Neuropad assessment improved in both trial arms during follow-up but being randomly assigned to CPAP was associated with a numerically lower proportion of abnormal Neuropad testing.

A biothesiometer was used to test the vibration perception on the great toe of each foot; an average of 3 measurements were considered.²⁸ Vibration perception was numerically better in association with being randomly assigned to CPAP vs no CPAP. The CPAP arm had a lower proportion of patients with abnormal Neuropad testing or abnormal vibration perception at baseline compared to no CPAP (Table 7).

Table 7.

Neuropod, vibration, and 10-gram monofilament.

	Baseline		Follow-Up		Adjusted Relative Riska (95% CI)
	CPAP, n (%)	No CPAP, n (%)	CPAP, n (%)	No CPAP, n (%)
Neuropod/both feetb
Incomplete/abnormal	24 (63.2%)	31 (73.8%)	12 (57.1%)	11 (68.8%)	0.87 (0.3, 2.3)
Partial	17	24	10	10
None	7	7	2	1
Complete	14 (36.8%)	11 (26.2%)	9 (42.9%)	5 (31.3%)
Missing	2	1	2	17
Vibration perception/average both feetc
Abnormal (> 25V)	8 (33.3%)	17 (60.7%)	4 (36.4%)	3 (60.0%)	0.64 (0.04, 9.2)
Normal (≤ 25V)	16 (66.7%)	11 (39.3%)	7 (63.6%)	2 (40.0%)
Missing	16	15	12	28
10-g monofilament/both feetd
Abnormal	15 (38.5%)	22 (52.4%)	6 (28.6%)	15 (88.2%)	0.49 (0.1, 1.5)
Reduced	12	13	4	10
Absent	3	9	2	5
Normal	24 (61.5%)	20 (47.6%)	15 (71.4%)	2 (11.8%)
Missing	1	1	2	16

^aAdjusted for ethnicity, sex, obstructive sleep apnea severity, age, and baseline value. A relative risk < 1 favors CPAP. Analysis is “incomplete” [“partial” + “none”] vs “complete.” Only participants with both baseline and follow-up values are included in this analysis. ^bIf either foot is scored as “ none,” ” then “both feet” is categorized as “none”; if either foot is categorized as “partial,” but neither as “none,” then “both feet” is categorized as “partial”; if both feet are scored as “complete,” then “both feet” is categorized as “complete.” ^cIf either foot is high, then “both feet” are high. ^dIf either foot is scored as “absent,” then “both feet” is categorized as “absent”; if either foot is categorized as “reduced,” but neither as “absent,” then “both feet” is categorized as “reduced”; if both feet are scored as “normal,” then “both feet” is categorized as “normal.” CI = confidence interval, CPAP = continuous positive airway pressure.

Monofilament:

Foot insensitivity to a 10-g monofilament (applied to 10 positions) was defined as a score of 8–10 being normal and a response of 0–7 abnormal.²⁶ An abnormal monofilament test was less common in the CPAP arm at baseline vs no CPAP. During follow-up, there was an improvement in the monofilament testing in the CPAP arm and a worsening in the no-CPAP arm (Table 7).

DISCUSSION

This article reports the longest CPAP trial in patients with T2D and is the first to focus on diabetes complications. This feasibility RCT highlights the challenges that can be faced when conducting CPAP trials in patients with T2D and the possible underlying causes.

We found that clinicians are willing to recruit patients and patients are willing to be randomly assigned. Neither the study teams nor the patients reported concerns with the portable sleep studies, despite the recruiting teams’ having no prior experience conducting OSA research. Study teams were trained to educate the patients about using portable sleep devices and how to download the data to the cloud-based system. This study also achieved high follow-up rates despite the challenging circumstances, likely due to the reliance on remote monitoring. However, 3 out of 5 feasibility criteria were not met.

Two unmet criteria are related to timelines (recruitment and CPAP initiation). The recruitment timelines targets were missed, mostly due to the lengthy delays in obtaining local approvals from the study sites despite no difficulties obtaining overall ethical approval for the study. We have no definite reason for this delay. We postulate that it is due to the high workload on the research and development departments that resulted in some studies’ being prioritized over others.

We met the target for scoring the sleep studies quickly after the studies were uploaded to the online portal. However, initiating CPAP with a specific timescale was not achieved due to constraints imposed by patients’ schedules. Evidently, these targets were unrealistic, and future studies should allocate more time for opening study sites in the United Kingdom and provide longer intervals between randomization and CPAP initiation, which should not affect future trials’ integrity.

In addition, we failed to achieve our target of CPAP compliance. It is well-recognized that achieving adequate CPAP compliance is challenging in real life or in clinical trials. After the first night of CPAP, 8–15% of patients refuse further treatment,²⁹ and at least 50% discontinue usage within 1 year.²⁹ Nonadherence rates of 29–83% were reported in different clinical trials.³⁰ Several factors have been linked to a lack of adherence, including higher disease severity, sleepiness, mood disorders, ethnicity, socioeconomic status, claustrophobia, unrealistic expectations, and poor disease and treatment knowledge.^31,32 Some factors are associated with better adherence, such as using CPAP auto-titration, heated humidification, self-efficacy, and the presence of a bed partner.³² In our study, we aimed to improve adherence through various strategies, including auto-titration CPAP with humidifiers, remote monitoring, intensive follow-up, and extensive education on CPAP initiation troubleshooting. However, factors such as excessive daytime sleepiness and a modest AHI (below 15 events/h) in some patients have contributed to low adherence. Withdrawals from CPAP occurred in the first 10 months of the trial, with reasons given to stop using CPAP, such as other health problems (eg, osteoarthritis), lack of support at home, or change in personal circumstances (eg, frequent travels). Our exploratory analysis showed that CPAP compliance from as early as 2 weeks could predict CPAP compliance at 1 year, which is consistent with other studies.³¹ A run-in period of as short as 2 weeks might have reduced the number of patients not using CPAP after randomization and improved adherence.

We have measured the impact of CPAP treatment on many clinical variables over the 2 years, with a specific focus on diabetes-related microvascular complications. Previous RCTs of CPAP in patients with T2D were relatively short (3–6 months) and mainly focused on glycemic outcomes or cardiovascular disease risk factors.³³ Hence, this trial adds many novel insights into the links between OSA and diabetes-related outcomes. However, due to the feasibility nature of the trial and to increase the burden on patients, we elected to use no treatment at the control.³⁴

Also, the study was not powered to detect statistically significant differences between treatment arms. These analyses should be treated as hypothesis-generating to inform the choice of future definitive RCTs. Despite not meeting the compliance criteria, we found a possible favorable association between being randomly assigned to the CPAP arm vs no CPAP in relation to several outcomes related to chronic kidney disease, peripheral neuropathy, and quality of life. If these findings are proven in future RCTs, this will provide another treatment option to reduce the burden of T2D on individuals and the health care system. Despite recent improvements in the treatment options for diabetes complications, there is still a big unmet need.^35,36

The observed favorable associations between CPAP use and improved nephropathy/nephropathy markers are plausible biologically. OSA has been linked to the development and progression of chronic kidney disease, nephropathy, and retinopathy in patients with T2D, as well as the worsening of intermediary mechanisms such as oxidative stress, inflammation, and endothelial function.^14,15,37,38 CPAP has been shown to have favorable effects on many of the above mechanisms and pathways that lead to microvascular complications.³⁹

We have not specified the routine care in the protocol or directly collected data about the routine care received. Hence, we cannot be sure there were no differences in routine care that might have affected the study outcomes. However, there is no meaningful impact of CPAP on glycemic control, lipids, blood pressure or adiposity (Table S2 ^{(372.4KB, pdf)} ). We have also found that there were some differences in the prescription of glucose- and lipid-lowering treatments, insulin, and antihypertensives in favor of the no-CPAP arm by the study’s end (Table S3 ^{(372.4KB, pdf)} ). Hence, if there were any differences in routine care, they are likely to be in favor of the no-CPAP arm and do not bias the observed findings in favor of CPAP.

This feasibility RCT has limitations, including low CPAP compliance and a high proportion of missing data regarding clinical variables attributed partly to the coronavirus disease 2019 pandemic. The study is not adequately powered to determine the clinical benefits of CPAP treatment. Also, by excluding patients with excessive daytime sleepiness from the RCT, we might have excluded patients with more severe diseases that might have better CPAP compliance. However, we needed to ensure patients’ safety by excluding patients where random assignment to no CPAP would have been unsafe. Based on the study protocol, patients initiated on CPAP had extra contact to initiate CPAP and support CPAP compliance. This extra contact was mainly in the early stages of the study. Differences in contact between study arms might affect the study outcomes. However, this is unlikely to be the case in this study due to the long time between the extra contact and the end points’ assessment time.

In our trial, we did not have study-end AHI measurement, so we cannot confirm that the findings are related to improvement in AHI. However, we have used an intention-to-treat analysis in order to reduce biases that might overestimate the effect of the intervention, and we found no meaningful effect of CPAP on important risk factors related to the outcome, such as blood pressure, glycemic control, lipids, or adiposity measures. Also, we have adjusted the comparison of the secondary study outcomes between study arms to some important biologically important variables. However, the list of variables was not exhaustive, and other variables could be relevant, such as body mass index. Body mass index improved in both arms, and the adjusted difference is clinically not meaningful. Due to the feasibility nature and small sample size, we elected not to adjust for further variables, including body mass index, especially because the findings of this trial need to be tested in full RCTs.

Furthermore, Clinical measures were used to assess peripheral neuropathy instead of more sensitive tools, such as nerve conduction studies, for practical reasons and based on previous trials.⁴⁰ Patients continued to receive routine care regardless of being randomly assigned to CPAP or no CPAP. On the other hand, this trial has notable strengths. It is the longest CPAP RCT trial in patients with T2D, assessing outcomes beyond glycemic control and related to hard end points. The trial also achieved high follow-up rates through the remote data collection method.

CONCLUSIONS

In conclusion, conducting long-term RCT with the current design is probably not feasible. However, having a run-in period and selecting those more compliant with CPAP could be feasible in a full RCT. It is important to assess the OSA treatment on diabetes-related microvascular complications considering the high burden of these diseases. It might significantly affect reducing those burdens on patients and the health care system. However, such potential benefits need to be examined in future RCTs.

DISCLOSURE STATEMENT

The manuscript has been read and approved by all named authors. This is a multicenter study that was mainly conducted at the University of Birmingham. However, participants were recruited from 13 different National Health Service Trusts throughout England (detailed in the supplemental material). This project was funded as part of National Institute for Health and Care Research (NIHR) Clinician Scientist (CS-2013-13-029). The views presented in this manuscript are those of the authors and not those of NIHR or the National Health Service. A.A.T. is currently an employee of and has shares in Novo Nordisk. The views expressed in this manuscript are those of the author and not Novo Nordisk. Novo Nordisk had no role in this manuscript. Srikanth Bellary has received speaker fees and honoraria from Novo Nordisk, Eli Lilly, Boehringer Ingelheim, and Astra Zeneca. Mayank Patel has received speaker fees from Eli Lilly and Company, Insulet, and Astra Zeneca. The other authors report no conflicts of interest.

ABBREVIATIONS

AHI

apnea-hypopnea index

CPAP

continuous positive airway pressure

MNSI

Michigan Neuropathy Screening Instrument

OSA

obstructive sleep apnea

T2D

type 2diabetes