We performed a comprehensive literature search of the MEDLINE, EMBASE, Cochrane Library, and CINAHL databases using their respective medical subject headings. A three-step search strategy was used to identify all potential articles of interest. First, all articles related to LBP or lumbar degenerative diseases were retrieved. Next, the subset of articles that were indexed under the subject headings of cost and cost analysis was identified. Finally, these articles were further filtered to only retain those studies that included quality-of-care data in addition to cost analysis. The search strategy included articles with medical subject headings of articles that included lumbar vertebra and low back pain and/or spinal disease with consideration of costs and utility including keywords of cost and cost analysis, quality of healthcare, quality of life, and quality-adjusted life years (Table ). Potentially relevant articles lacking online access were ordered from the National Library of Medicine.
MEDLINE, EMBASE, Cochrane Library, and CINAHL database search strategy details and results (including duplicates)
A total of 1004 articles (including duplicates) were retrieved on June 24, 2010, with the EMBASE database contributing 55%, MEDLINE contributing 26%, and Cochrane Library contributing 19% of the retrieved articles (Table ). All three articles retrieved from CINAHL were duplicates. The MEDLINE, Cochrane Library, and CINAHL medical subject headings used for identifying the studies were identical. The EMTREE medical subject headings corresponding to cost and quality search criteria were broader in scope compared with other databases, possibly contributing to the retrieval of the highest number of search results through this database.
The titles and the abstracts of 752 unique articles were initially screened by one of the authors (SI) to identify the subset of studies to be evaluated more closely. The criteria used for this preliminary screening included relevance of the article’s subject matter to the current review on the value of various LBP interventions as well as the availability of cost and utility data in the study.
Full-length texts of this initial subset of studies were evaluated by one of the authors (SI) based on the following inclusion and exclusion criteria to further identify studies eligible for inclusion in this review. Inclusion criteria were (1) English language; (2) adults, 18 years of age and older; (3) at least 15 patients per intervention group at the start of the study; (4) minimum 1-year followup (considered as long-term followup for nonoperative interventions by the Cochrane Back Review Group [6
]); (5) cost of intervention data reported; and (6) utility values (SF-6D or EQ-5D) or incremental quality-adjusted life-years (QALYs) gained or incremental cost-utility ratio (ICUR) reported. Exclusion criteria were (1) case reports; (2) meta-analyses or review articles; (3) PhD theses or conference abstracts; (4) modeling studies based on cost-effectiveness analyses; and (5) database search results with unavailable abstracts.
References of all previously published systematic reviews on cost-utility analysis (CUA) of LBP interventions as well as the studies selected for inclusion in this study were also reviewed to identify any additional studies not retrieved earlier or missed in the initial screening. Review of references led to identification of no additional articles for review or inclusion in the study.
Screening the title and abstract based on the relevance of the study in question to the topic of this review yielded 32 studies of interest. Four of the 32 articles [19
] were excluded as a result of a followup shorter than 1 year. The study by Ijzelenberg et al. [12
] was excluded because it investigated the cost-utility of a LBP prevention program, not an intervention to manage LBP. As a result, a total of 27 studies were selected for inclusion in the systematic review. The earliest study included was published in 1995 [18
] and the two latest studies were published in 2010 [8
]. The references of articles selected for this review and the previously published systematic review by Dagenais et al. [5
] did not yield identification of additional studies. Of the 27 studies, 16 studies included here present the comparative cost-utility of two or more nonoperative interventions in managing symptomatic LBP, four studies evaluate the cost-utility of two operative interventions against each other in managing LBP, and seven studies compare the cost-utility of a nonoperative intervention with an operative intervention. All 27 studies identified for inclusion in this review were examined independently by all the reviewers (SSI, SHB [treating surgeon], MHW, SKT, SSH) on three different aspects to ascertain their study quality: (1) source of risk of bias: each study was assigned a low or high risk of bias based on the 12 question criteria proposed in Cochrane Back Review Group’s (CBRG) 2009 guidelines [7
]; (2) strength of recommendation was determined as strong or weak based on the American Thoracic Surgeons (ATS) Guidelines and Recommendations [31
]; and (3) quality of evidence was rated as high, moderate, or low based on the ATS Guidelines and Recommendations [31
The composite quality score was the score for each dimension of quality (risk of bias, strength of recommendation, and quality of evidence). The composite quality score for each study was calculated based on the score it received from the majority of the reviewers on each dimension of quality mentioned. For example, a study receiving a high risk of bias from two reviewers but a low risk of bias from the remaining three was assigned an overall low risk of bias for the purposes of this systematic review. The same study receiving a scoring of strong recommendations from three reviewers and weak recommendations from two reviewers was determined to have made strong recommendations overall for the purposes of this review. A majority opinion on each dimension of the quality score (risk of bias, strength of recommendations, quality of evidence) emerged for each examined study, eliminating the need to reevaluate or resolve any inconclusive overall scores through consensus. There was moderate variability in interobserver grading of articles. Bias (graded as high or low) consisted of five of five authors agreeing 26% of the time (seven of 27 studies), four of five authors agreeing 44% of the time (12 of 27 studies), and three of five authors agreeing 30% of the time (eight of 27 studies). Strength of recommendations (graded as strong or weak) consisted of five of five authors agreeing 30% of the time (eight of 27 studies), four of five authors agreeing 37% of the time (10 of 27 studies), and three of five authors agreeing 33% of the time.
Quality of evidence (graded as high, moderate, low) consisted of five of five authors agreeing 15% of the time, four of five agreeing 33% of the time (nine of 27 studies), and three of five authors agreeing 52% of the time (14 of 27 studies).
Nine studies (six in Table , three in Table , and zero in Table ) were determined to have a high risk of bias [7
]. The majority of the studies were scored to have weak recommendations (17 of 27) and moderate evidence quality (18 of 27).
Design details of studies reporting the comparative CUA of two or more nonoperative interventions
Details of studies reporting the comparative CUA of a nonoperative intervention relative to an operative intervention
Details of studies reporting the comparative CUA of two operative interventions
The following data were extracted from each selected article: (1) study design; (2) source for utility cost data; (3) study length; (4) compared interventions; (5) number of patients recruited; (6) patients’ mean age; (7) percentage lost to followup; (8) baseline patient characteristics; (9) diagnosis or indication; (10) utility values; (11) QALYs gained; (12) total cost: direct charges or reimbursement and estimated indirect costs; and (13) CUA findings.
Many studies estimated an indirect cost, or a societal cost, based on considerations of time from work and lost productivity. All cost data were converted to US dollars (if necessary) using end-of-year currency exchange rates (available at http://www.oanda.com/currency/converter
). The inflation-adjusted 2010 US dollar cost of an intervention was calculated using the consumer price index inflation calculator (available at http://data.bls.gov/cgi-bin/cpicalc.pl
). For cost data collected over several years, the final year was used as the index year for currency conversion and inflation adjustment. If the year of cost data collection was not specified, the year of publication was used as the index year.