Introduction
Evergreening refers to owners of pharmaceutical products using numerous strategies, such as patent laws and minor drug modifications, to extend their monopoly privileges with their products [
1‐
3]. Typically, these strategies are developed before expiry of the patent of an original drug, usually a high-revenue drug [
4‐
6]. If they succeed, they result in an extension of the patent protection period or a new patent for a minimally modified version of the drug.
A consequence of evergreening is delayed entry of generic drugs into the market with extension of the original drug patent or competition between the patent-protected minimally modified version of the drug and generic drugs [
7]. This situation might increase drug reimbursement costs by keeping the cheaper generic versions completely or partly out of the market [
8]. Pharmaceutical companies defend evergreening practices and claim that revised formulas benefit patients (for example, by improving adherence) and the drug industry (for example, by providing incentives for companies to engage in incremental innovation) [
9‐
11].
Minimal modifications used in evergreening include use of a different salt or molecule as an additive to the main drug components, change in formulation, modified release or change in route of administration [
12,
13]. An enantiomer patent is another form of evergreening based on a chiral switch (that is, from a chiral drug developed as a racemic mixture to a single enantiomer) [
14]. Single-enantiomer drugs represent more than 50% of the top-selling 100 drugs worldwide [
15].
A typical example of this strategy was the case of citalopram/escitalopram, two antidepressants. The Lundbeck company's patent on citalopram has run out in many countries [
15]. The company launched a single-enantiomer drug, escitalopram, before the patent on the original drug expired and significantly increased its advertising campaigns to promote the new form [
16]. However, the clinical superiority of escitalopram over citalopram is still debated [
15]. Moreover, previous evergreening's societal burden analyses, dedicated to other evergreening examples, were based on projections of market shares and health insurance reimbursement costs [
4,
17].
We aimed to evaluate the impact of evergreening citalopram with the chiral switch form escitalopram on efficacy and treatment acceptability for patients, as well as health insurance costs for society.
Methods
We investigated the relative efficacy and acceptability of citalopram and escitalopram by performing meta-analyses for direct evidence (meta-analyses of results of head-to-head trials) and indirect evidence (adjusted indirect comparison of results of placebo-controlled trials). Health insurance costs were analyzed through reimbursement data for citalopram, its generic forms and escitalopram from the French national health insurance information system.
Assessment of relative efficacy of escitalopram and citalopram
Identification and selection of randomized controlled trials
We identified randomized controlled trials by systematically identifying reviews published from 2000 to 2011 and trial results published from 2011 to 2012 [
18], see Section 1, Additional file
1.
Eligible reviews assessed the efficacy of citalopram or escitalopram in adults with major depression based on randomized trials. We searched several bibliographical databases for reviews published between January 2000 and March 2011, and four repositories of national health technology agencies, as well as the FDA. See Section 1, Additional file
1.
Eligible randomized trials assessed acute treatment efficacy, which was defined as eight-week treatment efficacy of citalopram versus escitalopram or citalopram and/or escitalopram versus placebo in patients with major depression, see Section 1, Additional file
1 for detailed selection criteria. First, we screened selected reviews and listed all included trials. The eligibility of trials was assessed independently by two reviewers, with disagreements resolved by consensus. Then, we searched for trial results published from March 2011 to February 2012 in MEDLINE and EMBASE. Finally, we searched for trial results in databases from Lundbeck and Forest registries [
19,
20]. We also contacted Lundbeck/France for a list of clinical trials for the two medications.
Outcome measures
We assessed acute treatment efficacy, which was defined as eight-week treatment. When the depression outcome was measured at several timepoints, we extracted outcome data at eight weeks. If not reported, we extracted outcome data for the closest time points, ranging from 4 to 12 weeks [
21,
22], see Sections 1 and 2, Additional file
1 for details about data extraction. We used outcome data for the Montgomery-Åsberg depression rating scale (MADRS) and, if not reported, the Hamilton scale. Efficacy was assessed by the proportion of responders in each treatment group, defined as patients with a decrease in depression score from baseline to follow-up of at least 50%, see Section 2, Additional file
1.
We assessed treatment acceptability by the proportion of patients who did not drop out of the allocated treatment during the short-term treatment period (completers), see Section 2, Additional file
1. Data were extracted by two reviewers independently. Disagreements were resolved by discussion. If outcome data were available from FDA reports and other sources, priority was given to FDA data, because the FDA re-analyses of raw data from the sponsor adhered to the pre-specified statistical methods in the trial protocol [
23].
First, we performed a meta-analysis of head-to-head trials. Then, we performed an adjusted indirect evaluation of the relative efficacy of each treatment compared to placebo using Bucher's method [
24]. The effect of treatment was measured by odds ratios (ORs) and 95% confidence intervals (95% CIs). Combined estimates were calculated by fixed- and random-effects models. The two models always showed similar results [
25]. In cases of significant treatment effect, we re-expressed the results in terms of number needed to treat (NNT). We computed the NNT from the combined ORs and by considering low and high response rates for the control group, defined as the lower and upper bounds of the 95% CI for the combined response rate across control groups in the meta-analysis.
We assessed heterogeneity of treatment effect estimates across trials using the I² statistic. We assessed similarity (whether citalopram and escitalopram placebo-controlled trials were similar for moderators of relative treatment effect) by comparing clinical and methodological characteristics of randomized comparisons [
26]. We assessed inconsistency between the direct and adjusted indirect estimate by the difference between the two estimates and associated 95% CIs and tested whether it was statistically significant [
27].
We assessed small-study effects by funnel plots [
28]. Sensitivity analyses for direct and adjusted indirect comparisons involved re-analyzing data after excluding head-to-head trials without comparable dosages and excluding placebo-controlled trials as soon as the treated group did not receive defined daily dose (DDD) dosages, respectively. In addition, sensitivity analyses were performed excluding trials with imputed outcome data and trials of older adults only.
Assessment of reimbursement costs for citalopram, its generic drugs and escitalopram
To assess the reimbursement burden of the three drug forms on the French general health insurance regimes, we analyzed data from the French national health insurance information system (Système National d'Information Inter-Régimes de l'Assurance Maladie, SNIIR-AM).
Data sources
The SNIIR-AM contains anonymous and comprehensive data on health spending reimbursements from 2003. In 2009, it covered 86% of the French population, approximately 53 million people. We used a random representative sample (1/97) of SNIIR-AM, the Echantillon généraliste de bénéficiaires (EGB). In 2009, the EGB consisted of about 500,000 beneficiaries [
29,
30]. We searched the EGB database using French identifiers for drug products for the three drug forms. We examined all reimbursement claims for the drugs between January 2003 and December 2010, given that generic citalopram was introduced in December 2003 and escitalopram was introduced in June 2005 in France.
Reimbursement cost analysis
To illustrate changes in consumption pattern, we calculated the monthly number of reimbursements and monthly consumption in DDD units for the three drug forms. According to the World Health Organization, the DDD is the assumed average maintenance dose per day for a drug used for its main indication in adults. The DDD is 20 mg for citalopram and 10 mg for escitalopram [
31]. To illustrate the evolution of spending, we calculated the total monthly reimbursement costs for the three drug forms. For all data, we produced time series plots. Because the EGB sample is representative of the SNIIR-AM population, we projected our analyses by dividing all results by the sampling fraction so that estimates reflected the whole SNIIR-AM population [
30].
Analyses involved use of Stata MP v10.0 (Stata Corp., College Station, TX, USA). A P < 0.05 was considered statistically significant.
Discussion
We performed a large-scale systematic review to evaluate the impact of evergreening of citalopram with the chiral switch form escitalopram on efficacy, treatment acceptability and health insurance costs. We found substantial discrepancy between the direct and indirect comparisons of citalopram and escitalopram for short-term treatment efficacy and acceptability. The direct comparison showed clinical superiority of escitalopram over citalopram, but the indirect comparison showed no evidence of difference between the two. Our analysis of reimbursement costs for a large representative sample of French health insurance beneficiaries, showed that escitalopram took a large share of the market, with citalopram substantially lower. The generic forms of citalopram started to gain an expected share of the market, but the uptake was suppressed by escitalopram competition. Moreover, escitalopram consumption overcame citalopram and its generic forms combined.
Network meta-analysis can be used to combine direct and indirect estimates of efficacy or acceptability. The analysis borrows strength from all the available evidence and increases statistical power and precision of estimates. However, we could not perform such a network meta-analysis because of the observed discrepancy between direct and indirect evidence. A possible explanation for the discrepancy is dose difference among trials; however, sensitivity analysis for comparable dosages showed consistent results. Another possible explanation could be bias in the direct comparison or the indirect comparison [
27,
31]. Head-to-head trials are usually considered the gold standard. However, such trials may favor the sponsored treatment [
33,
34] or the newest treatment [
35‐
37]. In our analysis, most head-to-head trials, except two, were sponsored by the manufacturer (Lundbeck/Forest Lab). One of the two trials was sponsored by Arbacom, a Russian company with potential conflict of interest with the Danish manufacturer Lundbeck [
38]. The trial results contained outlying results, with strong superiority of escitalopram over citalopram. The second trial was an institutional funded trial [
32], and did not find any evidence of difference between the two drugs. Escitalopram was superior to citalopram in a network multiple-drug treatments meta-analysis of efficacy without placebo controlled trials [
22]. However, citalopram was superior to escitalopram, with a statistical non-significance, in an extensive multiple-drug treatments meta-analysis that considered the entire network of second-generation antidepressant drugs and including placebo-controlled trials [
21].
Our indirect comparison may have been biased. For instance, the characteristics of citalopram and escitalopram placebo-controlled trials may have greatly differed, which would have invalidated the adjusted indirect comparison. However, we found no evidence of unequal distribution of potential treatment effect modifiers. Placebo-controlled trials may have exhibited reporting bias. Nevertheless, this bias is unlikely because as compared with active comparator trials, placebo-controlled trials are frequently registered with the FDA, which is considered a gold standard for placebo-controlled trials in the antidepressant field [
39] and when we searched the FDA database, we identified five trials with unpublished data. As well, the indirect comparison may have had low statistical power [
40]. However, we ensured a balance in the number of included trials, with at least 10 randomized comparisons for both citalopram and escitalopram versus placebo.
Our study has some limitations. In the comparative effectiveness analysis, we assessed treatment acceptability only and did not assess safety outcomes [
21,
22]. Our findings could not be generalized to other patent-extended medications using chiral switch or other indications for citalopram/escitalopram usage. Second, we examined the EGB sample [
29], although the EGB is a representative sample of the SNIIR-AM database [
30]. The EGB data limited our analysis to begin with 2003, so we were not able to look at the cost and consumption trends earlier than 2003.
Conclusions
We found strong uncertainty about the clinical benefits of escitalopram over citalopram. Given the likelihood of sponsorship bias for head-to-head trials and the absence of reporting bias for placebo-controlled trials [
41], our adjusted indirect comparison may be less biased than the direct comparison. However, the market share of escitalopram increased substantially and suppressed that of the generic forms, which may have prevented a substantial cost savings for health insurance, especially if we assumed prescriptions shifted from escitalopram to generic citalopram. Finally, as evergreened medications are typically launched to the market before the patent of the original product expires - in the expectation of generic competition - it might be suitable to base the new product costs on the estimated price of the generic form, rather than on the current price of the originator. Moreover, health technology assessment agencies could redefine product innovation to reflect actual added benefits to patients and society. This might effectively make the evergreened product less cost effective and might help discourage this practice [
42].
Acknowledgements
We thank Dr. Annie Rudnichi from the Haute Authorité de Santé and Marie Dalichampt for their help with querying the SNIIR-AM database, Laura Smales (BioMedEditing, Toronto, Canada) for editing the manuscript.
An ethics statement was not required for this work and no funding was received for this work, no funding bodies played any role in the design, writing or decision to publish this manuscript. We had full access to all the data and take responsibility for the integrity of the data and the accuracy of the analysis.
Competing interests
The authors declare that they have no competing interests.
Authors' contributions
AA contributed to the study design and performed the literature search, data collection and data interpretation. AA was also involved in drafting the manuscript. LT substantially contributed to the design, data collection and data analysis, and was also involved in the manuscript drafting and revising its intellectual content. GB contributed in gathering reimbursement data and its analysis, and also critically revised the manuscript for intellectual content. MD contributed to the conception and design of the study, and also was involved in the manuscript revision and approving its intellectual content. PR contributed substantially to the conception, design of the evergreening study and the interpretation of data, and was involved in revising the manuscript critically for intellectual content and has given the final approval of the version to be published. All authors have read and approved the manuscript for publication.