Study	Reason for exclusion
Wilkinson 2003	20% of its participants had thoracic or cervical spinal pain and were not analysed separately. Study design involved crossover between experimental injections of bupivicaine/phenol/glycol and control injections of bupivicaine on the second treatment, making long term results uninterpretable.
Wilkinson 2005	Published report of previously unpublished Wilkinson 2003 study.
Yelland 2000	A multi‐centre pilot study comparing the effectiveness of prolotherapy with a range of other conservative treatments for the treatment of chronic low back pain. Excluded as randomisation was not used.