Appendices:

Appendix Document A1: Search strategy for Medline (1960 – November 2015)

Appendix Table A1: Detailed Risk of Bias Table

Appendix Figure A1: Funnel plot

Appendix Figure A2: Sensitivity analysis excluding cluster randomized trials

Appendix Figure A3: Subgroup analysis of studies by control intervention method

Appendix Figure A4: Subgroup analysis of studies by method of peer support

Appendix Figure A5: Subgroup analysis of studies by presence or absence of peer supervision

Appendix Figure A6: Meta-regression plot of effect of peer training hours on effect size

Appendix Figure A7: Meta-regression plot of effect of baseline Hemoglobin A1C on effect size

Appendix Figure A8: Meta-regression plot of effect of duration of observation on effect size

Appendix Document A1: Search strategy for Medline (1960 – November 2015)

01. exp Diabetes Mellitus/ or diabetes.ti.

02. Hemoglobin A, Glycosylated/ or (HbA1c or A1c or glycated h?emoglobin* or glycosylated h?emoglobin* or glyc?emic control or clinical

outcome*).mp.

03. Peer group/ or (peer*1 or peer health or peer support or peer led or peer educator* or peer coach* or peer counsel* or peer advis*).mp.

04. Patient Education as Topic/ or patient education.mp.

05. Self-Help Groups/ or self help group*

06. Social Support/ or (social support or promotora).mp.

This supplemental material has been supplied by the author and has not been edited by Annals of Family Medicine.

07. 4 or 5 or 6

08. peer*1.mp.

09. 7 and 8

10. 1 and 2 and 3 and 9

11. limit 10 to yr=”1960-Current”

12. randomized controlled trial.pt. or randomized controlled trials as topic/ or random*.ti,ab.

13. 11 and 12

Summary of Results

Ovid Medline and Ovid Medline In Process

62 MASR

68 RCT

Cochrane Central Register of Controlled Trials

62 RCT

Cochrane Database of Systematic Reviews

37 MASR

Scopus

16 MASR

44 RCT

CINAHL – Medline excluded

3 MASR

21 RCT

PsycINFO

1 MASR

This supplemental material has been supplied by the author and has not been edited by Annals of Family Medicine.

7 RCT

ProQuest Dissertations & Theses A&I, BioOne Abstracts & Indexes, Social Service Abstracts, and Sociological Abstracts

0 results

OCLC First Search

34 RCT

Author searches of Scopus, CINAHL, PsycINFO and Google Scholar

45 citations

Total 400 citations retrieved

Minus 195 duplicates removed

Leaves 205 unique citations

This supplemental material has been supplied by the author and has not been edited by Annals of Family Medicine.

Appendix Table A1: Detailed Risk of Bias Table

Study Name Adequate Sequence Generation

Allocation Concealment

Blinding of Participants, Personnel and Outcome

Assessors

Complete Outcome Data

Keyserling 2002 Stratified block randomization with randomly permuted blocks of size 3 and 6 from random numbers generated by a personal computer

Sealed envelopes No Yes: Dropout < 20%. Similar dropout in both groups. Follow-up data were analyzed according to group assignment

Lorig 2008 Yes: Random-number tables a

Yes: Random-number tables for each allocation that were used randomly a

No: “Participants could not be blinded.”

Yes: Dropout < 20%. Attrition reported. The “proportion of intervention non-completers compared with that of usual-care control non-completers was not statistically significant.”

Lorig 2009 Yes: Randomization using random-number tables

Yes: Random-number tables for each allocation that were used randomly a

No: “Participants could not be blinded.”

Yes: Dropout < 20%. Attrition reported. Proportion of intervention non-completers compared to usual-care control non-completers was not significant.

Dale 2009 Yes: Randomization protocol mentioned

Yes: Opaque sealed envelopes

Not reported Yes: Dropout < 20%.Attrition mentioned with some reasons.

Cade 2009 Yes?: Participants were randomized into intervention or the control arm but no information on sequence generation available.

Not reported Not reported Yes: Dropout < 20% from baseline measurements Attrition mentioned “Dropout between randomization and baseline measurements.” “At 12 months, clinical data (HbA1C) was available on 86 (77%) of the

This supplemental material has been supplied by the author and has not been edited by Annals of Family Medicine.

intervention participants and 118 (93%) control participants.”

Heisler 2010 Yes: “Random sequence generation and treatment group assignment were determined centrally just before the initial session.”

Yes: “Sequence was concealed until interventions were assigned.”

Yes: “Patients, research staff, and care managers were blinded to randomization results until after the baseline surveys and physiologic measures were completed. Data assessors remained blinded to group assignment throughout the study.”

Yes: Dropout < 20%. Attrition with reasons mentioned in flow diagram.

Philis-Tsamakas 2011

Yes: Blocked random assignment using a randomly generated numbers sequence

Allocation concealment was done a

No a Yes: Dropout > 20%Attrition mentioned with reasons of attrition. “Fifty-one (25%) participants were lost to follow-up (Project Dulce group, n = 35 [33.5%]; control group, n = 16 [15.5%]); however, at baseline, these participants did not differ significantly from those who completed at least one follow-up assessment on any demographic or outcome variable (p = 0.05).”

Smith 2011 Yes: Cluster randomization using minimization

Yesa: Allocation carried out independently of the research team Minimization was undertaken by an independent statistician

No Yes: 15% dropout Attrition with reasons mentioned.

Long 2012 Yes: Using the random-number generator function, each group assignment given a random number and put the ordered numbers in envelopes

Yes: “Envelopes were sealed, shuffled, and stacked, and the research assistant took the top envelope after consent was obtained to determine group assignment.”

Yes: “Study investigators were blinded to allocation and results until study completion.”

Yes: Dropout < 20% Attrition mentioned in flow diagram with some reasons.

This supplemental material has been supplied by the author and has not been edited by Annals of Family Medicine.

Yes indicates low risk of bias; Yes? Likely low risk of bias but no details on sequence generation available from article and author; No indicates high risk of bias; ?No is not reported or no response from author. a author response to email requests Selective outcome reporting risk was considered minimal since we only selected studies which reported hemoglobin A1C as the outcome. There were no significant other biases noted in the included studies.

Gagliardino 2013

Yes?: Random assignment mentioned in Methods but no information received on sequence generation

Not reported Not reported Attrition unclear

Siminerio 2013 Yes: Randomization was done by a computer program a

No a No a Yes: Dropout < 20% Attrition with reasons mentioned.

Thom 2013 Yes: “Assigned to usual care or peer-coaching study arm using randomly ordered opaque envelopes.”

Yes: Opaque envelopes.

No a Yes: Dropout < 20%.Attrition mentioned.

Chan 2014 Yes: Computer-generated, block randomization 1:1 assignment.

Yes: Opaque envelop opened by non-nursing staff not involved in research.

Not reported Yes: Dropout < 20% Attrition with reasons mentioned.

Simmons 2014 Yes: Clusters were randomized in blocks of four

Yes: Randomization by statistician who had no trial involvement after all clusters enrolled

Yes: Outcomes assessors and investigators were masked to arm allocation.

Yes: 23.9% dropout. Attrition mentioned.

Safford 2015 Yes: Random-number generator.

Yes: Randomization at the cluster level predated participant recruitment and was secure.

No: Participants and peer coaches were not blinded.

Dropout < 20% Attrition mentioned.

Ayala 2015 Yes: Conducted by study biostatistician and stratified by clinic.

?No: Not reported Not reported 27% dropout Attrition with reasons mentioned.

McGowan 2015 Yes: Block randomization, for each block random-number generator used

Yes: Anonymized study identification numbers were assigned to each subject.

No: Participants were not blinded.

34% dropout Attrition with reasons mentioned.

This supplemental material has been supplied by the author and has not been edited by Annals of Family Medicine.

-2.0 -1.5 -1.0 -0.5 0.0 0.5 1.0 1.5 2.0

0.0

0.1

0.2

0.3

0.4

Sta

nd

ard

Err

or

Std diff in means

Funnel Plot of Standard Error by Std diff in means

Appendix Figure A1: Funnel plot

This supplemental material has been supplied by the author and has not been edited by Annals of Family Medicine.

Study name Statistics for each study Std diff in means and 95% CI

Std diff Lower Upper in means limit limit p-Value

Keyserling 0.029 -0.311 0.369 0.866

Lorig 2008 0.240 0.047 0.433 0.015

Lorig 2009 -0.067 -0.279 0.144 0.534

Dale -0.306 -0.594 -0.017 0.038

Cade -0.159 -0.435 0.117 0.258

Heisler 0.377 0.124 0.630 0.004

Philis-Tsimakas 0.371 0.097 0.646 0.008

Long 0.661 0.202 1.120 0.005

Gagliardino 0.000 -0.279 0.279 1.000

Siminerio 0.108 -0.369 0.584 0.658

Thom 0.357 0.129 0.586 0.002

Chan 0.007 -0.150 0.163 0.934

Ayala 0.205 -0.009 0.419 0.061

McGowan 0.248 0.029 0.467 0.027

Pooled SMD 0.137 0.021 0.254 0.021

-1.00 -0.50 0.00 0.50 1.00

Favors control Favors peer support

I2 of 66.030%

p for heterogeneity <0.001

Appendix Figure A2: Sensitivity analysis excluding cluster randomized trials

This supplemental material has been supplied by the author and has not been edited by Annals of Family Medicine.

Supplement eFigure 3: Subgroup analysis of studies by control intervention

method

Group by control intervention

Study name Statistics for each study Std diff in means and 95% CI

Std diff Lower Upper in means limit limit p-Value

Education or Nurse care Keyserling 0.029 -0.311 0.369 0.866

Education or Nurse care Gagliardino 0.000 -0.279 0.279 1.000

Education or Nurse care Siminerio 0.108 -0.369 0.584 0.658

Education or Nurse care Chan 0.007 -0.150 0.163 0.934

Education or Nurse care Safford 0.181 -0.104 0.465 0.213

Education or Nurse care pooled effect 0.041 -0.072 0.153 0.477

Usual care Lorig 2008 0.240 0.047 0.433 0.015

Usual care Lorig 2009 -0.067 -0.279 0.144 0.534

Usual care Dale -0.306 -0.594 -0.017 0.038

Usual care Cade -0.159 -0.435 0.117 0.258

Usual care Heisler 0.377 0.124 0.630 0.004

Usual care Philis-Tsimakas 0.371 0.097 0.646 0.008

Usual care Smith 0.000 -0.208 0.208 1.000

Usual care Long 0.661 0.202 1.120 0.005

Usual care Thom 0.357 0.129 0.586 0.002

Usual care Simmons 0.029 -0.153 0.211 0.753

Usual care Ayala 0.205 -0.009 0.419 0.061

Usual care McGowan 0.248 0.029 0.467 0.027

Usual care pooled effect 0.147 0.021 0.273 0.023

-1.00 -0.50 0.00 0.50 1.00

Favors control Favors peer support

Appendix Figure A3: Subgroup analysis of studies by control intervention method

Group education or nurse care management control intervention: I2= 0.00,

p for heterogeneity 0.86

Usual care: I2= 70.53%, p for heterogeneity<0.001

This supplemental material has been supplied by the author and has not been edited by Annals of Family Medicine.

Supplement eFigure 4: Subgroup analysis of studies by method of peer support

Group by peer intervention method

Study name Statistics for each study Std diff in means and 95% CI

Std diff Lower Upper in means limit limit p-Value

group education Lorig 2008 0.240 0.047 0.433 0.015

group education Lorig 2009 -0.067 -0.279 0.144 0.534

group education Cade -0.159 -0.435 0.117 0.258

group education Philis-Tsimakas 0.371 0.097 0.646 0.008

group education Smith 0.000 -0.208 0.208 1.000

group education McGowan 0.248 0.029 0.467 0.027

group education pooled effect 0.106 -0.050 0.262 0.182

group+telephone Keyserling 0.029 -0.311 0.369 0.866

group+telephone Gagliardino 0.000 -0.279 0.279 1.000

group+telephone Simmons 0.029 -0.153 0.211 0.753

group+telephone Ayala 0.205 -0.009 0.419 0.061

Group + telephone pooled effect 0.076 -0.040 0.193 0.201

Telephone Dale -0.306 -0.594 -0.017 0.038

Telephone Heisler 0.377 0.124 0.630 0.004

Telephone Long 0.661 0.202 1.120 0.005

Telephone Siminerio 0.108 -0.369 0.584 0.658

Telephone Thom 0.357 0.129 0.586 0.002

Telephone Chan 0.007 -0.150 0.163 0.934

Telephone Safford 0.181 -0.104 0.465 0.213

Telephone pooled effect 0.179 -0.028 0.386 0.091

-1.00 -0.50 0.00 0.50 1.00

Favors control Favors peer support

Group education: I2 64.67%, p for heterogeneity 0.01

Group education + telephone support: I2 0.00%, p for heterogeneity 0.57

Telephone: I2 74.83%, p for heterogeneity 0.001

Supplement Figure A4: Subgroup analysis of studies by method of peer support

This supplemental material has been supplied by the author and has not been edited by Annals of Family Medicine.

Supplement eFigure 5: Subgroup analysis of studies by presence or absence of peer supervision

No supervision: I2 77.09%, p for heterogeneity 0.001

Supervised peers: I2 43.24%, p for heterogeneity 0.06

Group by peer supervision

Study name Statistics for each study Std diff in means and 95% CI

Std diff Lower Upper in means limit limit p-Value

No supervision Keyserling 0.029 -0.311 0.369 0.866

No supervision Dale -0.306 -0.594 -0.017 0.038

No supervision Cade -0.159 -0.435 0.117 0.258

No supervision Long 0.661 0.202 1.120 0.005

No supervision Thom 0.357 0.129 0.586 0.002

No supervision Simmons 0.029 -0.153 0.211 0.753

No supervision pooled effect 0.078 -0.157 0.313 0.515

Supervised Peers Lorig 2008 0.240 0.047 0.433 0.015

Supervised Peers Lorig 2009 -0.067 -0.279 0.144 0.534

Supervised Peers Heisler 0.377 0.124 0.630 0.004

Supervised Peers Philis-Tsimakas 0.371 0.097 0.646 0.008

Supervised Peers Smith 0.000 -0.208 0.208 1.000

Supervised Peers Gagliardino 0.000 -0.279 0.279 1.000

Supervised Peers Siminerio 0.108 -0.369 0.584 0.658

Supervised Peers Chan 0.007 -0.150 0.163 0.934

Supervised Peers Safford 0.181 -0.104 0.465 0.213

Supervised Peers Ayala 0.205 -0.009 0.419 0.061

Supervised Peers McGowan 0.248 0.029 0.467 0.027

Supervised Peers pooled effect 0.143 0.050 0.237 0.003

-1.00 -0.50 0.00 0.50 1.00

Favors control Favors peer support

No supervision: I2 77.09%, p for heterogeneity 0.001

Supervised peers: I2 43.24%, p for heterogeneity 0.06

Appendix Figure A5: Subgroup analysis of studies by presence or absence of peer supervision

This supplemental material has been supplied by the author and has not been edited by Annals of Family Medicine.

Regression of Std diff in means on Peer training hours

Peer training hours

-20.0 -10.0 0.0 10.0 20.0 30.0 40.0 50.0 60.0

Std

dif

f in

me

an

s0.80

0.60

0.40

0.20

0.00

-0.20

-0.40

-0.60

-0.80

Appendix Figure A6: Meta-regression plot of effect of peer training hours on effect size

Slope: 0.001

p = 0.8

This supplemental material has been supplied by the author and has not been edited by Annals of Family Medicine.

Regression of Std diff in means on baseline a1c

baseline a1c

5.5 6.0 6.5 7.0 7.5 8.0 8.5 9.0 9.5 10.0 10.5 11.0 11.5 12.0

Std

dif

f in

me

an

s

1.00

0.80

0.60

0.40

0.20

0.00

-0.20

-0.40

-0.60

-0.80

Appendix Figure A7: Meta-regression plot of effect of baseline Hemoglobin A1C on effect size

Slope 0.0657

P = 0.09

This supplemental material has been supplied by the author and has not been edited by Annals of Family Medicine.

Regression of Std diff in means on Months of observation after intervention completed

Months of observation after intervention completed

0.0 2.0 4.0 6.0 8.0 10.0 12.0

Std

dif

f in

me

an

s

0.80

0.60

0.40

0.20

0.00

-0.20

-0.40

-0.60

-0.80

Appendix Figure A8: Meta-regression plot of effect of duration of observation on effect size

Slope -0.004

P = 0.8

This supplemental material has been supplied by the author and has not been edited by Annals of Family Medicine.