Lester 2010.

Methods	Randomized controlled trial
Participants	The study location was The University of Nairobi Pumwani Clinic, Coptic Hope Center for Infectious Diseases, The Kajiado Clinic, Kenya 273 participants were randomized to the intervention group and 265 participants were randomized to the control group The inclusion criteria were over 18 years old, initiating ART for the first time, and able to access a mobile phone on a near‐daily basis and communicate via short message service (SMS). People who did not own mobile phones were eligible if they had shared access (with corroborative agreement by the phone owner), and illiterate patients were eligible if assisted by a literate partner. Participants used existing mobile phone services; phones and network airtime credit were not provided
Interventions	Intervention: SMS GROUP All intervention group participants were given brief training about the use of SMS intervention by the study clinicians. On Monday morning of each week, the site nurse or clinical officer sent a text message via SMS to patients in the intervention group to inquire about their status and thus to remind them about the availability of phone‐based support. Regular, structured mobile phone communication between healthcare workers and patients could improve patient outcomes by both reminding patients to take their ART and by providing support to the patients. Patients in the intervention group were instructed to respond within 48 hours that either they were doing well or that they had a problem. The clinician then called patients who said they had a problem or who failed to respond within 2 days. Participants were instructed that healthcare workers were available to respond during clinic hours only Control: USUAL CARE Control group is standard care. Standard care at Kijaiado study site included providing one counseling session at ART initiation and at the 2 Nairobi sites included providing 2 counseling sessions before and one session 1 month after ART initiation. Disclosure of HIV status, pairing up with a treatment adherence partner, and participation in support groups was encouraged but not insisted upon. Additional brief counseling was provided at each site during dispensation of the drugs in the clinic or pharmacy. Patients did not receive weekly SMS
Outcomes	The measures of adherence were self reported adherence, measured at 6 and 12 months followed. Participants were asked how many pills they missed in the past 30 days, and classified as adherent if they reported that they had taken more than 95% of the provided pills The patient outcome was suppression of plasma HIV‐1 viral load. Participants were classed as virologically suppressed if their plasma HIV‐1 RNA load at their 12‐month visit was 400 copies per ml or less. Patients who did not achieve this outcome were classed as virologic failures. Plasma was taken from patients at the 12‐month visit for assessment of HIV‐1 RNA viral load (Amplicor, Roche Diagnostics, Mannheim, Germany) and was stored and analyzed at a later time point in batches. Laboratory assays (CD4 count and HIV‐1 RNA load) were all done at a central laboratory (University of Nairobi Institutes for Tropical and Infectious Diseases). CD4 count testing (FACScan, Becton Dickinson, Sunnyvale, CA, USA) was part of routine care at the urban sites and was provided for study purposes at the rural site
Notes	―
Risk of bias
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Low risk	"Patients were randomised (1:1) by simple randomisation with a random number generating program to a mobile phone short message service (SMS) intervention or standard care." (pg 1838. "WelTel Kenya1 was an individually randomised, parallel, multisite controlled trial. Patients were randomly assigned (1:1) by simple randomisation15 to the SMS intervention or to standard care (control group). A project statistician generated the randomisation numbers with a random number generating program." (pg 1839)
Allocation concealment (selection bias)	Low risk	"Written allocation of assignment was sealed in individual opaque envelopes marked with study identification numbers, which were distributed to all three study clinics." (pg 1839)
Selective reporting (reporting bias)	Unclear risk	They do not report on all of the predefined secondary outcomes, but they state that "Other predefined secondary endpoints, including quality of life and social and economic outcomes, will be reported separately." (pg 1840). CD4 cell count is reported for baseline but no results from follow‐up is reported. "CD4 count testing (FACScan, Becton Dickinson, Sunnyvale, CA, USA) was part of routine care at the urban sites and was provided for study purposes at the rural site." (pg 1840)
Other bias	Unclear risk	Author's note: risk of bias is unclear because contamination would have to be quite substantial for it to pose a high risk of bias
Blinding of outcome assessment (detection bias) Adherence measure	Unclear risk	(PRIMARY) SELF REPORT ‐ INTERVIEW ‐ Not blinded (pg 1839), but probably would not have influenced patient responses
Blinding of outcome assessment (detection bias) Patient outcome	Low risk	(PRIMARY) VIRAL LOAD ‐ Objective outcome. Also, all analysis "...were done by investigators masked to treatment allocation."
Blinding of participants (performance bias) Adherence measure	High risk	(PRIMARY) SELF REPORT ‐ INTERVIEW ‐ The authors indicate this measure may have been biased: (pg 1842) The primary analysis classed all‐cause attrition as treatment failures. Thus, the higher follow‐up rates and lower mortality reported in the intervention group contributed to the positive intervention effect for the primary outcomes. When only available data were included in the complete‐case analyses, a significant reduction was preserved in viral suppression but not in self reported adherence. This could be because of a recall or social desirability bias in self reporting adherence among those who were followed up. Alternatively, patients might have been less likely to respond to follow‐up if they had not adhered to ART. Nonetheless, we chose an intention‐to‐treat analysis for the primary outcome because contributions of loss to follow‐up are important indicators for the durable implementation of ART programs
Blinding of participants (performance bias) Patient outcome	Low risk	(PRIMARY) VIRAL LOAD ‐ Study participants were not masked to treatment; but this is an measure of outcome. Thus the outcome measurement is not likely to be influenced by lack of blinding
Blinding of personnel (performance bias) Adherence measure	Low risk	(PRIMARY) SELF REPORT ‐ INTERVIEW ‐ Not blinded (pg 1839), but probably would not have influenced patient responses
Blinding of personnel (performance bias) Patient outcome	Low risk	(PRIMARY) VIRAL LOAD ‐ Objective outcome. Also, all analysis "...were done by investigators masked to treatment allocation"
Incomplete outcome data (attrition bias) Adherence measure	Low risk	(PRIMARY) SELF REPORT ‐ INTERVIEW ‐ Total attrition (missing) from SMS group is 53 (19%) and standard care is 61 (23%). The reasons are loss to follow‐up, mortality, withdrawal, transfer out. Refer to Table 2. Missing outcome data balanced in numbers across intervention groups, with similar reasons for missing data across groups
Incomplete outcome data (attrition bias) Patient outcome	Low risk	(PRIMARY) VIRAL LOAD ‐ Total attrition (missing) from SMS group is 53 (19%) and standard care is 61 (23%). The reasons are loss to follow‐up, mortality, withdrawal, transfer out. Refer to Table 2. Missing outcome data balanced in numbers across intervention groups, with similar reasons for missing data across groups