Study design
The present population-based study of a large unselected adult population replicate
the methodological design used in the Tromsø Study [29], including similar questionnaire-based
diagnoses for headache and insomnia, identical categories of hs-CRP, and replication
of the statistical analyses with adjustments for the same type of potential confounders.
The Nord-Trøndelag Health Study
The total population of the Nord-Trøndelag county aged ≥20 years has been invited
four times between 1984 and 2019. All participant answered questionnaires and were
invited to interview, clinical examination, and blood sampling [30].
Study population
In HUNT3 (2006-2008), a total of 94,194 individuals aged ≥ 20 years were invited,
whereof 93,860 were eligible. Of these, 50,807 (54%) accepted the invitation. The
diagnoses of headache and insomnia was questionnaire-based. The present study included
data from 38,813 participants (41%) who had answered questions about headache and
insomnia, and had hs-CRP measured from non-fasting blood samples (Fig. 1).
C-reactive protein
Hs-CRP was analyzed at the Central Laboratory, Levanger Hospital, using Architect
cSystem ci8200, by latex immunoassay method. The detection limit was 0.03 mg/L, and
samples without detectable hs-CRP were assigned this value. In the present study we
replicated the strategy used in previous published studies from the Tromsø study [29,
31] defining normal CRP as 0-3.00 mg/L and elevated as 3.01-20.00 mg/L. Participants with hs-CRP values >20.00 mg/L, which
probably indicate some acute or chronic disease [32], were excluded (n=527) (Fig.1).
Headache questionnaire
All details of the headache questionnaire have been published elsewhere [33], and
all questionnaire-based headache diagnoses were identical to those presented in the Tromsø study [29]. Subjects who answered “no” to the screening question (“Have you suffered from headache
during the last 12 months?”) were included in the reference group without headache.
Participants who answered “yes” completed 12 additional questions, mainly to diagnose
migraine according to the second version of the International Classification of Headache
Disorders (ICHD-2) [34]. Regarding attack duration, <4 hours were accepted for those
who reported visual disturbance before headache. Participants who fulfilled the migraine
diagnosis and reported visual disturbance prior to headache were classified as migraine
with aura (MA), whereas the remaining with migraine had migraine without aura (MO).
The remaining participants with headache were classified as having “other headache”.
The merged group of “any headache” consisted of participants with migraine or other
headache. Furthermore, based to answer on the question about headache frequency, subjects
with headache were subdivided in two groups; headache < 7 days/month or headache ≥7
days/month.
The validity of the questionnaire-based headache diagnoses has been published in a
separate paper [39]. The agreement between the validation interview and questionnaire-based
headache categories were very good for any headache (kappa value at 0.70, 95 % CI
0.61-0.79) and moderate for migraine (kappa value 0.50, 95 % CI 0.32-0.68), whereas
the specificity for MA was high (95 %) (kappa value 0.44, 95 % CI 0.38-0.50) (33).
Questionnaire-based diagnosis of insomnia
The questionnaire-based diagnosis of insomnia was based on the DSM-V (Diagnostic and
Statistical Manual of Mental Disorders, 5th ed.), and was nearly identical to the
insomnia diagnosis used in the Tromsø study [29]. To fulfil the proxy diagnosis of
insomnia the participants had to answer “several times a week” to at least one out
of the three questions, asking whether they had “difficulties falling asleep at night”,
whether they “woke up repeatedly during the night”, and/or whether they “woke up too
early and could not get back to sleep” [36]. In addition, they also had to report
“several times a week” to the question “felt sleepy during day” [36]. The overall
agreement between questionnaire and interview for the first three questions has previously
been found to be moderate (kappa value 0.51, 95% CI 0.40-0.63) [37].
Potential confounders
The selection of potential confounders used in the Tromsø study [29] was based on
previous literature [35, 37, 42]. To replicate the Tromsø study [29], we included
almost identical potential confounders, with two exceptions: employment status was
used instead of education level, and Hospital Anxiety and Depression Scale instead
of Hopkins symptom checklist (HSCL-10). The remaining potential confounders were identical;
age (continuous variable); gender; body mass index (BMI) (<25, 25.0-29.9, and ≥30
kg/m2) (28, 32); smoking (current, previous, and never); physical activity (never, ≤ 1
time per week, ≥2 times per week); alcohol consumption (never, <2 times/week, ≥2times/week);
self-reported diabetes (yes/no); self-reported stroke and/or heart infarction (yes/no); and self-reported hypertension (yes/no).
Ethics
This study was approved by the Regional Committee for Medical and Health Research
Ethics, the Faculty of medicine, mailbox 8905, 7491 Trondheim. The approval number
was #2018/2422/Rek Midt. The participants have given written informed consent.
Statistical analysis
The statistical method was identical to that presented in the Tromsø Study [29], and
based on multivariate analyses using multiple logistic regression. The precision of
the odds ratio (OR) was assessed with 95% confidence interval (CI). We present results
for three different statistical models separated by number of confounders included;
model 1 (age and sex), model 2 (age, sex and BMI), and model 3 (all other factors
were tested). In model 3 we excluded factors if they did not change OR when evaluating
each factor separately or when including several factors grouped together (i.e. self-reported
diabetes, stroke/heart infarction, and hypertension). Potential interaction between
two variables was evaluated by including the product of the variable in the logistic
regression analyses, and the interaction was tested using Wald χ2 statistics. Subjects with missing data of confounding factors (numbers reported in
Table 1) were included in the analysis to reduce the impact of possible bias.
To evaluate insomnia as a modifying factor on the association between hs-CRP and types
of headache, we repeated the multi-adjusted analyses in model 3 in those with and
without insomnia.
Analyses were performed with the IBM SPSS version 25 (SPSS, Chicago, Illinois, USA)