Phumipamorn 2008.

Methods	Randomized controlled trial
Participants	The study location was Krabi Province, Thailand 67 participants were randomized to the intervention group and 68 participants were randomized to the control group The inclusion criteria were Muslim diabetic patients, literate in Thai, over 18 years of age, and had A1C > 7% 3 months prior to the study
Interventions	Intervention: EXTRA PHARMACIST SERVICE Patients received their usual scheduled care by a primary care physician every 4 to 8 weeks. At each visit, fasting blood glucose was checked by the laboratory. Blood pressure and weight were recorded by a qualified nurse. A research pharmacist checked the pill count. Each patient then met a physician who assessed the patient and issued a repeat or modified prescription which the dispensing pharmacist filled and they also gave general advice on the medication uses. This was done over the dispensary counter on a routine basis. In addition to the usual care, each patient of the study group had a scheduled meeting with the research pharmacist for 4 consecutive visits at 2‐month intervals. Each visit was on the same date as the physician's appointment and in addition, to avoid a missed appointment, a health personnel co‐ordinator attached to the primary health care center in the area, where a patient was living, reminded the patients of the scheduled visit 3 days prior to each visit date. At each visit, the research pharmacist refilled prescriptions, discussed the uses of medication, and checked the pill count. Education on diabetes which included appropriate lifestyles and correct diet was also provided apart from a companion diabetic pamphlet which covered the diabetic complications, the targets of treating diabetes, lifestyle change, and antidiabetic medications Control: USUAL CARE The control group received their usual scheduled care by a primary care physician every 4 to 8 weeks. At each visit, fasting blood glucose was checked by the laboratory. Blood pressure and weight were recorded by a qualified nurse. A research pharmacist checked the pill count. Each patient then met a physician who assessed the patient and issued a repeat or modified prescription, which the dispensing pharmacist filled and they also gave general advice on the medication uses. This was done over the dispensary counter on a routine basis
Outcomes	The measure of adherence was per cent pill count. It was measured at baseline and then by a research pharmacist at each follow‐up appointment the patient had with their primary care physician, every 4 to 8 weeks. Patients were asked to bring their medications with them to appointments to facilitate pill counts The patient outcomes were changes in A1C level and changes in lipid profiles. Patients fasting blood glucose was checked at each appointment by the laboratory. A patient's A1C level was measured using NycoCard®, manufactured by Rapport Company Limited, USA. The hospital laboratory employed a DataPro® Random Access Clinical Analyser, from the U.S. Summit Oversea Company, to determine fasting glucose, total cholesterol, triglyceride and HDL‐C. No specific reagent for measuring LDL‐C was available at the setting studied. The albuminuria status level was identified using urine test strips. It should be noted here that the A1C and albuminuria measurements are not routinely performed in the setting studied and, as required by the study, these tests and results were performed specifically to collect these data
Notes	―
Risk of bias
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Unclear risk	Random numbers were used. "They were selected randomly by drawing numbers from a container that included "1" for the study group (N = 67) and "2" for the control group (N = 68)." (pg 32). "The randomization method used was apparently not optimal as more women were drawn into the study group and more patients in the control group took single antidiabetic agent and tended to have higher pill count.The randomization method used was apparently not optimal as more women were drawn into the study group and more patients in the control group took single antidiabetic agent and tended to have higher pill count." (pg 36)
Allocation concealment (selection bias)	Unclear risk	It does not specify that an opaque container was used. "They were selected randomly by drawing numbers from a container that included "1" for the study group (N = 67) and "2" for the control group (N = 68)." (pg 32)
Selective reporting (reporting bias)	Unclear risk	No protocol available; although it appears that everything was reported it is difficult to determine this without a protocol
Other bias	Unclear risk	As noted by the authors, "...some contaminations between the control and the study groups in the single site as in this study, as well as their close proximity of living in confined Muslim community could have diluted the effect of the intervention. The Hawthorn effect could not also be excluded. In addition, there might also be an increase in the overall patient care by other care providers due to a lack of blinding in the pharmacist's activities." ‐ (pg 34 "This study has a number of limitations. Expectations and needs of this specific group of patients were not considered while planning for the interventions. The randomization method used was apparently not optimal as more women were drawn into the study group and more patients in the control group took single antidiabetic agent and tended to have higher pill count. Net A1C reduction set in the study may be too large to be obtained with the interventions. We measured two points of A1C, i.e., at the baseline and at the end, and as a consequence there was a lack of observation that could have possibly led to a more accurate trend of A1C change. The pill count method to gauge medication adherence may be inadequate. Though blinding to the randomization groups, the physicians and nurses were aware of pharmacist carrying out the study thus increases in overall care might be possible. Our intention was not to affect or evaluate the process of physician care for diabetes but it would be useful to measure changes in their diabetes management that may also have an impact on the outcome measures. The pharmacist's extra input involved combined strategies. Thus, it was impossible to label which component was associated with the improvements although multifaceted interventions were proved to enhance patient outcomes in diabetes care. It should be borne in mind that the management of diabetes mellitus in real practice is life‐long and complex in nature. The results may not be conclusively generalisable, but they imply that similar strategies employed for general diabetes patients might have favourable effects on cardiovascular risks in this specific group." (pg 36)
Blinding of outcome assessment (detection bias) Adherence measure	High risk	(PRIMARY) PILL COUNT ‐ Pharmacist was not blinded. As noted by the author, "In addition, there might also be an increase in the overall patient care by other care providers due to a lack of blinding in the pharmacist's activities" (pg 34)
Blinding of outcome assessment (detection bias) Patient outcome	Low risk	(PRIMARY) A1C ‐ This is an objective laboratory test
Blinding of participants (performance bias) Adherence measure	High risk	(PRIMARY) PILL COUNT ‐ Participants were not blinded
Blinding of participants (performance bias) Patient outcome	Low risk	(PRIMARY) A1C ‐ This is an objective laboratory test
Blinding of personnel (performance bias) Adherence measure	Unclear risk	(PRIMARY) PILL COUNT ‐ No information on blinding given. There is insufficient information to permit judgment of 'Low risk' or 'High risk'
Blinding of personnel (performance bias) Patient outcome	Low risk	(PRIMARY) A1C ‐ This is an objective laboratory test.
Incomplete outcome data (attrition bias) Adherence measure	Unclear risk	(PRIMARY) PILL COUNT ‐ Few dropouts, but they were not evenly spread across the 2 groups
Incomplete outcome data (attrition bias) Patient outcome	Unclear risk	(PRIMARY) A1C ‐ Only 5 dropouts but they were unbalanced across the groups