Carroll 2009.

Study characteristics
Methods	RCT cross‐over
Participants	N = 9 Age mean (range) 49.4 (38‐67) Sex: 1 male Lower limb CRPS‐I, with duration of pain of at least 6 months, spontaneous pain rating > 6/10, unsuccessful therapy with at least 2 non‐opioid medications (for neuropathic pain), at least a 50% reduction in pain for > 5 h but < 2 weeks from a previous lumbar sympathetic injection Inciting events: tarsal tunnel surgery n = 1, bunionectomy/cast n = 1,  crush injury n = 1, plantar fasciectomy n = 1, foreign body removal n = 1, ankle arthroscopy n = 1, ankle fracture/cast n = 1, metatarsal fracture n = 1, back surgery n = 1 Diagnostic criteria: IASP (Merskey 1994) Medico‐legal factors: not reported Mean duration of pain (range): 3.8 years (2‐14) Baseline mean pain levels, 10 cm VAS (range): 7.2 cm (4.7‐8.9) Concomitant treatments: not reported but participants asked not to cease existing therapies, but not to start new therapies during study period
Interventions	Lumbar sympathetic blocks: Anterolateral border of L2 vertebral body, fluoroscopy guided Active: 10 ml of 0.5% bupivacaine with an added 75 units botulinum toxin A Control: 10 ml of 0.5% bupivacaine
Outcomes	Primary outcome: time to analgesic failure (time for pain to return to baseline level) Daily pain intensity ‐ 10 cm VAS, measured for 7 days prior to first injection and recorded daily until participants reported their pain returned to baseline or 1 month (whichever was longer) Adverse events reported
Country of origin	USA
Study aim	To determine whether adding BTA to lumbar sympathetic blockade increases the duration of analgesia
Notes	Authors declared that they had filed a patent for BTA in sympathetic blocks
Risk of bias
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Unclear risk	Quote: "Patients were randomly assigned to which injection they received first" Comment: method of randomisation not described
Blinding of participants and personnel (performance bias) All outcomes	Low risk	Quote: "All physicians and patients involved in the study were blinded to which injection contained botulinum toxin A. Data were not unblinded for any patient until the study was completed, and no interim analyses were performed"; "in the crossover injection, the patient received an identical injection" Comment: injections were identical and participants blinded
Blinding of outcome assessment (detection bias) All outcomes	Low risk	Quote: as above Comment: self reported outcomes and participants were blinded.
Incomplete outcome data (attrition bias) All outcomes	Unclear risk	2/9 participants did not complete the study (one due to technical issues related to the block ‐ malpositioning of the needle ‐ and one because outcome forms were not returned). Only 1 participant received BTA first, and this participant dropped out. Due to complete blinding and use of a cross‐over study design, the effect that this drop‐out has on the results is unclear.
Selective reporting (reporting bias)	Unclear risk	Full data not presented for the secondary end point (VAS pain scores over time); comparison of within‐injection group change over time provided, but comparison of between‐injection group differences not provided.
Adequate sample size?	High risk	N = 9
Adequate duration of follow‐up?	Unclear risk	Quote: "Patients continued to record daily VAS until they reported their pain had returned to baseline or 1 month, whichever was longer" Comment: Follow‐up was observed until pain returned to baseline levels ‐ for some this was only 4 weeks
Free of other bias?	Low risk	Cross‐over study design ensured similarity between groups; participants allowed to continue current medications but could not start new medications. Quote: "patients were eligible for their crossover injection 1 month after they reported their pain had returned to baseline" Comment: 1 month washout period observed, after pain had returned to baseline levels