Dear Editor,

The European Medicines Agency (EMA) recently concluded a safety review of fluoroquinolone antibiotics, which are commonly prescribed in Europe [1]. The EMA recommended that the marketing authorization of medicines containing cinoxacin, flumequine, nalidixic acid, and pipemidic acid should be suspended and that the use of the remaining fluoroquinolone antibiotics should be restricted due to the risk of disabling and potentially long-term side effects [2]. This led to a restriction in indication following a harmonization procedure as part of the referral. We recently published a nested case–control study characterizing the risk of tendon rupture with fluoroquinolone therapy demonstrating that the relative incidence associated with fluoroquinolone exposure varied by timing, type of tendon rupture and concomitant exposure to systemic corticosteroids [3]. Whilst an absolute measure of risk was calculated, the number needed to harm (NNH) may be easier to interpret for policy makers, clinicians and patients. In this regard, the larger the NNH the more patients can on average be treated with fluoroquinolones before one person is harmed. We have therefore calculated the NNH for tendon rupture associated with different durations of fluoroquinolone therapy, overall and stratified by age and gender.

The NNH was calculated using the reciprocal of the adjusted rate difference for tendon rupture per 10,000 person-years of fluoroquinolone exposure and was calculated for different durations of fluoroquinolone therapy (5-, 7-, 10-, 14-, 21- and 28-day courses). These durations were chosen to reflect the variation in clinical practice according to the condition being treated [4]. NNH for different durations of fluoroquinolone therapy varied by age, gender and duration of fluoroquinolone therapy (Table 1). For any tendon rupture: the overall NNH ranged from 44,951 patients (28–day course) to 251,724 patients (5-day course); NNH was lower in women and in those aged ≥ 60 years. A similar distribution was seen with concomitant exposure to fluoroquinolone and systemic corticosteroid therapy but NNH were lower in this case (Table 1). For any tendon rupture associated with concomitant fluoroquinolone and corticosteroid exposure: the overall NNH ranged from 11,959 patients (28–day course) to 66,972 patients (5-day course); NNH was lower in women and in those aged ≥ 60 years. For Achilles tendon rupture, the most common type of tendon rupture associated with the largest increased relative incidence, estimates for NNH followed a similar distribution as for any tendon rupture apart for a lower NNH in men and the NNH being marginally greater (Table 2).

Table 1 Number needed to treat to cause one additional tendon rupture with different durations of fluoroquinolone and fluoroquinolone/corticosteroid exposure
Full size table
Table 2 Number needed to treat to cause one additional Achilles tendon rupture with different durations of fluoroquinolone and fluoroquinolone/corticosteroid exposure
Full size table

Absolute risk estimates are informative for public health and there are limited data on absolute risk of tendon rupture associated with fluoroquinolone therapy. A recently published nested case–control study reported incidence rate differences (IRD) in line with those from our original study but with less precision (IRD 3.73, 95% CI 2.08–5.39 per 10,000 person-years of current exposure for all patients) giving further confidence in the validity of our original findings [5]. Our results show the frequency of tendon rupture events varies across the population depending on several factors: namely duration of therapy, age, gender and concomitant exposure to systemic corticosteroids. The Summary of Product Characteristics (SmPC) is a document that includes information such as how to use a medicine, its method of administration, special warnings and adverse effects. In many fluoroquinolone SmPCs the frequency of tendon rupture is listed as very rare (<1/10,000) but in others it is listed as rare (≥ 1/10,000 to < 1/1000) [6] or frequency unknown [7]. Our results also highlight that absolute risk may vary markedly due to different baseline risk among subpopulations. Further work on understanding how this affects extrapolation of benefit–risk to stratified populations or those commonly excluded from clinical trials, could be valuable.

A strength of this study is that we provide confidence intervals for each NNH. However, other potential subgroups of patients may exist in which the NNH may vary that have not been studied. This may be important if linear relationships between exposure and risk do not exist. Although it is completely appropriate that different indications may require different durations of antibiotic therapy, evidence underpinning the minimum effective duration to successfully treat infections can be limited. From a safety perspective, a medicine should be used for the shortest possible duration at the lowest effective dose. Shortening the duration of antibiotic therapy could therefore reduce the risk of fluoroquinolone-associated adverse effects, a potential area for further evidence generation with antibiotic stewardship interventions [8, 9]. This may also affect other potential adverse reactions associated with fluoroquinolones including antibiotic resistance, peripheral neuropathy and potentially aortic aneurysm [10,11,12,13,14]. The NNH in the highest risk group that we evaluated, those aged ≥ 60 years and prescribed fluoroquinolone therapy for 28 days with concomitant corticosteroid therapy, was 6651 patients that would equate to a frequency of rare in an SmPC [15]. The number of fluoroquinolone products listed across Europe is extensive and the extent of variation in the frequency of tendon ruptures listed in all SmPCs across Europe is uncertain. Despite this, a range of frequencies presented in the SmPC of fluoroquinolone products may better reflect differences in risk compared to a single figure.