Antidepressants are believed to help manage select pain syndromes, with fewer adverse effects than nonsteroidal anti-inflammatory drugs (NSAIDs) or opioids. One of these antidepressants, duloxetine, was approved for treating chronic musculoskeletal pain, including osteoarthritis (OA). Is this agent effective alone or in combination with NSAIDs for the relief of knee pain due to OA? An evidence-based assessment raises some doubts.
Leslie Citrome, MD, MPH, and Amy Weiss-Citrome, MD, both of New York, assessed recent clinical evidence on duloxetine — an oral, centrally-acting selective serotonin and norepinephrine reuptake inhibitor (SNRI) — introduced as an antidepressant in 2004 and FDA-approved in 2010 for chronic musculoskeletal pain, including back pain and osteoarthritis. Writing in an early, online edition of the International Journal of Clinical Practice, the authors looked at clinical trials examining benefits and harms of duloxetine monotherapy or in combination with NSAIDs [Citrome and Weiss-Citrome 2012-A].
Study outcomes were analyzed using number-needed-to-treat (NNT) and number-needed-to-harm (NNH) metrics. This further allowed calculation of a unique Benefit-to-Harm ratio, or what the authors call “Likelihood to be Helped or Harmed (LLH).”
These statistical measures were described in a very recent UPDATE [here]. Briefly, they quantify how many patients need to be treated with one intervention versus another before one additional patient experiences a desired outcome (NNT) or a harmful event, such as a side effect (NNH). The smaller the NNT and the larger the NNH the better. The Benefit-to-Harm ratio, or “LHH” as Citrome and Weiss-Citrome label it, is calculated simply as NNH/NNT; it answers the vital clinical question, “How many patients might benefit from a therapy for each patient who experiences a harmful effect of some sort?” The larger this ratio the better.
Most studies in the pain management field report differences between groups on outcomes of interest, whether beneficial effect or harm, and the statistical significance (P-value), if any, of these measures — but stop there. Going a step further to calculate NNT, NNH, and NNH/NNT transforms the data into language that is more meaningful for practitioners and patients; that is, whether the therapy will likely do more good than harm, and by how much.
Systematically examining 13-week trials of duloxetine by itself, as monotherapy, for moderate knee pain due to OA, the authors found that the NNT for duloxetine vs placebo on a composite measure of treatment effectiveness was 6 (95% Confidence Interval, 4–10) [reported in Citrome and Weiss-Citrome 2012-A and 2012-B]. Or, for every 6 patients treated with duloxetine rather than placebo, one additional patient would find the drug effective.
This was slightly less favorable than earlier research on OA examining the NSAIDs etodolac after 4 weeks (NNT 5; 95% CI, 3–25) and tenoxicam after 8 weeks (NNT 4; 95% CI, 3–8). Of interest, pain relief with duloxetine was not dependent on improvement in any symptoms of depression; however, the CIs of duloxetine and the 2 NSAIDs were overlapping, so it cannot be said that one was significantly better than the other.
In terms of harms, compared with placebo, duloxetine exhibited NNHs of 16, 17, and 19 for nausea, fatigue, and constipation, respectively, which were the most common side effects. These are considered to be somewhat favorable NNHs for such side effects. However, approximately 16.3% of patients receiving duloxetine compared with 5.6% receiving placebo discontinued treatment due to one or more adverse effects, for an NNH of 10 (rounded up from 9.35; 95% CI 7-20). In other words, for every 10 patients treated with duloxetine rather than placebo, one additonal patient might drop out due to adverse effects.
Another way of looking at these data [not calculated by Citrome and Weiss-Citrome] is that the Benefit-to-Harm ratio (NNH/NNT) for discontinuation of duloxetine monotherapy was roughly 2 (10/6=1.67 rounded up). That is, for every 2 patients successfully treated with duloxetine monotherapy rather than placebo for moderately painful knee OA, one more patient might find the drug intolerable and drop out of treatment. The individual NNH/NNT ratios for nausea, fatigue, and constipation were all approximately 3 (rounded numbers).
Citrome and Weiss-Citrome [2012-A] further observe that, although duloxetine is approved as monotherapy for OA, it is commonly used in combination with other agents, such as NSAIDs. Would this improve analgesic efficacy? How might tolerability be affected?
To answer those questions, the authors assessed the findings of a recent study examining the potential synergy of duloxetine and NSAIDs [Frakes et al. 2011]. This adequately-powered, double-blind trial of 524 patients with moderately painful knee OA, found that those who took a combination of duloxetine (60 mg/d or 120 mg/d) plus an NSAID reported greater pain reductions at week 8 than the control group taking an NSAID plus a placebo. Ibuprofen and naproxen were the most commonly used NSAIDs, at typical daily doses for moderate to severe pain.
NNTs for ≥30% and ≥50% improvements in pain severity with duloxetine+NSAID compared with placebo+NSAID were roughly identical (NNT 6; 95% CI 4-9). The NNHs for nausea, fatigue, and constipation were 10, 19, and 18, respectively. The NNH for treatment discontinuation due to adverse effects was 16 (95% CI, 9–130).
The authors also express concern about gastrointestinal bleeding events occurring more frequently in the duloxetine+NSAID group; although, the NNH=89 for this adverse effect was not statistically significant. At the same time, this was a relatively short trial and they observe that a serotonergic antidepressant, which can be associated with platelet dysfunction, combined with NSAIDs over time may incur additional bleeding events.
The Benefit-to-Harm ratio (NNH/NNT) for discontinuation of duloxetine+NSAID therapy due to adverse events was approximately 3 (16/6=2.67 rounded up). So, for every 3 patients who benefit from duloxetine+NSAID therapy for knee OA, one additional patient might find the combination intolerable and drop out of treatment.
However, as noted above, the confidence interval range for the treatment discontinuation NNH was wide (9-130), and the upper limit quite high, so it is possible that the true Benefit-to-Harm ratio might be much more favorable. Note, however, that a wide CI also suggests a high degree of variance in the difference between groups on this measure and a lack of precision in the NNH point estimate to begin with.
Overall, Citrome and Weiss-Citrome [2012-A] conclude from their analyses that clinicians managing patients suffering from knee osteoarthritis should consider prescribing antidepressants, either alone or adjunctively. However, from the above data, advantages gained by adding duloxetine to the pharmacotherapeutic mix for treating OA may be questionable.
COMMENTARY: Duloxetine Doubtful for Osteoarthritis
An evidence-based assessment of research using data converted to NNT, NNH, and NNH/NNT — as demonstrated by Citrome and Weiss-Citrome — is, in our opinion, a very meaningful way to view outcomes in terms of their everyday relevance for clinical practice. Research-report writers should always provide the NNT and NNH for outcomes whenever possible, along with their 95% Confidence Intervals — unfortunately, this is seldom done in pain research.
In the above analyses, the NNT for duloxetine monotherapy was comparable to the NSAIDs etodolac and tenoxicam for the treatment of knee OA. At the same time, however, the Benefit-to-Harm ratio of 2 for treatment discontinuations with duloxetine monotherapy, as compared with placebo, could fall short of practitioner and patient expectations for a safe and effective therapy.
Citrome and Weiss-Citrome take a rather sanguine view of Benefit-to-Harm ratios, or what they call the LHH (Likelihood to be Helped or Harmed). They would say that a ratio of 2 means that duloxetine is twice as likely to help as to harm the patient. This is true, but it is important to keep in mind how risk is being portrayed — that for every 2 patients helped by duloxetine therapy one additional patient may find the drug so intolerable as to discontinue treatment. Practitioners and patients need to know this, and looking only at the NNT (6, in this case) would not alert them to the concurrent possibility of harms.
Adding an NSAID to duloxetine had no impact on nausea, fatigue, and constipation; the Benefit-to-Harm ratios for these adverse effects in the two studies were roughly the same. Furthermore, compared with duloxetine monotherapy the duloxetine+NSAID combination only slightly improved the NNH/NNT ratio for treatment discontinuations, from 2 to 3. However, it also must be acknowledged that the combination therapy was compared with NSAID, whereas duloxetine monotherapy was compared with placebo, so a direct comparison of NNH/NNT ratios in the two studies could be misleading.
Still, it appears from the various studies that duloxetine confers relatively little added benefit vs risk as monotherapy, compared with NSAID monotherapy, or when combined with an NSAID for the treatment of knee OA. Add to this the considerable extra cost for duloxetine and the advantages of this agent seem dubious.
In fact, just-released updated guidelines on pharmacologic therapies for osteoarthritis from the American College of Rheumatology (ACR) has “no recommendation” regarding duloxetine for the initial management of knee OA [Hochberg et al. 2012, ref below]. And, in patients who have failed all other pharmacologic and nonpharmacologic approaches, the ACR strongly recommends opioid therapy and only “conditionally” recommends the use of duloxetine.
There are, however, some caveats regarding the above observations to consider:
- NNT, NNH, and NNH/NNT are clinically intuitive measures of effect size, but there are limitations to their use as we noted in our earlier UPDATE [here]. The discussion above demonstrates how these metrics need to be carefully interpreted when making indirect comparisons across studies, as when duloxetine was compared with placebo in one study, and duloxetine+NSAID was compared with placebo+NSAID in the other. Sussing out any specific advantages of duloxetine alone versus adding duloxetine to NSAID from these comparisons is tenuous.
- Citrome and Weiss-Citrome [2012-B] also caution that those metrics do not take into account time-to-onset of the adverse event(s) for NNH or time to therapeutic response for NNT, as well as their duration. For example, commonly occurring but short-lived side effects could lower the NNH, but may not truly diminish the Benefit-to-Harm ratio in everyday practice. Or, adverse events may occur before therapeutic effects begin, which might exacerbate early treatment discontinuations unless both patients and practitioners are aware of this.
- It could be important that the studies of duloxetine and duloxetine+NSAID were done in patients with only moderate OA pain (≥4 on a 0–10 rating scale). Whether stark differences in effectiveness and/or adverse events might be found in patients with more severe pain is unknown.
- It also is unknown whether other antidepressants might foster better outcomes than duloxetine.
What could be most beneficial in the treatment of OA are therapies that decrease the NNT and/or increase the NNH. One approach might be pairing low-dose duloxetine or NSAIDs with nonpharmacologic modalities. Researchers have yet to fully explore combinations of minimal pharmacotherapy with “low-risk” modalities such as thermotherapy (heat, cold), exercise, massage therapy, acupuncture, and others. Added pain relief with fewer adverse effects might also be conferred by topical analgesics and supplements, such as vitamin D; although, the ACA guidelines specifically advise against topical capsaicin and glucosamine-chondroitin for initial treatment of OA [Hochberg et al. 2012].
A key message appears to be that NNT, NNH, and NNH/NNT can be important metrics for helping to determine clinical significance of any therapy. Along with that, effective pain management needs to look beyond the prescription pad and consider multimodal approaches that confer the greatest Benefit-to-Harm ratios, as confirmed by appropriate, unbiased, and valid research initiatives.
> Citrome L, Weiss-Citrome A. Antidepressants and the relief of osteoarthritic pain - Findings from a study examining adjunctive duloxetine. Int J Clin Prac. 2012-A(Mar); online ahead of print [access here].
> Citrome L, Weiss-Citrome A. A Systematic Review of Duloxetine for Osteoarthritic Pain: What is the Number Needed to Treat, Number Needed to Harm, and Likelihood to be Helped or Harmed? Postgrad Med. 2012-B(Jan);124(1):83-93 [access here].
> Frakes EP, Risser RC, Ball TD, et al. Duloxetine added to oral nonsteroidal anti-inflammatory drugs for treatment of knee pain due to osteoarthritis: results of a randomized, double-blind, placebo-controlled trial. Curr Med Res Opin. 2011(Dec)27(12):2361-2372 [abstract here].
> Hochberg MC, Altman RD, April KT, et al. American College of Rheumatology 2012: Recommendations for the Use of Nonpharmacologic and Pharmacologic Therapies in Osteoarthritis of the Hand, Hip, and Knee. Arth Care Res. 2012(Apr);64(4):465–474 [PDF here].
Don’t Miss Out. Stay Up-to-Date on Pain-Topics UPDATES!
Register [here] to receive a once-weekly e-Notification of new postings.