Methods

Desing trial: parallel (2 groups).
Study phase: not described.
Baseline observation period: no (Dr Graham Serjeant, interviewed by one author of this review (JKM), on February 4, 2011).
Follow‐up period: 8 weeks.
Randomisation unit: patient.
Unit of analysis: ulcers.
Intention to treat: unclear.

Participants

Total: 30.
Antibiotic therapy: 16.
Placebo: 14.
Withdrawals: 7% (2/30) (1 from each group).

Reported leg ulcers at the start of the trial: 40 (data supplied by Dr Graham Serjeant, Jamaica, February 4, 2011).

Age: median (range). Antibiotic therapy group: 29 years (14 ‐ 44 years); control group: 25 years (17 ‐ 49 years).

Number of ulcers: 40 (antibiotic therapy: 20; placebo: 20).

Duration (median):
Antibiotic therapy group: 3.6 years (range: 0.3 to 15 years).
Control group: 3.5 years (range: 0.3 to 12 years).

Mean baseline ulcer area (cm²) ± SD. Antibiotic therapy group: 20.0 (14.9); placebo: 20.8 (19.0).

Gender (male:female). Antibiotic therapy group: 4:9; control: 6:9.

Inclusion criteria:
1. Hb SS;
2. leg ulceration of at least 3 months duration;
3. bacterial swabs revealed at least one of the skin pathogen;
4. to attend at 2‐weekly intervals;
5. to agree to the trial protocol;
6. multiple ulcers: yes, all used (data supplied by Dr Graham Serjeant, Jamaica, February 4, 2011).
7. new ulcers discussed: not information (Data supplied by Dr Graham Serjeant, Jamaica, February 4, 2011).

Exclusion criteria:
1. large leg ulcers;
2. chronic renal failure.

Interventions

Antimicrobial therapy (aerosol preparation, twice daily):
neomycin, bacitracin and polymyxin B.

Control (aerosol dispenser, twice daily).
Placebo: sterile normal saline with 1:1000 red food colouring.

Co‐intervention: disinfectant (Eusol).
Rest with elevation of the affected leg(s).

Systematic antibiotics: not allowed.
Oral zinc sulphate was allowed.

Outcomes

Primary: pain.
Likely secondary outcomes (these end points were not described at 'methods section'): healing rate; bacteriological investigation.

According with Dr Graham Serjeant, interviewed by one author of this review (JKM), 'ulcer area‐change' was the primary outcome (February 4, 2011).

Notes

1. This trial was described as "randomised controlled crossover trial" at Proceedings of the Commonwealth Caribean medical Research Council (31st Scientific Meeting, 1986, April 16‐19, Port Spain, Trinidad and Tobago.
2. Country: University Hospital of the West Indies, Kingston, Jamaica.
3. Study time: 10 June to 9 September 1993.
4. Sample size estimation a priori: not stated.
5. 3M Company Incorporated donated Rikospray^®, an aerosol preparation of neomycin, bacitracin, and polymyxin B.

Risk of bias

Bias

Authors' judgement

Support for judgement

Random sequence generation (selection bias)

Unclear risk

Quote: "patients were assigned, using size 4 block randomization" (page 847).
Comments: there is an imbalance in two characteristics baseline: zinc therapy was only received by control patients and marked pain (14 placebo patients vs 8 patients treatment).

Insufficient information about the sequence generation process to permit judgement of ‘low risk’ or ‘high risk’.

Allocation concealment (selection bias)

Unclear risk

Insufficient information to permit judgement of ‘low risk’ or ‘high risk’.

Blinding of participants and personnel (performance bias)
All outcomes

Unclear risk

Insufficient information to permit judgement of ‘low risk’ or ‘high risk’

Blinding of outcome assessment (detection bias)
All outcomes

Unclear risk

Insufficient information to permit judgement of ‘low risk’ or ‘high risk’.

Incomplete outcome data (attrition bias)
All outcomes

Unclear risk

Insufficient reporting of attrition and exclusions to permit judgement of ‘low risk’ or ‘high risk’.

Selective reporting (reporting bias)

High risk

One or more clinically relevant and reasonably expected outcomes were not reported on; data on these outcomes were likely to have been recorded (i.e. safety).
According with Dr Graham Serjeant, interviewed by one author of this review (JKM), 'ulcer area‐change' was the primary outcome (February 4, 2011).

Other bias

High risk

Ascertaiment bias and bias in the presentation of data (Appendix 2).
Table of characteristics of patients at entry study only shows 28 patients.
There is inconsistency relating to the study design and the healing rate end point.
The inconsistency between 'Randomisation unit: patient' and 'Unit of analysis: ulcers' could generate the design and confusion bias (Appendix 2).

Methods

Desing trial: parallel (3 groups).
Study phase: not described.
Baseline observation period: 3 months.
Follow‐up period: 12 weeks.
Randomisation unit: ulcers.
Unit of analysis: ulcers.
Intention to treat: unclear.

Participants

Sample size: 32 participants.
Randomised leg ulcers: 49 (Solcoseryl^® 14; hydrocolloid dressing 14; placebo 21).
Loss patients reported: 14 (Solcoseryl^® 4; hydrocolloid dressing: 8; placebo: 2).
(Dr. Graham Serjeant, interviewed by one author of this review (JKM), on February 4, 2011).

16 (Solcoseryl^®: 3); hydrocolloid dressing: 8); (Placebo: 5)
Loss (%): 50%.

Age: Mean (± SD): Solcoseryl^®: 27.0 years (8.2); hydrocolloid dressing: 35.5 years (9.8); placebo: 30.7(8.2).

Gender (males): Solcoseryl^®: 93%; hydrocolloid dressing: 50%; placebo: 71%.

Inclusion criteria:
1. Patients with Hb SS.
2. Age (years): ≥ 15
3. Leg ulcer duration: ≥ 16 months.
4. Ulcer size: ≥3 cm.
5. Patients attending for a 3‐month baseline observation period.

Exclusion criteria: not described.

Interventions

1. Solcoseryl^® (de‐proteinized extract of calf blood) supplied in tubes as jelIy or ointment, twice daily after cleaning with Eusol, covered with a gauze dressing and supported by an Elastoweb bandage.

2. Hydrocolloid dressing (hydroactive dressing) supplied as sheets of occlusive dressing and as granules. and covered with gauze and supported by an Elastoweb bandage. The dressing was replaced at weekly intervals or more often if indicated.

Control: Twice‐daily cleaning with a mild antiseptic agent (Eusol), wet dressing and a firm supportive elastic bandage.

Outcomes

Healing ulcer.
According with Dr Graham Serjeant, interviewed by one author of this review (JKM), 'ulcer area‐change' was the primary outcome (February 4, 2011).

Notes

Centre and Country: Sickle‐cell Clinic of the University Hospital of the West Indies, Kingston, Jamaica.
Solcoseryl^® was supplied by Solco Basle Ltd.
Hydrocolloid dressing by Convatec.
Sample size estimation a priori: no described.

Risk of bias

Bias

Authors' judgement

Support for judgement

Random sequence generation (selection bias)

Unclear risk

Quote: "Unilateral ulcers were randomized into one of three treatment schedules in blocks of three" and "In bilateral ulcers, one leg was randomized to one of the two treatment agents and the other leg to control therapy" (page 121).

Insufficient information about the sequence generation process to permit judgement of ‘low risk’ or ‘high risk’.

Allocation concealment (selection bias)

Unclear risk

Insufficient information to permit judgement of ‘low risk’ or ‘high risk’.

Blinding of participants and personnel (performance bias)
All outcomes

Unclear risk

Insufficient information to permit judgement of ‘low risk’ or ‘high risk’

Blinding of outcome assessment (detection bias)
All outcomes

Unclear risk

Insufficient information to permit judgement of ‘low risk’ or ‘high risk’

Incomplete outcome data (attrition bias)
All outcomes

High risk

Dropouts: 50% (16/32) (Placebo 24% (5/21); Solcoseryl^® 21.4% (3/14); hydrocolloid dressing 57.1% (8/14).
Reasons: none described.

Imbalance between hydrocolloid dressing group and the other groups > 10%.

Selective reporting (reporting bias)

High risk

1 or more clinically relevant and reasonably expected outcomes were not reported on; data on these outcomes were likely to have been recorded (i.e. safety).

Other bias

High risk

Ascertaiment and design bias and bias in the presentation of data (Appendix 2).

Comments: there is imbalance in sex and age.

Methods

Design trial: parallel (2 groups).
Study phase: II.
Follow‐up period: 12 weeks.
Randomisation unit: patients.
Unit of analysis: ulcers
Intention‐to‐treat: unclear.

Participants

Sample size: 26 participants
Randomised leg ulcers: 62 (arginine butyrate plus standard local care: 37; standard local care: 25).

Age: median (range): arginine butyrate plus standard local care: 36.6 (21–60); standard local care: 34.7 (20–57).

Gender (female): arginine butyrate plus standard local care: 57%; standard local care: 58%.

Mean baseline ulcer area (cm²) ± SE: arginine butyrate plus standard local care: 25.7 cm² (8.1); standard local care: 50.6 cm² (13.9).

Total number ulcers: arginine butyrate plus standard local care: 37; standard local care: 25.

Number of ulcers per patient: arginine butyrate plus standard local care: 3.5; standard local care: 2.1.

Inclusion criteria:
1. Age: >18 years.
2. Sickle cell anaemia or sickle beta thalassaemia.
3. Presence of one or more leg ulcers which had been refractory to healing with National Institutes of Health (NIH)‐defined standard local care, consisting of twice‐daily cleaning and wet‐to‐dry dressing changes, for at least 6 months.
4.Patients were also eligible to participate if an ulcer had recurred and not healed with at least 3 months of standard care.

Exclusion criteria:
1. Renal or hepatic compromise.
2. Current chronic transfusion therapy.

Interventions

Intervention*: arginine butyrate at a total daily dose of 500 mg/kg/dose given 5 d/week plus standard local care.
Arginine butyrate formulation: 5% butyric acid (50 g/l) and 7.5% L‐arginine (75 g/l) by intravenous infusion through a port‐a‐cath or peripheral pass‐port, generally over 6–8 h/d.
The administration rate: 85 mg/kg/h and those patients infused with rates > 60 mg/kg/h were pre‐medicated with acetaminophen or ibuprofen and an anti‐emetic to prevent headache or nausea. Patients were encouraged and assisted to ambulate frequently during the 6‐h infusion.

Control arm: standard local care alone.

Co‐intervention: topical antibiotics and surgical debridement. Only 1 subject received concomitant hydroxycarbamide (in the control arm).

Outcomes

Wound healing: partial healing (decrease in ulcer area by at least 25% of the baseline ulcer area) and complete healing (complete closure of the ulcer (to an area of 0 cm²).

Notes

* If healing was objectively documented during the first 12‐week treatment cycle, as determined by a decrease in measured ulcer area by at least 25% of the baseline area, arginine butyrate could be continued for two additional courses of 8‐week cycles, although the responses to the extended treatment were not analysed as study endpoints. Control arm participants could cross‐over to the treatment arm if their ulcer did not heal after 12 weeks of closely monitored and supervised standard local care. Their remaining ulcers were then assessed on the treatment arm for 12 weeks.

Sample size estimation a priori: no.
Country: USA.
Founders: the Food and Drug Administration, Office of Orphan Product Development (Grant FD‐R‐000176) and the General Clinical Research Center at Boston University (Grant M01 RR00533).

Risk of bias

Bias

Authors' judgement

Support for judgement

Random sequence generation (selection bias)

Low risk

Quote "table of random numbers prepared by a blinded statistician." (page 517).

Allocation concealment (selection bias)

Unclear risk

Insufficient information to permit judgement of ‘low risk’ or ‘high risk’.

Blinding of participants and personnel (performance bias)
All outcomes

Unclear risk

Insufficient information to permit judgement of ‘low risk’ or ‘high risk’

Blinding of outcome assessment (detection bias)
All outcomes

Unclear risk

Insufficient information to permit judgement of ‘low risk’ or ‘high risk’

Incomplete outcome data (attrition bias)
All outcomes

Unclear risk

Insufficient reporting of attrition/exclusions to permit judgement of ‘low risk’ or ‘high risk’.

Selective reporting (reporting bias)

Low risk

‐

Other bias

High risk

Bias in the presentation of data (Appendix 2).
The inconsistency between 'randomisation unit: patient' and 'unit of analysis: ulcers' could generate the design and confusion bias (Appendix 2).

Methods

Design trial: parallel (two groups).
Study phase: III.
Baseline observation period: no (Dr Graham Serjeant, interviewed by one author of this review (JKM), on February 4, 2011).
Follow‐up period: 6 months.
Randomisation unit: patients.
Unit of analysis: ulcers.
Intention‐to‐treat analysis: no (information was supplied by Dr Graham Serjeant, interviewed by one author of this review (JKM), on February 4, 2011).

Participants

Enrolled: not described.
Randomised: 40 (unclear). This number was confirmed by Dr Graham Serjeant, interviewed by one author of this review (JKM), on February 4, 2011.

1. Reported patients: 40
2. Loss patients: 25% (10/40) (isoxuprine hydrochloride group: 4/ placebo group: 6) (This number was supplied by Dr Graham Serjeant, interviewed by one author of this review (JKM), on February 4, 2011).
3. Reported leg ulcers at the start trial: 63

• 3.1. 1 ulcer reported by 23 patients: 23 ulcers.

• 3.2. 2 ulcers reported by 12 patients: 24 ulcers.

• 3.3. 3 ulcers reported by 4 patients: 12 ulcers.

• 3.4. 4 ulcers reported by 1 patient: 4 ulcers.

• Total: 63 ulcers.

4. Ulcers developed during trial: 9 (# patients: no reported).
5. Total ulcers: 72 (63 + 9): 72.
6. Reported leg ulcers completing trial: 46 (75% 30/40 patients).
7. Ulcers developed during trial which completed trial: 8 (# patients: no reported).

8. Number of ulcers completing trial: 54 (isoxuprine: 32; placebo: 22).

9. Duration (median):
Isoxuprine group: 3.6 years (range: 0.3 to 15).
Placebo group: 3.5 years (range: 0.3 to 12).

Allocated group: not mentioned.

Information about haemoglobin disorder type:
Homozygous sickle cell disease "Diagnosed by haemoglobin electrophoresis on cellulose acetate and agar gel, and quantification of A₂ levels": 38.
Sickle cell‐haemoglobin C disease: 1.
Sickle cell‐haemoglobin O (Arab) disease: 1.

Age (years): 17 to 67.
Gender (% male): 52.5 (21/40).

Inclusion criteria: information was not supplied.
Exclusion criteria: information was not supplied.

Interventions

Experimental group:

Isoxuprine hydrochloride: 40 mg, per oral (bd).

Control group:

Placebo: 40 mg, per oral (bd).

Co‐interventions:
Local ulcer therapy: dressings plus mild antiseptic agent (not named): bd.

Outcomes

Outcomes were not describe as primary or secondary. They were reported into 'results' section:

Variation ulcer size; healed.

According with Dr Graham Serjeant, interviewed by one author of this review (JKM), 'ulcer area‐change' was the primary outcome (February 4, 2011).

Notes

Centre and Country: Medical Research Council Laboratories, University of West India, Jamaica.

Sample size estimation a priori: no.

Support: Mead Johnson and Company.
Founder role: it supplied isoxsuprine hydrochloride and financial assistance.

Risk of bias

Bias

Authors' judgement

Support for judgement

Random sequence generation (selection bias)

Unclear risk

Quote " patients were allotted serial numbers in a code..." (page 164).

Insufficient information about the sequence generation process to permit judgement of ‘low risk’ or ‘high risk’.

Allocation concealment (selection bias)

Unclear risk

Insufficient information to permit judgement of ‘low risk’ or ‘high risk’.

Blinding of participants and personnel (performance bias)
All outcomes

Unclear risk

Insufficient information to permit judgement of ‘low risk’ or ‘high risk’

Comment: Dr Searjent referred this paper as blinded (Interviewed by one author of this review (JKM), on February 4, 2011).

Blinding of outcome assessment (detection bias)
All outcomes

Unclear risk

Insufficient information to permit judgement of ‘low risk’ or ‘high risk’.

Incomplete outcome data (attrition bias)
All outcomes

High risk

The number or reasons for dropouts and withdrawals were not described.
Comments: this RCT loss 25% (10/40) of the patients and 27% (17/63) of the leg ulcers.

Selective reporting (reporting bias)

High risk

One or more clinically relevant and reasonably expected outcomes were not reported on; data on these outcomes were likely to have been recorded.

Other bias

High risk

Quote "The code was kept by the producing company".
Comments.
1. Characteristic baseline table: not shown.
2. It is unclear data on randomisation unit reported by the authors. Dr Graham Serjeant pointed out unit of randomisation were the patients, and unit of analysis were the ulcers (Interviewed by one author of this review (JKM), on February 4, 2011).
3. Ulcers developed during trail were included which were analysed.
4. There is inconsistency regarding numbers of the total ulcers.
5. Ascertaiment bias and bias in the presentation of data (Appendix 2).
6. The inconsistency between 'Randomisation unit: patient' and 'Unit of analysis: ulcers' could generate the design and confusion bias (Appendix 2).

Methods

Design trial: parallel (2 groups).
Study phase: no described.
Baseline observation period: 12 weeks.
Follow‐up period: 12 weeks.
Randomisation unit: patients.
Unit of analysis: patients.
Intention‐to‐treat analysis: declared. It was unclear, according with Dr Graham Serjeant, interviewed by one author of this review (JKM) on February 4, 2011.

Participants

1. Enrolled: not described.
2. Randomised: 15 (experimental group: 7; placebo: 8).
3. Excluded: 2 (experimental group:1; placebo:1).
4. Active participants: 13 (87%) (experimental group: 6; placebo: 7).
5. Reason for exclusion (2/15 = 13%)
Quote: "One subject in the placebo group defaulted after 27 days, and 1 participant in the PLC group, admitted to the hospital with acute chest pain after 46 days, failed to take the treatment for 54 days, and then resumed, but compliance was erratic" (Page 492).

6. Reported leg ulcers at the start trial: 34 (Data supplied by Dr Graham Serjeant, Jamaica, February 4, 2011).

7. Sex (male): 12 (80%).

8. Age (years): (arithmetic range):
Total Group: 17 to 40.
Experimental group: not described / Placebo group: not described.

9. Clinical characteristics of the leg ulcers: Leg ulcers of at least six months' duration and at least 3 cm diameter at the onset of treatment.

10. Inclusion criteria:
Patients with sickle cell disease.
Age (years): 17 to 40.
Leg ulcer duration: ≥ 6 months.
Leg ulcer diameter: ≥3 cm at the onset of treatment.

11.‐ Exclusion criteria: not described.

Interventions

Baseline treatment (12‐weeks): all patients
1. Measurement of ulcers: at 2‐week intervals.
2. Photographic images: at 4‐week intervals.
3. Dressing with 0.01% potassium permanganate: twice daily.
4. Debridament of dirty ulcers with crushed unripe papaya.
5. Oral zinc sulphate: 200 mg per oral, 3‐times‐daily, after meals.

Experimental group:
Propionyl‐L‐carnitine (PLC): 2 gr/ per oral/ twice‐daily.

Control group:
Placebo: lactose and microcrystalline cellulose/per oral/twice‐daily.

Co‐interventions:
As for baseline period.

Outcomes

Healing rate.
Quote: "... was standardized to a 12‐weeks period and expressed as (1) absolute change in area (change in ulcer area over treatment period divided by number of weeks on treatment ) x 12; and (2) percentage change in area (absolute change in area + area at start of treatment)" (page 491).

According with Dr Graham Serjeant, interviewed by one author of this review (JKM), 'ulcer area‐change' was the primary outcome (February 4, 2011).

Notes

RCT start date: not described.

Centre and Country: Conducted at Medical Research Council Laboratories, University of West India, Jamaica.

A priori sample size estimation: 14 patients (7 patients by group).

• P‐value: 0.05.

• Healing rate: 15% over three months.

• Power: 80%.

• Percentage healing rate: 74%.

• Trail number: not described.

• Expected loss: not described.

Support: Sigma‐Tau Pharmaceuticals, Gaittherborg, Md (Partial grant).

Founder role: not described.

Risk of bias

Bias

Authors' judgement

Support for judgement

Random sequence generation (selection bias)

Unclear risk

Quote "...were randomly assigned to a treatment group, seven in the PLC group and eight in the placebo group" (pages 491‐2).

Quote "randomisation within small blocks (2‐4)" This information was supplied by Dr Serjeant on February 4, 2011.

Insufficient information about the sequence generation process to permit judgement of ‘low risk’ or ‘high risk’.

Allocation concealment (selection bias)

Unclear risk

Insufficient information to permit judgement of ‘low risk’ or ‘high risk’.

Blinding of participants and personnel (performance bias)
All outcomes

Unclear risk

Insufficient information to permit judgement of ‘low risk’ or ‘high risk’

Comment: Dr Searjent referred this paper as blinded (Interviewed by one author of this review (JKM), on February 4, 2011).

Blinding of outcome assessment (detection bias)
All outcomes

Unclear risk

Insufficient information to permit judgement of ‘low risk’ or ‘high risk’

Incomplete outcome data (attrition bias)
All outcomes

Unclear risk

Insufficient reporting of attrition/exclusions to permit judgement of ‘low risk’ or ‘high risk’.

Selective reporting (reporting bias)

High risk

if one or more clinically relevant and reasonably expected outcomes were not reported on; data on these outcomes were likely to have been recorded.

Other bias

High risk

Design bias (Appendix 2).
Comment: there is imbalance: ulcer size before treatment (cm²) (±SD): 45.4(46.9) [4.9 to 146.28 range] in placebo group vs 19.7(4.7) [13.7 to 27.9] in experimental group.

Methods

Design trial: parallel (2 groups).
Multicenter study: yes
Study phase: no described.
Follow‐up period: 10 weeks.
Baseline observation period: none declared.
Randomisation unit: patients.
Unit of analysis: patients.
Intention‐to‐treat analysis: yes.

Participants

Randomised: 55 (RGD peptide matrix: 32) / placebo (23).
Hb genotype: HbSS (50), HbSP thalassaemia (2), HbSC(1), HbSC Harlem (1), HbSD (1).

Completed study: 48 (87%). RDG peptide matrix (27/32 = 84%), placebo (21/23 = 91%)
Loss: 7 (RGD peptide matrix: 5/32:16% ) / placebo (2/23:9%).
Imbalance between both groups: 7%

Age: median(± SE) and range: RGD peptide matrix: 32.9(1.8) years, range 18 ‐ 65 years; placebo: 36.6(2.3) years, range 26 ‐ 65 years.

Baseline ulcer duration (months) median(± SE) and range: RGD peptide matrix: 52.0 months (12.6), range 1 ‐ 312; placebo: 45.9 months (12.4), range 1 ‐ 192).

Baseline ulcer area (cm²) median(±SE) and range: RGD peptide matrix: 14.8(3.3), range 0.4 ‐91.3; placebo: 11.0(2.4), range 0.9‐36.7.

Gender (male: female): RGD peptide matrix: 21:11; placebo: 12:11.

Inclusion criteria:

1. isolated full‐thickness lower leg or ankle ulcers that did not involve bone or tendon and had persisted at least 1 month;

2. age: ≥18 years old;

3. informed written consent.

Exclusion criteria:

1. medical conditions that might retard healing (immune system, diseases, uncontrolled diabetes, bleeding disorders, neurological disorders, or cancer requiring chemotherapy or radiation treatment);

2. receiving medications that might adversely affect healing (systemic corticosteroids or antineoplastic agents);

3. history of chronic transfusion therapy within the 3 months preceding study commencement.

Interventions

RGD peptide matrix (Argidene gel: formerly Telio‐Derm gel, Telios Pharmaceuticals, Inc, San Diego, CA).
Applications and dressing changes: once per week by 10 weeks.

Placebo: saline solution.
Applications and dressing changes: once per week by 10 weeks.

Co‐intervention: debridament and cleaning at each visit.

Outcomes

Primary: changes in per cent ulcer closure.
Secondary: none declared.

Notes

Country: conducted at USA.
Sample size estimation a priori: no.
Support: Telios Pharmaceuticals, Inc.
Founder role: none declared.

Risk of bias

Bias

Authors' judgement

Support for judgement

Random sequence generation (selection bias)

Unclear risk

Quote "sequentially assigned to treatment groups based on a unique randomization number list" (page 1776).

Comments: imbalance exists on 'baseline ulcer area (cm₂ ).
No Information reported about how "unique randomization number list" was generated.

Allocation concealment (selection bias)

Unclear risk

Insufficient information to permit judgement of ‘low risk’ or ‘high risk’.

Blinding of participants and personnel (performance bias)
All outcomes

Low risk

Patient and outcome assessor were blinded.

Blinding of outcome assessment (detection bias)
All outcomes

Low risk

Patient and outcome assessor were blinded.

Incomplete outcome data (attrition bias)
All outcomes

Low risk

Dropouts < 20%

Selective reporting (reporting bias)

Low risk

‐

Other bias

High risk

Design bias (Appendix 2).