Page 3 - Cannabis for the Management of Pain: Assessment of Safety Study

Ware et al The Journal of Pain 1235
and forced expired flow over the middle half of the vital in the control group above .20 case/person-year. These
capacity (FEF 25–75% ). estimates were derived from interim safety analyses dur-
Other safety parameters. Blood tests measured ing a protocol revision (see Supplementary Materials S-3)
hematological, biochemical, liver, kidney, and endocrine and are consistent with estimates from a meta-analysis of
function (prolactin, testosterone, thyroid-stimulating AEs from prescription cannabinoids. 16
hormone).
Efficacy measures. Pain intensity was measured using
visual analog scales (0, no pain; 10, worst pain possible)
as the average, highest, and lowest in the past 7 days, Statistical Analysis
and current pain intensity at the time of visit. Pain quality Primary Analysis
was assessed using the McGill Pain Questionnaire, which
measures sensory, affective, and evaluative dimensions Demographic and clinical characteristics were
of pain. Other symptoms were measured using the modi- compared between the cannabis and control groups us-
fied Edmonton Symptom Assessment Scale. Mood was ing parametric and non-parametric statistics as appro-
measured using the Profile of Mood States. Quality of priate. Reasons for withdrawals were tabulated for
life was measured using the SF-36. both groups. AEs were coded and tabulated using the
MedDRA headings ‘‘system organ classes’’ and ‘‘preferred
terms.’’ AEs were characterized by severity, causality, and
Study Procedures
outcome.
Baseline Assessment For incidence rate estimates, cumulative person-years
were calculated from the date of the baseline visit until
All patients underwent baseline history and physical
the date of discontinuation, death, or completion of
examinations, addiction screening (Drug Abuse the study, whichever came first. The 95% confidence in-
Screening Test [DAST-20]), neurocognitive testing, and
tervals (CIs) for the rates were calculated using the Pois-
urine drug testing (enzyme-linked immunosorbent son distribution assumption.
assay). Blood tests and pulmonary function tests were
An overdispersed Poisson regression model was used
conducted in the cannabis group only.
to assess the occurrence of AEs among cannabis users
Follow-Up or controls. 2,5,10 The results of the regression analyses
are presented as incidence rate ratios (IRRs) with
Intended follow-up was for 1 year. Six clinical visits (1,
corresponding 95% CIs. Logistic regression analysis was
2, 3, 6, 9, and 12 months after baseline) and 3 telephone also performed to explore the association between the
interviews (1, 2, and 3 weeks after the baseline visit) were
risk of having at least 1 AE and medical cannabis use.
scheduled for patients in the cannabis group; 2 clinical
Odd ratios (ORs) with 95% CIs were calculated.
visits (6 and 12 months after baseline) and 5 telephone
interviews (1, 2, and 3 weeks, 3 and 9 months after base- Subgroup Analysis
line) were scheduled for control patients.
To further control for confounding by past cannabis
Neurocognitive and efficacy assessments were con-
use, we estimated the stratified incidence rate of AEs
ducted at 6 and 12 months in all patients. Pulmonary by past cannabis use in the cannabis and control groups.
function tests were repeated in the cannabis group at
We grouped past cannabis use into 3 categories. ‘‘Cur-
12 months. Blood tests were conducted in the cannabis rent cannabis users’’ were those who reported using
group at 1, 6, and 12 months. Patients were not specif-
cannabis at the baseline interview; ‘‘ex-cannabis users’’
ically instructed to abstain from cannabis use before were those who reported having previously used
any study visits.
cannabis but not at the baseline interview; ‘‘naive users’’
AE Reporting were those who reported never having used cannabis
before the baseline interview. We carried out a Poisson
AEs were captured during interviews at clinic visits, regression analysis to explore whether the incidence
during telephone contacts, or spontaneously by calling
rate was consistent among participants with different
the study nurse. At site visits, the study monitor reviewed histories of cannabis use.
patients’ hospital charts to ensure that serious events
were not missed. Secondary Analyses
A random effects model with a random intercept for
Sample Size and Power Considerations patient was used to model neurocognitive and pulmo-
For the primary outcome, the incidence of AEs among nary function, pain, mood, symptom severity, and quality
cannabis users was compared with controls. It was of life. Age, gender, disability status, average pain inten-
assumed that SAEs followed Poisson distributions in the sity, concomitant pain medication use, alcohol use (cur-
2 study groups. The intended sample size of this study rent vs former or never users), tobacco use (current vs
350 cannabis-using participants and 350 controls (see former or never users), past cannabis use (ever vs never),
Supplementary Materials S-3) meant that a rate ratio of and study sites were incorporated as covariates.
1.5 could be detected at powers above 60% for a control Statistical analyses were undertaken with SAS soft-
group incidence rate of SAEs above .15 case/person-year, ware (version 9.1; SAS Institute Inc, Cary, NC). No adjust-
and at a power above 70% for the incidence rate of SAEs ments for multiple comparisons were made.