#NephJC Chat
Tuesday Feb 20 2024, at 9 pm Eastern
Wednesday Feb 21, 2024 at 9 pm Indian Standard Time
Ann Intern Med . 2024 Jan 16. doi: 10.7326/M23-1794. Online ahead of print.
SLEEP-HD: Effectiveness of Existing Insomnia Therapies for Patients Undergoing Hemodialysis : A Randomized Clinical Trial
Rajnish Mehrotra 1, Daniel Cukor 2, Susan M McCurry 3, Tessa Rue 4, Maria-Eleni Roumelioti 5, Patrick J Heagerty 4, Mark Unruh 5
PMID: 38224591
Introduction
Insomnia, lack of sleep, sleep deprivation... As Nobel laureate Aleksandr Solzhenitsyn relates in his book The Gulag Archipelago or CIA reports (Hinkle LE & Wolff HG, AMA Arch Neurol Psychiatry, 1956), lack of sleep was used as an effective torture method, even if in short terms it could determine cognitive impairment and hallucinations. In a wider, more appropriate sense to our medical purpose, insomnia disrupts homeostasis, both short- and long-term, contributing to hypertension, cardiovascular disease, metabolic syndrome, type 2 diabetes, and chronic pain (Medic G et al, 2017).
According to CDC, in the adult population, the prevalence of insomnia in the US is between 25 and 49% (Dopheide JA, 2020, Ellis J et al, 2023). Insomnia is more prevalent in population needing dialysis approx. 68% (Tan LH et al, 2022). Sleep is recognized by SONG-HD as a tier 3- important outcome, and as tier 2- critically important by SONG-PD. Higher prevalence of insomnia compared to the general population can be explained by dialysis or chronic kidney disease (CKD) related factors- as depicted in the infographic below.
Trazodone is commonly used for insomnia in many parts of the world, because of its perceived efficacy and safety, though it is difficult to envisage how a pill can overcome the complex phenotype as shown in this table. On the other hand, though insomnia guidelines (Edinger et al, JCSM 2021) recommend CBT, there is weak evidence to demonstrate its efficacy in dialysis (see in the table below). This was the rationale for conducting the SLEEP-HD (Short and Long-term Effectiveness of Existing insomnia therapies for Patients undergoing HemoDialysis) trial.
The Study
Methods
SLEEP-HD was a parallel-group, multicenter, placebo-controlled trial of three arms of which one (CBT) was open-label, while the other two (trazodone or placebo) were blinded.
Setting and participants
The trial enrolled participants from 26 community-based dialysis units in Albuquerque, New Mexico, and Seattle, Washington. Patients undergoing thrice-weekly in-center hemodialysis were included. In the pre-screening phase, the Insomnia Severity Index questionnaire was used to divide patients into mild (ISI score 8–14), moderate (ISI score 15–21), and severe insomnia (ISI score 22–28). Patients with an ISI score ≥10 and those who reported sleep disturbances for 3 or more nights per week or more months (chronic insomnia) were considered for inclusion.
The insomnia severity index, adapted from Morin et al, 2011
Exclusion criteria:
severe cognitive impairment
suicidal ideation
severe depressive symptoms
alcohol or substance abuse
severe restless leg syndrome
chronic trazodone treatment
allergy to trazodone
on medications that prolong rate-corrected QT
pregnancy or lactation
anticipated life expectancy shorter than 3 months
anticipated transplant or transition to home dialysis in the next 6 months
Intervention
During 1-week run-in-phase, all eligible participants were required to keep daily sleep diaries. Following this, participants were randomly assigned to 6 weeks of telehealth cognitive behavioral therapy for insomnia (CBT-I), trazodone, or placebo.
Participants undergoing telehealth CBT-I had six 30-minute weekly sessions via Zoom. Certified therapists delivered sessions covering all components of insomnia. Adaptations have been made for specific behaviors- like napping during hemodialysis. Participants could receive therapy at home or in a dialysis unit. Daily sleep diaries were maintained. Some randomly selected sessions were recorded.
Components of Telehealth CBT-I Intervention from Unruh M et al, 2020
Study drug started at 50 mg nightly and could be titrated to 100 mg in week 2 or 3 and was maintained until tthe end of the study. Patients could use other OTC sleep aids during the study, and 6 weeks after the study concluded, they could request trazodone regardless of their trial assignment.
Figure: An overview of the design of the SLEEP-HD Trial, from Unruh M et al, 2020
Outcomes
Primary outcome: severity of insomnia measured by 7-item ISI (score 0-28).
Secondary outcomes included:
patient-reported outcomes with various scales
cumulative use of sedative and hypnotics
nighttime sleep efficiency using actigraphy
Table: Patient Reported Outcome Instruments Collected by CATI, from Unruh M, et al, 2020
Computer-assisted telephone interviews were conducted by a blinded assistant in a fixed sequence, starting with ISI. Actigraphy used monitors worn for 7 days at baseline, 7 weeks and 25 weeks, with data analyzed using Actiware Sleep software.
COVID-19 pandemic
Enrollment paused twice due to COVID-19 pandemic, first in March 2020 (79 participants), and again in February 2021 (91 participants). Despite halts, assigned treatments and study procedures continued with adapted remote methods. Follow-up visits were conducted over the phone, medication delivery via interoffice mail, and actigraphy resumed with single-use wristbands and device decontamination between participants.
Statistical analysis
The study aimed to assess the effectiveness of 6 weeks of CBT-I, trazodone, or placebo for chronic insomnia in hemodialysis patients. A difference in scores of 4 or greater was conidered clinically important. The primary analysis utilized an intention-to-treat approach with a longitudinal linear mixed-effects model, incorporating baseline ISI scores and treatment-by-time interactions. Sensitivity analysis involved 40-fold multiple imputation.
Funding
The trial was funded by the National Institute of Health/National Institute of Diabetes and Digestive and Kidney Diseases, which had no input on study conduct, data analysis, or writing the manuscript.
Results
Over 42 months, 923 out of 1068 patients undergoing in-center hemodialysis completed ISI. Among 411 patients with an ISI score ≥10 and symptoms on 3 or more nights per week for at least 3 months, 126 met eligibility criteria. They were randomly assigned to CBT-I (43), trazodone (42), or placebo (41). (Appendix figure).
Appendix figure: CONSORT diagram describing enrollment, randomization, and follow-up of participants in SLEEP-HD trial, from Mehrotra R et al, 2024
Patients from the trazodone group were somewhat younger, with a higher dialysis vintage and a higher prevalence of ischemic heart disease. Otherwise, about half the patients were female, and about 20% were Black.
Table 1: Baseline characteristics from Mehrotra R et al, 2024
Patients were followed for a median of 174 days post-randomization. In CBT-I, arm 3 withdrew before the 7-week assessment, with 1 lost follow-up. They completed a median of 6 sessions during 6-week treatment, and 79% completed at least 4 sessions. During long-term follow-up, 2 patients died and 1 dropped out. In the trazodone group, 1 patient died before the 7-week assessment. At 6 weeks, the median prescribed dose was 100 mg, with 81% still on the drug. Two participants died during long-term follow-up. In the placebo group, 1 died before 7-week assessment, and 2 died during follow-up. Treatment participation remained consistent before and during the pandemic.
After 6 week treatment, when every patient could ask for trazodone, 17 participants reported trazodone use at week 13 visit, and 18 at 25-week visit.
Supplement Table 4. Participants reporting use of trazodone in the previous seven days at 13 weeks and 25 weeks after randomization, by treatment assignment from Mehrotra R et al, 2024
Primary outcome
The mean ISI scores decreased in all 3 groups. There was no difference in change of ISI score from baseline at 7 weeks (omnibus P value= 0.68) and 25 weeks (omnibus P value= 0.84)
Table 2: Primary outcome from Mehrotra R et al, 2024
Figure. Comparative effects of CBT-I, trazodone, and placebo in participants with chronic insomnia undergoing long-term hemodialysis, from Mehrotra R et al, 2024
The sensitivity analysis, using multiple imputation for missing data, was consistent with primary analysis.
Supplement Table 5. Primary patient-reported outcomes using multiple imputation for missing data, from Mehrotra R et al, 2024
Secondary endpoints
Even if at 7 weeks the PSQI score was lower with CBT-I compared with trazodone or placebo, by 25- week assessment, there was no difference between groups. The initial better scores with CBT-I were attributed to participants’ use of aid medication. There was no difference between groups in other patient reported outcomes.
Figure. Patient reported secondary outcomes, from Mehrotra R et al, 2024
There was no difference in the cumulative use of sedatives and hypnotics (excluding study drug).
Table 2: Secondary outcomes from Mehrota R et al, 2024
Exploratory endpoints
There were no substantial differences observed among treatment groups regarding the number of participants attaining remission of insomnia (ISI score ≤7) or experiencing significant clinical improvement (indicated by a decrease of ≥8 points in ISI scores).
Supplement Table 6. Number of participants with response to treatment using various measures on the insomnia severity score, from Mehrotra R et al, 2024
Safety
The annualized incidence rate for cardiac-related hospitalizations was 12 times higher with trazodone vs CBT-I and 3 times higher than placebo
Serious adverse events were noticed in all groups, as expected in a dialysis population among whom baseline hospitalization and mortality are quite common). The highest annualized incidence rate for serious adverse events was associated with trazodone (1.85, 95% CI, 1.29 to 2.53), and the lowest with CBT-I (0.79, CI 0.44 to 1.3). Notably, the difference comes from higher hospitalization rates, due to cardiac causes. The annualized incidence rate for cardiac-related hospitalizations was 12 times higher with trazodone (RR 12.22, CI 1.83 to 519.2) vs CBT-I and 3 times higher than placebo (RR 3.14, CI 0.97 to 13.22)
Table 3: Annualized incidence rate and incidence ratio ratios for adverse events, from Mehrotra R et al, 2024
Supplement Table 7 depicts an overview of cardiac adverse effects, showing that even with no history of ischemic heart disease, cardiac effects were more common in trazodone and placebo arms. Moreover, among those with preexisting ischemic heart disease, in the trazodone group, about a third of participants had cardiac adverse effects.
Supplement Table 7. Percent of participants experiencing at least one serious cardiac adverse event by history of ischemic disease (MI, angina, coronary revascularization), from Mehrotra R et al, 2024
In contrast, the rate of non-serious adverse events was similar across the groups.
Supplement Table 8C. Incidence rate ratios (95% CIs) for adverse events for the first 7 weeks and for the 25-week study period by treatment assignment, from Mehrotra R et al, 2024
Discussion
In this well conducted study of three arms, including a double-blind placebo arm, there was no difference in efficacy between placebo, trazodone, and CBT. However, trazodone was associated with a greater adverse effect burden, and CBT with the lowest. Thus trazodone was not only not better at helping people sleep than placebo, it was actually harmful.
This is the largest RCT addressing a very common problem that affects almost half of the hemodialysis population. It was clearly designed, and the two pharmacotherapy arms were double-blinded; the CBT arm could inherently not be blinded. The outcomes were validated scores used in the insomnia literature. Despite COVID, the investigators persevered, and completed the trial albeit on a longer timeline.
One can possibly understand trazodone not being effective - after all, it is approved by the US FDA for treatment of depression, and not insomnia. The use for helping people sleep is off-label, and this RCT supports it remaining that way. The purported mechanism of trazodone is from its mechanism as a serotonin antagonist and reuptake inhibitor (SARI) and not a classic SSRI. It inhibits the reuptake of serotonin and blocks the histamine and alpha-1-adrenergic receptors. It also induces significant changes in 5-HT presynaptic receptor adrenoreceptors. These are, needless to say, mere theoretical considerations given its lack of efficacy here. However, the effect on alpha adrenergic receptors (and on hERG potassium channels, see Tarantino et al EJCP 2005) may explain the adverse effects. Among its listed known adverse effects include orthostatic hypotension, QT prolongation, and torsades. In hindsight, we should not be surprised at all at the toxicity reported with trazodone here.
Does this mean nothing works for insomnia (given that a placebo was just as useful or useless?). One may reach for CBT, since it was without harm, if it is available and accessible? There are other - all small, short-term, and mostly single-centre RCTs - with a wide variety of interventions including acupressure and melatonin. Rather than try a variety of interventions, we owe it to perform rigorous RCTs such as SLEEP-HD for this common problem that leaves hemodialysis patients sleepless.
Conclusion
There was no difference between CBT, trazodone, and placebo for helping dialysis patients sleep. However, trazodone was harmful, and we should stop using this drug. Trazadont!
Summary by
Cristina Popa and Swapnil Hiremath
Header Image created by AI, based on prompts by Evan Zeitler