Ethnic differences in the association between depression and chronic pain: cross sectional results from UK Biobank

This study utilises baseline cross-sectional data from UK Biobank (http://www.ukbiobank.ac.uk/). UK Biobank is a large cohort study of over 500,000 persons aged 40–70 years across
England, Scotland, and Wales. Individuals registered with the National Health Service
and living within a 25 mile radius of one of 22 UK Biobank assessment centres were
invited to participate 21]. Participation involved completion of touch-screen questionnaires and nurse-led interviews
to collect epidemiological data on demographic, health, environmental and lifestyle
factors. Questions on mood disorder were added part-way during recruitment and the
172,745 participants who completed the detailed questions on lifetime features of
mood disorders were eligible for inclusion in this study.

Ethnic groups

Participants were asked in the touch screen questionnaire “What is your ethnic group?”
Response options were: White, Mixed, Asian or Asian British, Black or Black British,
Chinese, and other ethnic group. Due to insufficient numbers in the other ethnic groups,
only those who considered themselves as White, Asian or Asian British (including Indian,
Pakistani, Bangladeshi and “any other Asian background”), and Black or Black British
(including Caribbean, African and “any other Black background”) were included in this
study; these groups will be referred to in this manuscript as White, Asian, and Black
ethnic groups.

Classification of chronic pain

Within UK Biobank, participants were asked whether they had experienced pain in the
past month that interfered with their usual activities, in any of the following sites:
headache, facial, neck or shoulder, back, stomach or abdominal, hip, and knee; and
an eighth option of “pain all over the body”. Multiple sites could be selected, unless
participants reported “pain all over the body”. For all sites selected a subsequent
question asked whether the pain had been present for more than three months. Chronic
pain was defined as pain that had persisted for more than three months. The presence
of chronic pain was classified if participants reported any site of chronic pain;
those reporting no sites of chronic pain were classified as being free of chronic
pain. The extent of chronic pain was calculated by summing the number of sites of
chronic pain reported and the following categories were defined: free of chronic pain;
chronic pain at one site; chronic pain at 2–3 sites; chronic pain at 4–7 sites; and
“pain all over the body”. This classification of number of pain sites has previously
been reported 22].

Classification of depression

The definition of probable lifetime depression history among UK Biobank participants
has been described in detail previously 23]. Using the touch screen questionnaire, participants answered questions regarding
their past history of mood disorder symptoms and related medical help-seeking, which
were similar to questions within the Structured Clinical Interview for DSM-IV Axis
I Disorders 24]. Probable major depressive disorder included participants that we classified as having
a lifetime history of recurrent severe, recurrent moderate or single episode depression;
the criteria for these definitions are outlined in Fig. 1. Current depressive symptom score was calculated from four individual questions that
assessed the frequency of the following depression symptoms “over the past 2 weeks”:
depressed mood, unenthusiasm/disinterest, tenseness/restlessness, and tiredness/lethargy.
Questions were answered on a four point scale: not at all; several days; more than
half the days; and nearly every day. Individual answers were summed to give a score
of 0–12 for each participant; with higher scores reflecting greater levels of current
depressive symptoms. This score was independent of the classification of lifetime
depression status described above and in Fig. 1.

Fig. 1. Detailed description of the classification of a probable lifetime history of depression
using available data fields in UK Biobank; a positive classification for probable
lifetime depression included all three depression categories: single, recurrent (moderate),
and recurrent (severe)

Potential confounding variables

Sociodemographic, lifestyle and morbidity factors are associated with both depression
and chronic pain. Alongside age and sex, Townsend Score, an area level measure of
deprivation based on the participant’s postcode of residence 25], was used to describe the sociodemographic profile of the study population. Quintiles
of the Townsend score were generated for the study population; quintile 1 represents
the least deprived and quintile 5 the most deprived areas. Current smoking status
(never, current, or former smokers) and frequency of alcohol consumption (daily/almost
daily, 3–4 times per week, 1–2 times per week, 1–3 times per month, special occasions
only and never) were self-reported. Body mass index (BMI) was calculated from height
and weight measurements taken at the assessment centre and categorised as underweight
(18.5 kg/m
²
), normal weight (18.5–24.9 kg/m
²
), overweight (25.0–29.9 kg/m
²
), and obese (?30.0 kg/m
²
) to describe the population but was treated as a continuous variable in the regression
analysis. A count of self-reported, non-cancer long-term conditions was generated
from the information provided during the nurse interview. Inclusion of conditions
was based on previous literature 16], author experience (FM) and a UK Biobank prevalence of ?0.1 %. Painful and psychiatric
conditions were excluded from the count for this study, resulting in 36 conditions
being included in the count.

Ethical approval

Electronic, informed consent was given in person at the assessment centre before participation
in the study. This study forms part of UK Biobank project 7155 and UK Biobank has
full ethical approval from NHS National Research Ethics Service (approval letter dated
17 June 2011, Ref 11/NW/0382).

Statistical analysis

Chronic pain, depression, and comorbid pain and depression prevalence are reported
for each ethnic group, along with sociodemographic, lifestyle and morbidity variables.
Chi squared tests for categorical variables and Kruskal Wallis tests for continuous
variables are used to assess differences between ethnic groups for each of the pain,
depression and potential confounding variables. To quantify the association between
depression and the presence of chronic pain, logistic regression models were used;
results are presented as odds ratios (OR) and 95 % confidence intervals. To quantify
the association between depression and the extent of chronic pain reported, multinomial
logistic regression models were used; results are presented as relative risk ratios
(RRR) and 95 % confidence intervals. The association between depression (independent
variable) and chronic pain (dependent variable) was quantified using a separate regression
model for each ethnic group as we were specifically interested in the relationship
between depression and chronic pain within ethnic groups. Interaction between depression
and ethnicity with both chronic pain outcomes was assessed and was indicated for some
sub-groups at approximately p?=?0.10. For both chronic pain outcomes (presence and extent of chronic pain), levels
of adjustment for the regression models were as follows: 1) univariate analysis (no
adjustment); 2) adjusted for sociodemographic variables – age, sex and quintiles of
Townsend score; 3) adjusted for sociodemographic and lifestyle variables – as in 2)
plus smoking status, frequency of alcohol consumption and BMI; 4) adjusted for sociodemographic,
lifestyle and morbidity count – as in 3) – plus number of long-term conditions reported.
In order to investigate the impact of current depressive symptom score as this may
impact on an individual’s current pain reports (and vice versa), a final adjustment
for current depressive symptom score is described in the text. All analysis was conducted
using Stata V13.0 26]. The number of participants included in each model varied according to the proportion
of missing data for each of the potential confounding variables; however, the completion
rate of confounding variables included in the regression models was at least 99.3 %,
with the exception of current depressive symptom score, which was completed by 92.0 %.