This study was conducted at the Aga Khan University Hospital (AKUH) in Karachi, Pakistan. AKUH is a private, university hospital, focusing on the delivery of the highest quality
health care in Pakistan for over 30 years since 1985.
This was a hospital based single blinded, block randomized controlled trial with one
intervention and one control arm. Participants were recruited from all endocrine clinics
at AKUH. Participants aged 30 to 70 years old who had been diagnosed with Type II
diabetes in the last 2 to 15 years, with an HbA1c of more than 7% and who were on
oral hypoglycemic medication were included in the study. Moreover, participants who
owned mobile phones, and who could read and respond to text messages written in Urdu
were included in the study. Those participants who had serious chronic conditions
like diabetic ketosis, diabetic foot, nephropathy, neuropathy, retinopathy, chronic
renal failure and liver cirrhosis, pregnant mothers, those who were involved in any
other study and those who did not give informed consent were excluded from the study.
Recruitment of the study participants started on the 21st June 2012 and ended on the
12 June 2015. The overall follow up period for each participant was 3 months.
Potential participants were identified from endocrine clinics at the AKUH by obtaining
information about the status of the disease from both hospital online records and
medical record files. WHO criterion was used to diagnose diabetes; which is fasting
blood glucose ≥ 126mg/dl or 7mmol/l and 2 hours random sugar ≥200mg/dl or 11.1mmol/l.
Recent HbA1c (%) level and fasting blood sugar (mg/dl) were recorded from the electronic
laboratory records. Each eligible participant was given detailed information about
the study. A written informed consent was taken from all the willing participants.
Consent form was composed of both English and Urdu. Data collectors were provided
proper training for administration of the questionnaires. Questionnaires were explained
in detail to the data collectors and description of how to collect data for each variable
was formulated in a manual of operations. At baseline an interview was conducted with
each participant to obtain information about their socio-demographic status, social
and psychological support and dietary intake using validated questionnaires. Anthropometric
measurements including height and weight were measured in order to assess the body
mass index by using weight and height scales respectively. Waist and hip circumference
were recorded using measuring tapes to assess the waist to hip ratio. Dietary intake
information was assessed using semi quantitative validated food frequency questionnaires
(FFQ) both at baseline and after the follow-up period of three months. This FFQ contained
both traditional and mixed food dishes eaten in Karachi. Mobile phone numbers of each
participant were recorded. All participants were provided with dietary guidelines
and explained in detail, so it would help maintain their blood glucose level. Randomization
was performed by block randomization strategy with an uneven varying block size of
2 and 4. Alphabet “A” was assigned to intervention arm while “B” was considered as
control arm. There were only two possible combinations of A&B for block size 2, while
for block size 4, there were 6 possible combinations. These were generated by the
primary investigator. Each combination was put in an opaque envelop, which was selected
for randomization by a third person. After randomization, identification numbers (different
for intervention and control arm) were assigned to each participant.
Dietary text message reminders were sent to participants three times a week using
the Frontline Short Message Service (SMS) software. These were automated messages,
similar for all the participants in the intervention arm, and were sent at the same
time. Each time, different templates of messages were used. These dietary text messages
as well as dietary guidelines were developed based on suggestions provided by the
“American Diabetes Association Complete Guide to Diabetes”(13), as well as with advice
given by a local expert nutritionist and other investigators of the study. Messages
were based on 8 food components; vegetables, fruits, cereals, starch and whole grain
foods, nuts and legumes, dairy products, fish, poultry and eggs, meat products and
sweets and drinks.
Dietary adherence was the main outcome of interest and was measured using the FFQ.
Frequency of intake of each food item falling under the 8 food components was recorded
in one of the 9 categories which were: Never, ≤ once/month, 1-3 times/month, 1 time
/week, 2-4 times/ week, 5-6 times/ week, 1 time/ day, 2-3 times/day, 4-5 times/ day
and ≥ 6 times/day. Mean daily intake of each component was compared with the daily
intake recommended in the dietary guidelines. This was done by converting the frequency
of intake into “intake per day” and then multiplying it with the amount of food consumed
at one time. After adjusting for the amount of intake, mean daily intake for each
food component was calculated in order to consider the number of food items in each
food component. Diet quality scores were assigned to each component (range was from
0 to 2). If the consumption of the food component was exactly equal to the dietary
guidelines, it was given a score of “2”. If the consumption was greater or lesser
than the dietary guidelines, a score of “1 and 0” was given based on the food group’s
effect on the level of glucose management. If the food is beneficial in terms of management
of glucose level in the body and consumed greater than the prescribed amount, we assigned
1 and if consumed less than the mentioned amount we assigned 0. However, if the food
could increase the blood glucose level and could be unhealthy for diabetic patients
and the participant consumed more than the amount assigned, we gave 0 score and for
less consumption we assigned 1.
In addition, a fortnightly two-item questionnaire was sent to assess the dietary intake
of specific food components in both the groups of participants i.e. fruits and vegetables.
Two questions were designed in a text message. It was stated as: in the last 24 hours
how many fruits do you have? If yes write the number of fruits if no, please write
zero. Same question was asked for the vegetables as well. Participants responded back
to questions via text message. For this response credit was transferred through easy
loads to each participant. This approach also helped in the reduction of loss to follow-up
in the control arm as well as give us an idea of compliance to text messages.
After 3 months of the intervention period, another FFQ was administered to measure
dietary adherence from both groups via telephone calls. The outcome assessor was blinded
to the randomization status of the participants i.e. whether they belonged to the
intervention arm or the control arm. Full trial protocol is available on ANZCTR website.
Text message reminders and dietary guidelines are attached in the supplementary material.
A total of 112 subjects in each arm were required to achieve 80% power for the parent
study with 5% level of significance, range of compliance score of 0-16 and expected
mean difference in the dietary compliance score of 1.5 between the two groups for
the complete duration of follow-up. After considering 10% loss to follow-up an additional
sample of 12 participants in each arm was decided to recruit in the study.
Total scores for all FFQ were then calculated by adding diet quality scores of all
8-food components. The overall possible scoring for FFQ ranged from 0 to 16.
Means and standard deviations were calculated for normally distributed continuous
data, while median and inter quartile range were computed for skewed data. Frequencies
and proportion were computed for categorical variables. We were unable to conduct
the planed analysis because we ended up with count data for compliance score and because
of less variability in the dietary score. Hence Poisson regression was conducted to
see the association between dietary text message reminders and the adherence score
for dietary guidelines while adjusting for different variables such as age, gender,
body mass index (BMI), education and household income. We also checked the interaction
between different variables while building the final model. SPSS version 19 was used
to conduct the analysis.