This was a randomized, double-blind, parallel group clinical trial comprising a 2-week, single-blind, placebo run-in period (Baseline), a 3-week double-blind treatment period (treatment weeks 1-3) followed by a 26- week double-blind extension period (treatment weeks 4 to 29) in which patients were randomized to receive PRM (Circadin® 2 mg, Neurim Pharmaceuticals Ltd, Tel Aviv, Israel) or placebo, given orally as one tablet per day 2 h before bedtime, and a 2-week single-blind placebo run-out period (withdrawal). The study protocol and relevant documents were approved by Huntingdon Multi-centre Research Ethics Committee, Cambridge, UK. Participants provided written informed consent.
Patients were recruited from Glasgow and the surrounding areas (West of Scotland) and were pre-screened by telephone using the Sleep History Questionnaire (SHQ) within 1 month of the baseline screening visit. The SHQ was adapted from The Management of Insomnia Guidelines for Clinical Practice and resembled that recommended by Clinical Practice Guideline - Adult Insomnia [27, 28]. Suitable patients were invited to Visit 1 during which they were consented and assessed for inclusion.
Men and women aged between 18 and 80 years suffering from primary insomnia according to the Diagnostic and Statistical Manual for Mental Disorders (DSM-IV) criteria with sleep latency longer than 20 min were included in the study.
The major exclusion criteria for the study included the use of benzodiazepine or non-benzodiazepine hypnotics within the previous 2 weeks or any psychoactive treatment within the previous 3 months, sleep disorders associated with a psychiatric disorder (for example, depression, anxiety, dementia), sleep disorders secondary to another medical condition (for example, sleep apnoea, circadian rhythm sleep disorder), use of prohibited concomitant medication [psychotropic treatments - neuroleptics, antiepileptics, barbiturates, antidepressants, anxiolytics and lithium, first generation antihistamines, hypnotics or treatments used as a hypnotic (for example, all benzodiazepines, zopiclone, zolpidem and zaleplon, barbiturates, buspirone and hydroxyzine)] or excessive alcohol consumption, any chronic medical condition that was likely to be the cause of the sleep problem (for example, chronic pain, benign prostatic hypertrophy) or might interfere with the conduct of the study or a lifestyle likely to interfere with sleep patterns (for example, shift work, jet-lag).
A four-step process was used for screening out patients with secondary sleep disorders including depression and other sleep disorders in the study according to DSM-IV criteria.
Step 1: The initial prescreening for primary insomnia as defined in DSM-IV was performed on a telephone interview and was based on the SHQ. The SHQ characterizes the primary sleep complaint according to the differential diagnostic criteria (DSM-IV and International Classification of Diseases-10) and also helps in differentiating primary insomnia from insomnia due to medical and psychiatric disorders (including depression and anxiety) and specific insomnia disorders such as circadian rhythm disorders, movement disorders, parasomnias and breathing related sleep disorders.
Step 2: At the screening visit, a physical examination was performed by a qualified clinician to exclude patients with physical causes of insomnia.
Step 3: At the screening visit the patients went through a detailed psychological assessment that included the Raskin Depression scale, Covi anxiety scale and the Mini Mental State (MMS) in order to exclude psychiatric disorders, including depression anxiety and dementia. In addition, a history of severe psychiatric disorders, especially psychosis, anxiety and depression were major exclusion criteria.
Step 4: Patients who were using psychotropics (neuroleptics, antiepileptics, barbiturates, antidepressants, anxiolytics or lithium) in the 3 months before the study were excluded. A urine drug screen for benzodiazepines and morphine derivatives was undertaken at baseline. Patients with a positive result were excluded. Hypnotic use was monitored throughout the study. Patients were asked at each visit whether they took a hypnotic beside medication and were withdrawn if they did. The common analgesics used in UK for self limiting intermittent problems such as headache frequently contain codeine. Patients for whom pain was a cause of insomnia were excluded from the study. However, due to the long-term nature of the study intermittent use of common analgesics was allowed.
The screening and run-in periods were used to wash-out previously administered medicinal products which were incompatible with the trial, for confirmation of a stable disease and compliance with study medication and procedures and for the qualitative and quantitative baseline assessments of patients. Patients with major short-term fluctuations of their condition and non-compliance with study procedures were excluded. A history of sleep latency of >20 min, required for patients inclusion, was assessed once in the telephone interview (SHQ), confirmed at the screening and then at the baseline visits using the PSQI [28–30]. Eligible patients entered the baseline screening run-in period and received 2 weeks of single-blind treatment with placebo. Patients still eligible after the 2-week placebo run-in and who were compliant with respect to treatment, had a negative drug screen and correctly completed study assessment forms were randomized in a 1:1 ratio to receive either PRM 2 mg or placebo for 3 weeks in a double-blind manner. Randomization was stratified by trial site, 6-SMT levels (low ≤8 μg versus high >8 μg/night) and age group (< 65 versus ≥65 years).
After the 3-week treatment period, completing patients were allowed to proceed into the extension period. All PRM patients stayed on PRM and all placebo patients were randomised in a 1:1 ratio to receive either PRM 2 mg or placebo for 26 weeks, resulting in a 3:1 ratio of PRM to placebo.
At the end of the extension period, all patients received 2 weeks of single-blind placebo in the run-out period to evaluate withdrawal effects. The overall duration of the study was 33 weeks.
Patients were instructed to take one tablet daily of study medication orally, 1-2 h before going to bed (preferably between 2100 h and 2200 h) and after food, and were asked to fill in a diary each morning, reporting on sleep latency, sleep maintenance, total sleep time, time going to bed, sleep offset time, refreshed on waking score, morning alertness score, and sleep quality in the previous night).
The treatment period was double-blind with two parallel treatment groups. Selection for a treatment group was determined by a computer generated randomization list in a 1:1 ratio (PRM mg to placebo). The list was constructed using the method of randomized permuted blocks. Randomization was stratified by trial site, 6-SMT levels (low/high) and age group (< 65/> 65). A centralized randomization system (Interactive Voice Response System [IVRS]) was used. Sites called the IVRS, using a toll-free number. The patient's status: trial site, 6-SMT level (low excretors/high excretors) and age (< 65/>65) were entered into the IVRS. The IVRS randomised the patients and provided a double-blind treatment kit assignment.
By the end of the 3 week treatment period, the PRM patients remained on the active medication and placebo patients were randomized again for the double-blind extension period. Selection for a treatment group was determined by a computer generated randomization list in a 1:1 ratio (PRM 2 mg to placebo). This procedure was performed for all participants using the centralized IVRS randomization system keeping the patient and the study personnel (investigator and nurses) blind to the allocated treatment. The list was constructed using the method of randomized permuted blocks.
Blinding was maintained by use of a matching placebo identical in appearance taste and smell to the active medication and use of an independent IVRS to allocate randomized treatment. During the 33 weeks of the study (run-in, double-blind treatment, double-blind extension and run-out), patients were blinded regarding the type of medication they were receiving (placebo or active). During the double-blind treatment and extension periods, the investigator(s) and their staff were also blinded regarding the treatment administered (double-blind). However, they were aware that the patient was receiving placebo during the run-in and run-out periods.
The blinding was not to be broken (unless in an emergency) until the database was locked to perform planned analyses. The IVRS was used to break a code in case of emergency. This was to be done only when the investigator decided that knowledge of which study treatment the patient had been randomized to, would affect the management of an adverse experience. There were three people for whom the blind was broken all due to a serious adverse event (SAE), two were on PRM and one was on a placebo.
The main objective of this study was to assess the effects of short-term (3-week) therapy with PRM versus placebo on patient reported sleep latency (sleep diary) in their natural setting, in patients with low endogenous melatonin levels (≤8 versus >8 μg urinary 6-SMT/night) and in elderly patients (65-80 years old). Additional sleep and daytime parameters, safety and maintenance of PRM efficacy and safety over a 6-month period were also evaluated.
Efficacy variables were recorded at baseline and each visit. PSQI [29, 30] global score, component scores, and questions 2 and 4 filled in by the investigator with the patient; sleep diary (National Sleep Foundation diary) parameters (daytime and night-time) were filled in by patient each day in the morning in the 7 days preceding each visit; World Health Organization (WHO)-5 Well-being Index (1998 version) and Clinical Global Impression of Improvement (CGI) - Severity of Illness Scale (CGI-S)  were filled in by the investigator with the patient at screening and baseline and the improvement (CGI-I) at each subsequent visit.
The PSQI has been recommended as an essential measure for global sleep and insomnia symptoms in recent expert consensus recommendations for a standard set of research assessments in insomnia . It comprises nine questions relating to the patient's usual sleep habits during the previous 2 weeks; the second and third weeks of active treatment. It addresses possible reasons for trouble in sleeping as well as daytime behaviour. An algorithm is used to calculate seven component scores and these are added to give a global PSQI score. The PSQI component scores, Question 2 (Sleep Latency) and Question 4 (Total Sleep Time) after 3 weeks' double-blind treatment, and the change from baseline levels of these parameters. It has been shown that each of the PSQI individual component scores measures a particular aspect of the overall construct. Furthermore, control subjects differ from insomnia patients in all individual components . However, the correlation between individual items and global score ranged from 0.83 (subjective sleep quality) to 0.07 (cough or snore during sleep) . In the evaluation of the drug effects it was therefore interesting to look at each component.
Safety variables and vital signs (pulse, blood pressure) were assessed at each visit including spontaneously reported adverse events (AEs); unusual events and AEs observed by the investigator. Physical examination was performed at screening, end of run-in, after three and 29 treatment weeks and at discontinuation. An electrocardiograph was recorded at end of run-in, after three, seven and 29 treatment weeks and at discontinuation. Vital signs (pulse, blood pressure) were recorded at all visits. Laboratory tests (haematology, biochemistry and urinalysis) were assessed at screening, after three, seven and 29 treatment weeks and at discontinuation. Endocrine evaluations were performed at screening and after 29 treatment weeks in 80 patients who were not using any hormonal contraceptives or hormonal replacement therapies and who were not suffering from any significant endocrine disease. Cortisol was assessed at screening and after 29 treatment weeks in 56 patients before and after synacthen test. The Tyrer scale was completed by the investigator after 29 treatment weeks and at withdrawal.
The predefined primary efficacy variable was the comparison of sleep latency as measured by the sleep diary at 3 weeks treatment weeks with PRM (2 mg) or placebo in the pre-defined subgroups of patients who were low excretors of melatonin regardless of age (primary endpoint) and the patients aged 65-80 years, regardless of melatonin levels. The comparison was done using a linear regression model with terms for treatment (PRM versus placebo), baseline sleep latency and age group (≥65 or <65 - only for the primary endpoint). In compliance with US Food Drug Administration regulatory procedures, no correction for multiple comparisons were performed for the primary outcome measure.
All other efficacy endpoints were pre-defined as exploratory and aimed at confirming the results of the primary analysis using additional instruments (for example, PSQI) or adding information on other aspects of the sleep and daytime consequences of the treatment including: (1) time going to bed and sleep offset times, sleep maintenance, total sleep time, sleep quality and morning alertness from the sleep diaries; (2) the PSQI global score; (3) PSQI questions 2 (sleep latency in minutes) and 4 (total sleep time in minutes) and the individual PSQI components; (4) the CGI-I score assessed by the clinician at three to 29 treatment weeks) quality of life derived from the WHO-5 Well-being index covering positive mood, vitality and general interests.
Our main conclusion was based on sleep latency, the predefined primary variable. No correction was made for multiple statistical testing for the exploratory variables. Accordingly, the overall conclusions from the results are based on the accumulation of evidence for between-treatment differences which were, in many cases, correlated or complementary, rather than on isolated P-values.
Sleep latency as recorded in the sleep diary was summarized for low excretors aged 18-80 years and for patients aged 65-80 years, at baseline after the 2-week run-in period and after 3 weeks double-blind treatment (actual and change from baseline) for each treatment group and, as a whole, using descriptive statistics for continuous variables. At each visit, the mean value of the 7 days prior to the visit was used. Sleep latency as measured by the sleep diary after 3 weeks double-blind treatment was compared using a linear regression model with terms for treatment (PRM versus placebo), baseline sleep latency and age group (≥65 or <65 years).
The other short-term variables were summarized by the mean values at baseline and after 3 weeks of double-blind treatment (actual and change from baseline) using descriptive statistics of continuous variables for each treatment group. These included: (1) sleep diary variables, calculated as the mean of the values recorded in the 7 days prior to each study visit; (2) PSQI global, individual component, question 2 and question 4 scores; (3) WHO-5 Well-being Index score; (4) CGI-S (Visit 2) and CGI-I (Visit 3) scores.
Efficacy variables were summarized by the first randomization for outcomes at baseline and treatment weeks 4-29, or those visits at which the outcome was recorded. Summaries are given at each visit, and for changes between post-baseline visits and baseline. For those outcomes recorded at withdrawal, summaries are given for the change between treatment week 29 and withdrawal weeks. In addition, the changes of PSQI and WHO-5 between treatment week 29 and withdrawal weeks in the run-out period are summarized.
For efficacy outcomes measured at treatment weeks 4 to 29, a linear mixed effects model for repeated measures [MMRM] was used to compare outcomes at treatment weeks 4 to 29, in relation to the treatment currently received. For treatment week 3 measures, treatment was defined by the first randomization; for subsequent visits, treatment was defined by the second randomization. Each model included a random individual effect and assumed a general covariance structure for the residuals over time. For each outcome, a model was fitted which included terms for treatment, visit (as a categorical variable), the baseline values of the outcome measure, age group (≥65 or <65 years, except for analyses for the ≥65 years) and baseline 6-SMT (≤8 or >8 μg/night, except for analyses of low excretors). This model was used to estimate the global treatment effect, with a 95% confidence interval (CI) and P-value.
For each outcome, the above model was extended by including treatment-by-visit interaction terms. These models were used to estimate the treatment effect at each visit, with 95% CIs and P-values. A P-value for the treatment-by-visit interaction is provided, based on a likelihood ratio test. The global and visit-specific treatment effect estimates and 95% CIs are provided graphically, along with the estimated mean values and 95% CIs at each visit for each treatment group. In addition, a model was also fitted for each outcome, including a treatment-by-visit interaction, assuming a linear trend in the treatment effect over Visits 3 to 7, with the treatment effect changing by a fixed amount between each consecutive pair of visits; a likelihood ratio test P-value is given for this trend.
Adverse event data, clinical laboratory data, including hormones, vital signs and withdrawal symptoms, were summarized for all randomized study participants who took at least one dose of study medication, regardless of their subsequent participation in the study. No formal statistical testing was performed on the safety data.
In order to achieve 95% power at the 5% level for the primary objective of assessing the change in sleep latency in the Intention to Treat (ITT) low excretors at Week 3, assuming a treatment effect of 19 min and a residual standard deviation of 40.6 min, 120 participants were required per treatment group. Assuming equal numbers of low and high excretors, 480 patients were required to complete treatment Week 3. Assuming a 10% dropout rate between baseline and Week 3, 540 patients would have to be randomized at baseline. In order to achieve 90% power at the 5% level for the first secondary objective of assessing the change in sleep latency in the ITT dataset of patients aged 65-80 years, assuming a treatment effect of 14 min and a residual standard deviation of 40.7 min, 179 patients ≥65 years were required per group (active and placebo). Therefore, 400 patients in this age range would need to be randomized at baseline. Assuming 45% of 540 patients already randomized would be ≥65 years old (245), an additional 150 patients would need to be randomized at baseline in this age group.