Setting and Participants
Twenty consecutive staff physician volunteers were recruited from the doctors' lounge of a large urban teaching hospital during the first week of May 2008 following a hospital wide poster campaign advertising study recruitment location and timing. The physicians selected two typical, similar work days to be scheduled as the baseline and intervention study periods during May and/or June 2008. All data were collected on site at the hospital. Ethics approval was obtained from the Conjoint Ethics Review Board of the University of Calgary. Written consent was obtained from participants.
Study Design
This prospective study compared physicians' nutritional intake and cognitive function during work hours on two separate work days, a baseline day and an intervention day. A before and after study design was chosen rather than assigning participants to intervention or control days in random order, given the possibility that physicians assigned to first receive the intervention may be influenced to alter their typical nutritional habits. On the baseline day, the physicians followed their usual eating and drinking habits. On the intervention day, they were fed nutritious meals, snacks and fluids at scheduled intervals. Participants chose two typical and similar work days (in terms of workload, hours, sleep patterns, and other professional and personal commitments) within a two week period to serve as baseline and intervention days. Most physicians chose daytime work hours as the study period (17/20) while three chose evening and overnight work hours.
Intervention
The intervention, that of ensuring that physicians consumed nutrients and fluids at regular intervals throughout their work day, was designed based on previous research where physicians and other health care professionals described barriers to achieving adequate nutritional intake during work hours [[1–4], unpublished data, Lemaire, Wallace, Dinsmore, Roberts]. The intervention had four key elements: providing healthy nutrition choices; enforcing nutrition breaks; maximizing ease of accessibility; and offering cost free nutrition. Nutrition provided during the study period on the intervention day was based on the recommendations of Canada's Food Guide [19] and a projected 24 hour total intake of 30.8 kcal/kg body weight with 15 percent of energy from fat, 15% from protein, and 70% from carbohydrate. On average, food and beverages were provided in six small meals, with selections offered based upon participant preference, and ease of storage, delivery and consumption in the hospital setting. This varied according to the number of hours worked by each participant and their individual nutritional choices. At each scheduled nutrition break, the research team contacted participants through hospital paging. Ready-to-consume and cost free nutrition was either waiting for physicians at the centrally located doctors' lounge or was brought to their practice location.
Outcomes and Measurements
The primary outcome was cognition. Secondary outcomes were blood glucose levels and "hypoglycemic" nutrition-related symptoms. Baseline demographic characteristics were recorded at study enrolment. Fluid and nutrient intake and physical activity were measured on both days. At the beginning of each day, participants were weighed and fitted with an activity and heart rate monitor. At that time and at approximately two hour intervals on both days, measures of cognitive function, capillary blood glucose, "hypoglycemic" nutrition-related symptoms, food and fluid intake, and volume of urine excreted over the previous two hours were captured. Participants were weighed again at the end of each day. On the baseline day, the physicians maintained their usual eating and drinking habits. On the intervention day, the physicians reported to the study center fasting, and all nutrition for the day was delivered to the physician and recorded. The participants were blinded to their glucose and cognitive function test results at the time of testing.
Cognition
Cognition was measured using Brain Checkers software, Version 3.01 (Behavioural Neuroscience Systems LLC, Springfield MO) run on Palm Tungsten E2, (Palm Inc. Milpitas, CA). Two software programs were used. The simple reaction test was designed to measure the speed of motor response to a visual cue with repeated testing over thirty seconds. The complex reaction test, a choice reaction time and continuous performance task, was designed to measure running memory, attention and visual information processing with repeated testing over two minutes. This task requires the subject to indicate whether the current number (1 through 9 appearing randomly on a screen) matches the previously displayed number (with random time delay between the two) by tapping on the appropriate text box (labelled "same" or "different") located below the number. For both tests, the reaction times of each unique response as well as the mean reaction time for the session were recorded for each participant. Accuracy was documented in terms of percent correct responses, lapses (the subject did not respond to the stimulus) and impulses (the subject anticipated and acted before the prompt). A Tput statistic was calculated that captures the correct responses per minute of time available to respond. It represents a combination of speed and accuracy with a higher Tput statistic indicating a superior performance. Based on the manufacturer's recommendation, subjects completed three practice tests prior to baseline data collection. This approach eliminates any learning effect during the study period and prevents learning effects from confounding actual study measurements [20–22].
Glucose and "hypoglycemic" nutrition-related symptoms
Capillary blood glucose samples were collected from participants' fingertip and analyzed immediately using the Precision Xtra Blood Glucose Monitoring System (glucose measured in millimoles per liter). Participants were asked to report from a checklist of "hypoglycemic" nutrition-related symptoms, including those produced by falling glucose and counterregulatory hormones and by reduced brain glucose. The seventeen symptoms covered manifestations of adrenergic responses (sweating, sensation of warmth, anxiety, tremor or tremulousness, palpitations and tachycardia), glucagon responses (hunger, nausea), and neuroglycopenic responses (fatigue, dizziness, headache, visual disturbance, drowsiness, difficulty speaking, inability to concentrate, abnormal behavior, loss of memory, and confusion) [23, 24]. The checklist response data were collapsed to a binary yes/no variable for the presence or absence of each symptom.
Hydration and nutrients
Body mass was measured using SR Model SR241 scales (accuracy = 0.2% ± 1 digit, resolution = 0.1 kg, SR Instruments, New Jersey). The measure of weight, performed by either of the two research assistants at the beginning and end of each study period, was standardized by using a single digital scale at the same location and ensuring participants' equivalent post urinary void state and clothing status (e.g. shoes off, pockets empty, pagers removed). Volume of fluid consumed and urine voided were quantified. Dietary analyses were performed using individual physicians' recorded diet history (instructions on how to record all food and drink consumption accurately were provided). Two-hour diet recall was also taken at each blood glucose sampling in order to enhance the validity of the dietary record. Only nutritional intake during the study period was analyzed using Diet Analysis+, Canadian version 4.0 (Wadsworth/Thompson Learning, Scarborough Ontario). Nutritional requirements were based upon the Dietary Reference Intakes (DRI 2002) [25], which reflect the current state of scientific knowledge.
Activity, patient load, stress and wellbeing
A triaxial accelerometer that records acceleration in three planes (Actiheart system, Mini Mitter Co. Inc, Bend OR) recorded activity level and heart rate simultaneously every fifteen seconds. Physicians were asked to rate both days on scales of 0 (low) to 10 (high) for workload, stress and general well being.
Statistical Analysis
After determining the appropriateness of parametric analytical methods, the statistical significance of mean differences in blood glucose levels and cognitive test scores were calculated using a generalized estimating equation to take into account the repeated measurements taken during each study day; change in body mass, and fluid and nutrient intake on baseline and intervention days were assessed for normalcy and means were compared using paired two sided t-tests; where the assumption of normalcy was not met, results were presented as medians plus interquartile range, and compared using a Wilcoxon signed rank test. Variability in glucose values was calculated using the coefficient of variation (CV), which describes variability relative to the mean [CV = (standard deviation/mean)*100%]. Analysis of variance (ANOVA) was used to assess within-day differences in mean cognitive test scores across the sampling times. A Fisher's exact test was used to compare proportion of physicians reporting "hypoglycemic" nutrition-related symptoms on baseline and intervention days.
This study was originally conceived as a pilot study for preliminary testing of a nutrition based intervention, and for determination of multiple physiological and nutritional measurements in twenty working physicians. Given this, there were no a priori sample size considerations. However, based on mean glucose and cognition (Tput) values obtained, and corresponding standard deviations, we determined post-hoc that we had 96% power to detect a difference of 0.28 mmol/L in glucose values, and 97% power to detect a difference of 5 in Tput scores from the complex cognition test for the intervention day versus the baseline day.
All statistical analyses were performed using Stata 10 (StataCorp LP, College Station, Texas USA).