QPPV.COM

Signal Evaluation: Bisphosphonates and Atypical Femoral Fractures

A historical signal-evaluation case study showing how reports of unusual femoral fractures evolved into evidence of a rare bisphosphonate-associated adverse effect, and how duration, discontinuation, biological plausibility and absolute risk shaped the final benefit-risk assessment.

Audio Lesson 12 min
Knowledge Assessment Test your understanding of this article. Take the assessment →

Signal Evaluation: Bisphosphonates and Atypical Femoral Fractures

Introduction

The association between bisphosphonate treatment and atypical femoral fractures is one of the clearest examples of a pharmacovigilance signal developing from unusual clinical observations into a recognised, rare adverse effect.

It is also an unusually useful example because the signal did not depend on a single dramatic epidemiological finding.

Instead, the evidence accumulated through several complementary routes:

This is precisely the type of signal that should be evaluated as an evolving body of evidence rather than reduced to a single relative-risk estimate.

The case is also important because the signal involved a medicine class with well-established clinical benefit.

Bisphosphonates reduce osteoporotic fracture risk.

Therefore, identifying an uncommon but potentially serious adverse effect did not lead simply to the conclusion:

"Bisphosphonates are unsafe."

The more useful pharmacovigilance question became:

"How large is the atypical-fracture risk, which patients are most susceptible, how does risk change with duration of treatment, and how does that risk compare with the fractures prevented by treatment?"

That is a much more clinically meaningful signal-evaluation question.

1. Define the Signal

The signal can be expressed as:

Bisphosphonate exposure → atypical femoral fracture.

The important refinement is that the event is not simply any fracture of the femur.

The phenotype is a distinctive fracture pattern involving the subtrochanteric region and/or femoral shaft and occurring with little or no trauma.

The definition of the event therefore matters.

A study that simply counts all femoral fractures may mix:

Those events do not represent the same clinical phenomenon.

A central lesson from this signal is therefore:

Before estimating whether a medicine is associated with an adverse event, make sure that the event has been defined with enough clinical precision to represent the suspected phenomenon.

2. What Are Atypical Femoral Fractures?

Atypical femoral fractures are unusual fractures occurring in the subtrochanteric region or femoral shaft.

They differ from the more common hip fractures associated with osteoporosis.

They typically occur after minimal trauma and have characteristic radiographic features.

The precise diagnostic criteria have evolved over time.

Important features include:

The distinction is clinically important because the signal concerns a particular fracture phenotype rather than femoral fractures in general.

3. Why the Signal Was Difficult to Detect Initially

Bisphosphonates are widely used.

Osteoporosis itself is common.

Femoral fractures are also common in older populations.

Therefore, a relatively uncommon drug-associated fracture phenotype can initially be hidden within a large background of fractures.

This creates a classic pharmacovigilance detection problem.

The signal was not simply:

"Patients taking bisphosphonates fracture."

That would have very little specificity.

The more informative observation was:

"Some patients receiving long-term bisphosphonates are developing unusual low-trauma fractures with a distinctive radiographic appearance."

That clinical pattern was much more informative.

4. The Initial Clinical Observations

Reports of unusual femoral fractures in patients receiving long-term bisphosphonate treatment began appearing before the signal was formally characterised.

Early reports and case series described fractures in the subtrochanteric and femoral-shaft regions that differed from typical osteoporotic fractures.

A 2010 report in the New England Journal of Medicine described a series in which 20 of 152 subtrochanteric or shaft fractures were classified as atypical. Seventeen of those patients were receiving oral bisphosphonate treatment, predominantly alendronate. The authors noted that the atypical radiographic pattern appeared highly specific to bisphosphonate-treated patients in their series. 2

This was not definitive causal evidence.

But it was exactly the kind of clinical observation that should trigger formal signal evaluation.

5. Why Case Reports Were Valuable

Case reports cannot establish incidence.

They cannot establish relative risk.

They are also vulnerable to:

Nevertheless, they can be extremely valuable when the adverse event has a distinctive phenotype.

In this case, the radiographic appearance itself was part of the signal.

That makes the case different from a nonspecific symptom such as fatigue, headache or nausea.

The question was not simply whether fractures occurred.

It was whether a characteristic type of fracture was disproportionately represented among patients receiving bisphosphonates.

6. A Key Signal-Evaluation Question: Is the Phenotype Specific?

Suppose a pharmacovigilance database contains:

10,000 reports of femoral fracture.

That alone says very little.

Now suppose investigators identify:

500 reports of unusual low-trauma subtrochanteric or shaft fractures with a consistent radiographic pattern.

The second dataset is much more informative.

This illustrates an important principle:

Signal specificity can be more informative than raw case count.

A well-defined clinical phenotype can make a relatively small number of cases highly informative.

7. The Biological Context

Bisphosphonates inhibit osteoclast-mediated bone resorption.

This reduces bone turnover.

That mechanism is therapeutically useful because excessive bone resorption contributes to osteoporosis.

However, bone remodelling also has physiological functions.

Microdamage accumulates within bone over time.

Normal remodelling helps remove damaged areas and replace them with new bone.

Therefore, prolonged suppression of bone turnover provided a biologically plausible hypothesis:

excessive or prolonged suppression of remodelling → accumulation of microdamage → reduced ability to repair stress injury → atypical fracture.

This mechanism was plausible.

But it is important to distinguish:

mechanistic plausibility

from:

demonstrated clinical causality.

8. What Was Known About the Mechanism?

The pharmacological effect of bisphosphonates on osteoclast-mediated bone resorption was well established.

The proposed connection between prolonged suppression of remodelling and atypical femoral fracture was biologically coherent.

However, the exact pathogenesis of atypical femoral fractures is more complicated.

The fracture is not simply the result of "too little bone turnover."

Other factors may contribute, including:

The precise biological pathway should therefore not be overstated.

A careful signal assessment distinguishes:

Established

Bisphosphonates suppress osteoclast-mediated bone resorption.

Strongly plausible

Long-term suppression of remodelling may impair repair of accumulated microdamage in susceptible bone.

Not fully established

The exact biological sequence through which bisphosphonate exposure produces the characteristic atypical fracture phenotype in an individual patient.

This distinction is important in pharmacovigilance writing.

9. The 2010 Swedish Epidemiological Study

A major advance came from a population-based study in Sweden published by Schilcher and colleagues in the New England Journal of Medicine in 2011.

The investigators examined:

12,777 women aged 55 years or older

who had sustained a femoral fracture in Sweden in 2008.

Radiographs were reviewed to identify atypical fractures.

Among 12,777 women with femoral fractures, 59 were classified as having atypical fractures.

The age-adjusted relative risk of atypical fracture associated with bisphosphonate use was:

47.3 (95% CI 25.6–87.3).

The estimated absolute difference was:

5 additional atypical fractures per 10,000 patient-years.

The corresponding estimated number needed to harm was approximately:

2,000 patients treated for one year for one additional atypical fracture.

The study also found that risk increased with longer duration of use. 3

The relative risk was striking.

The absolute risk was much smaller.

Both findings were important.

10. Why the Relative Risk Was So Large

A very high relative risk can occur when the baseline incidence of an event is extremely low.

For example, moving from:

1 event per 10,000

to:

5 events per 10,000

represents a five-fold increase.

But the absolute increase is:

4 events per 10,000.

Therefore, the magnitude of the relative association should never be interpreted without understanding the background incidence.

This is particularly important in pharmacovigilance because rare-event signals can generate dramatic relative measures.

11. Duration of Treatment

One of the most important findings in the Swedish study was the relationship between duration of bisphosphonate exposure and atypical fracture risk.

The risk increased with increasing exposure.

The study reported an odds ratio of approximately:

1.3 for every additional 100 prescribed daily doses.

The authors found that the risk was substantially higher after longer treatment durations. 4

This is important because a duration-response relationship can strengthen a causal hypothesis.

It does not prove causality.

But when a suspected adverse effect becomes more frequent with increasing cumulative exposure, the finding is generally more supportive of a drug-related mechanism than an association with no relationship to exposure duration.

12. Discontinuation and Dechallenge

Another particularly informative finding was the reduction in risk after discontinuation.

The Swedish investigators reported approximately a:

70% reduction in risk for each year since the last prescription.

The adjusted odds ratio for risk reduction per year after discontinuation was approximately:

0.28 (95% CI 0.21–0.38). 5

This provides an epidemiological analogue of dechallenge.

In an individual adverse-event case, dechallenge asks:

Does the event improve after the suspected drug is stopped?

For a fracture, that formulation is not straightforward because stopping the drug does not immediately repair an established fracture.

Instead, population-level evidence can ask:

Does the incidence of new atypical fractures decline after treatment is discontinued?

The answer appears to be yes.

That strengthens the overall causal interpretation.

13. A Subtle Mechanistic Problem

The Swedish study produced an interesting observation.

Bisphosphonates remain incorporated into bone for prolonged periods.

If atypical fractures were caused simply by the long-term physical presence of bisphosphonate in bone, one might expect risk to decline slowly after discontinuation.

Instead, the study observed a relatively rapid reduction in risk.

The authors therefore noted that this pattern was difficult to reconcile with a simple model based solely on prolonged persistence of the drug in bone.

They proposed that ongoing treatment may influence processes at active microdamage or fracture sites in a way that is more rapidly reversible than the overall skeletal persistence of the medicine. 6

This is an excellent example of how pharmacovigilance evidence can challenge an apparently intuitive mechanism.

The association can be real even when the first mechanistic explanation is incomplete.

14. The EMA Review

The European Medicines Agency reviewed bisphosphonates following accumulating reports and published evidence.

The regulatory history is particularly useful because it shows the signal evolving from a product-specific concern into a class-level conclusion.

In 2008, the Pharmacovigilance Working Party identified an increased risk of atypical femoral fractures with alendronic acid and a warning was added to alendronate product information.

At that stage, the possibility of a class effect could not be excluded.

By April 2010, additional literature and post-marketing evidence supported the possibility that atypical femoral fractures represented a class effect.

The CHMP subsequently concluded that atypical femoral fractures were likely to be a class effect of bisphosphonates. 7

15. Regulatory Action

The EMA concluded that:

For bisphosphonates used in osteoporosis, the regulatory communication also advised periodic review of treatment, particularly after five or more years. 8

This is a classic example of risk minimisation being proportionate to the evidence.

The regulatory response was not:

Stop bisphosphonate treatment.

Instead, it was effectively:

Recognise the rare risk, identify suspected cases, and periodically reassess the need for continuing treatment.

16. Why the Class Effect Matters

The original signal was particularly associated with alendronate.

But the regulatory assessment eventually extended the warning to the bisphosphonate class.

This raises a classic pharmacovigilance question:

When should evidence from one active substance be extrapolated to a class?

The answer depends on several considerations.

These include:

The EMA concluded that the totality of evidence supported treating atypical femoral fractures as a class effect. 9

17. The Importance of Avoiding Over-Generalisation

A class effect does not mean:

all bisphosphonates have identical risk.

Nor does it mean:

every patient receiving any bisphosphonate has the same susceptibility.

The evidence supports a class-level safety concern.

The magnitude of risk may still vary according to:

This distinction matters when translating regulatory findings into clinical practice.

18. Risk Factors Beyond Drug Exposure

The later literature identified additional risk factors.

In a large 2020 prospective cohort study published in the New England Journal of Medicine, among 196,129 women, 277 atypical femur fractures occurred.

After multivariable adjustment, important risk factors included:

This is important because it demonstrates that the drug exposure is not the entire risk model.

The adverse event emerges from an interaction between:

drug exposure + patient susceptibility + skeletal/mechanical factors + duration.

19. The 2020 Large Cohort Study

The 2020 study provided particularly useful evidence because the investigators directly adjudicated fractures radiographically.

This is important.

Administrative coding alone can misclassify atypical fractures.

Radiographic review provides greater confidence that the outcome being analysed actually corresponds to the phenotype under investigation.

Among 196,129 women, 277 atypical femur fractures were identified.

The adjusted hazard ratio increased substantially with longer bisphosphonate exposure.

Compared with less than three months of treatment:

This provides strong evidence for a duration-response relationship.

20. But the Absolute Risk Remained Low

The same study provides the critical counterweight.

Although relative risk increased dramatically with prolonged treatment, the absolute risk of atypical fracture remained low.

The investigators modelled the numbers of atypical fractures associated with treatment against osteoporotic fractures prevented.

Among White women, after three years of treatment, they estimated:

149 hip fractures prevented

for approximately:

2 bisphosphonate-associated atypical fractures.

Among Asian women, the balance was less favourable:

91 hip fractures prevented

for approximately:

8 atypical fractures.

The authors nevertheless concluded that the absolute risk of atypical femur fracture remained very low compared with the reduction in osteoporotic fractures, although the risk-benefit balance appeared less favourable in Asian women. 12

This is exactly the type of evidence required for a mature benefit-risk assessment.

21. Relative Risk Can Increase While Absolute Risk Remains Small

The 2020 study demonstrates a fundamental statistical principle.

As treatment duration increases:

relative risk can become very large

while:

absolute event frequency remains low.

This is not contradictory.

It simply reflects the low baseline incidence of atypical femoral fractures.

For pharmacovigilance communication, both measures should therefore be considered.

A statement such as:

"Long-term bisphosphonate treatment increases the risk of atypical femoral fracture."

is incomplete.

A more informative statement is:

"Risk increases with longer treatment, but atypical femoral fractures remain rare compared with the osteoporotic fractures prevented by treatment."

That is closer to the actual clinical decision.

22. The Importance of Treatment Duration

Duration became one of the most clinically useful features of the signal.

The evidence increasingly suggested that:

short exposure → lower atypical-fracture risk

and:

prolonged exposure → substantially higher relative risk.

This has implications for treatment review.

A patient who has received a bisphosphonate for several months is not necessarily in the same risk category as a patient who has received treatment continuously for many years.

The signal therefore evolved from:

drug → event

to:

drug + duration → event.

That is a much more clinically useful risk model.

23. Risk Reduction After Discontinuation

The 2020 cohort also found that the risk declined relatively quickly after bisphosphonate discontinuation.

Compared with recent discontinuation, the hazard ratio decreased to approximately:

0.52

after more than three months to fifteen months since discontinuation.

In later periods, the risk reduction was approximately:

74–79%. 13

This finding is consistent with the earlier Swedish evidence.

It strengthens the view that ongoing exposure contributes to the risk.

Again, this does not mean that every fracture risk disappears immediately after stopping therapy.

It means that the population-level incidence of atypical fractures decreases after treatment cessation.

24. Contralateral Fractures

An important clinical feature of atypical femoral fractures is that the other femur may also be affected.

This became part of the regulatory communication.

The EMA advised that if an atypical fracture is suspected in one leg, the other leg should also be examined. 14

This is a good example of how signal evaluation can generate a practical risk-management intervention.

The scientific observation:

bilateral susceptibility

becomes the clinical recommendation:

evaluate the contralateral femur.

25. Prodromal Pain as a Signal

Patients with atypical femoral fractures may experience thigh or groin pain before complete fracture occurs.

This is clinically important because it provides an opportunity for earlier investigation.

From a pharmacovigilance perspective, it also illustrates the difference between:

the adverse event

and:

a clinical feature that may precede the adverse event.

A safety warning that simply says "fracture" is less useful than one that helps clinicians recognise a possible impending event.

26. A Hypothetical Case

Consider a 72-year-old woman with osteoporosis.

She has received oral bisphosphonate treatment for seven years.

She develops persistent unilateral thigh pain without a fall.

An initial examination does not identify a clear traumatic injury.

The clinician obtains femoral imaging.

Imaging shows cortical thickening and an incomplete transverse fracture in the subtrochanteric region.

How should this be interpreted?

The patient has several features that increase suspicion:

The signal hypothesis is therefore clinically relevant.

The next step is not simply to record:

"Bisphosphonate-associated fracture."

The evaluator should document the evidence supporting and opposing the causal attribution.

27. A Stronger Case

Consider the same patient with:

The causal hypothesis is substantially stronger.

The pattern is highly compatible with the recognised adverse-event phenotype.

This is a good example of how:

clinical phenotype + exposure duration + biological plausibility + absence of strong alternatives

can provide strong case-level evidence.

28. A Confounded Case

Now consider another patient.

She has been receiving a bisphosphonate for four years.

She develops a femoral fracture after a high-energy motor-vehicle accident.

The fracture is comminuted and occurs in a location typical of traumatic fracture.

In this situation, the mere fact of bisphosphonate exposure provides little evidence for an atypical-fracture signal.

The exposure is present.

The event is present.

But the phenotype does not fit well.

This illustrates an important principle:

Exposure plus event does not equal signal-compatible event.

Clinical phenotype must be considered.

29. A Borderline Case

Consider a patient with:

This case should not automatically be classified as definitive.

Instead, the evaluator should seek:

The correct conclusion may initially be:

The case is compatible with an atypical femoral fracture, but available information is insufficient for definitive classification.

This is preferable to forcing a binary yes/no conclusion.

30. Signal Strengthening Factors

Evidence that strengthens the causal hypothesis includes:

The combination is considerably stronger than any individual factor.

31. Signal Weakening Factors

Evidence that weakens the causal hypothesis includes:

This illustrates why case-level signal assessment should not rely on exposure alone.

32. The Importance of Radiographic Adjudication

One of the methodological strengths of later studies was direct radiographic review.

Administrative databases may identify:

femoral fracture.

But that does not necessarily identify:

atypical femoral fracture.

Radiographic adjudication improves specificity.

This is particularly important when:

A useful general principle follows:

The rarer and more phenotype-specific the suspected adverse event, the more important accurate outcome classification becomes.

33. Why the Signal Was Not Simply "Confounding"

The presence of confounding does not mean the association was false.

The later evidence demonstrated that patient characteristics and treatment duration influence risk.

But the association persisted after multivariable adjustment.

The 2020 cohort, for example, found strong duration-related associations even after adjustment for multiple clinical variables. 15

Therefore, the mature interpretation is not:

"The early signal was just confounding."

It is:

"The association is supported by multiple lines of evidence, while its magnitude and susceptibility are influenced by duration and patient characteristics."

That is a much more defensible conclusion.

34. Signal Evolution

The historical development can be summarised as follows:

Period Evidence Signal interpretation
Early reports Unusual low-trauma femoral fractures Clinical signal generated
2008 Alendronate-specific regulatory review Warning added; class effect not yet established
2009–2010 Case series and accumulating literature Stronger association with distinctive fracture phenotype
2010–2011 Population-based epidemiology Association demonstrated; duration-response and absolute risk quantified
2011 EMA class review Atypical femoral fractures considered likely a bisphosphonate class effect
2010s Additional epidemiology Duration, discontinuation and patient susceptibility increasingly characterised
2020 Large radiographically adjudicated cohort Strong duration-response; absolute risk remained low relative to fractures prevented
Mature interpretation Totality of evidence Rare, clinically important adverse effect with duration-dependent risk and favourable overall benefit-risk in appropriate patients

This timeline illustrates how a pharmacovigilance signal becomes progressively more precise.

35. From Signal to Risk Minimisation

The evidence eventually translated into several practical measures.

These included:

The purpose of risk minimisation is not simply to announce that a risk exists.

It is to make the risk actionable.

36. What the Regulatory Decision Did Not Mean

The EMA did not conclude that bisphosphonates should generally be discontinued.

The Agency explicitly concluded that the benefits continued to outweigh the risks. 16

This distinction is fundamental.

A safety signal can be:

real

without making the medicine:

unacceptable.

Benefit-risk assessment asks whether the magnitude and clinical consequences of the adverse effect outweigh the benefits of treatment in the population and, where possible, in individual patients.

37. Benefit-Risk Is the Central Question

Bisphosphonates prevent osteoporotic fractures.

Those fractures can be common and clinically serious.

Atypical femoral fractures are rare.

Therefore:

rare serious harm

must be compared with:

common serious disease-related harm prevented by treatment.

This is the correct pharmacovigilance framework.

The question is not:

"Does the drug have a risk?"

Almost every effective medicine does.

The question is:

"What is the magnitude of the risk, who is exposed to it, how can it be reduced, and how does it compare with the benefit?"

38. A Simple Numerical Illustration

Imagine a hypothetical population in which treatment prevents:

100 osteoporotic fractures

while contributing to:

2 atypical femoral fractures.

It would be incorrect to conclude that the treatment is unsafe simply because the adverse event exists.

The relevant question is whether those 100 prevented fractures represent sufficient clinical benefit to outweigh the 2 additional atypical fractures.

The actual answer depends on:

This is why benefit-risk assessment cannot be reduced to a single adverse-event frequency.

39. Duration Changes the Benefit-Risk Balance

The benefit-risk balance can change over time.

Early in treatment:

After many years:

This creates a rationale for periodic treatment review.

The signal therefore changed the clinical question from:

"Should bisphosphonates be used?"

to:

"For this patient, should treatment continue for another period of time?"

That is a much more sophisticated risk-management question.

40. The Role of Treatment Holidays

The concept of a treatment holiday emerged partly from the broader question of long-term bisphosphonate therapy.

A treatment holiday is not synonymous with:

"The drug is dangerous."

Rather, it reflects the fact that:

The atypical-fracture signal therefore became one component of a broader treatment-duration decision.

41. What Is Established?

Established

Strongly supported but more nuanced

Not fully established

42. What Would Strengthen the Mechanistic Hypothesis?

Mechanistic evidence would be strengthened by findings showing:

However, mechanistic studies should complement rather than replace epidemiological evidence.

The clinical question is ultimately whether patients exposed to the medicine experience the adverse event at an increased frequency.

43. What Would We Look for in Future Signal Evaluation?

If a similar signal emerged today, a QPPV or signal evaluator should consider:

Case series

Epidemiology

Dechallenge

Susceptibility

Mechanism

Regulatory evidence

Benefit-risk

44. A QPPV Perspective

A QPPV reviewing a signal such as this should avoid two opposite errors.

Error 1: Dismissing the signal because the event is rare

A rare event can be clinically important if:

Error 2: Overreacting to a high relative risk

A relative risk of 40 does not mean that 40% of treated patients will experience the event.

The background incidence must be known.

The absolute excess risk may remain small.

The correct approach is therefore:

severity + specificity + relative risk + absolute risk + duration + susceptibility + benefit

rather than any single measure.

45. Why This Signal Is Particularly Useful for Teaching

This case teaches several core pharmacovigilance concepts simultaneously.

1. Case reports can identify a new phenotype

The signal began with unusual clinical observations.

2. Phenotype definition matters

Atypical femoral fractures are not synonymous with all femoral fractures.

3. Epidemiology can quantify the signal

Population-based studies established the magnitude of association.

4. Duration-response can strengthen causality

Risk increased with longer exposure.

5. Dechallenge can provide supporting evidence

Risk declined after discontinuation.

6. Mechanistic plausibility is supportive but not definitive

Bone-remodelling suppression provides a plausible explanation, but the precise mechanism is more complex.

7. Regulatory action can evolve

The concern moved from alendronate to a likely class effect.

8. Risk does not automatically overturn benefit

The overall benefit-risk balance remained favourable.

This makes the case particularly valuable for training people in real-world signal evaluation.

46. The Historical Product-Information Question

A useful historical exercise is to compare product information over time.

The evaluator should examine:

early product information

versus:

post-signal product information

and identify:

The exact wording should always be verified against the historical version of the relevant SmPC.

It should not be reconstructed from current wording.

47. Why Historical SmPC Comparison Matters

The SmPC is not merely a static document.

It can provide a record of how a safety concern was translated into regulatory knowledge.

For this signal, the historical sequence is particularly instructive:

clinical observation

→

substance-specific concern

→

additional evidence

→

class-effect assessment

→

warning

→

duration and clinical management considerations.

This is exactly the type of evolution that a historical signal-evaluation series should document.

48. Signal Evaluation Versus Signal Detection

The initial reports detected a possible problem.

The subsequent work evaluated it.

Those are different activities.

Signal detection asks:

Is there something unusual that warrants investigation?

Signal evaluation asks:

What does the totality of evidence tell us about whether the medicine is actually associated with the event, how strong the association is, who is at risk, and what action is appropriate?

This distinction is central to pharmacovigilance practice.

49. A Practical Signal-Evaluation Framework

For future historical signal articles, the following framework can be applied consistently.

Step 1 — Define the drug-event pair

What exactly is the suspected exposure and outcome?

Step 2 — Define the phenotype

What does a genuine case look like?

Step 3 — Identify the first credible signal

What observation caused concern?

Step 4 — Establish the background rate

How common is the event without exposure?

Step 5 — Review epidemiological evidence

What are the relative and absolute risks?

Step 6 — Examine confounding

Which factors could create or exaggerate the association?

Step 7 — Examine exposure-response

Does risk change with dose, duration or timing?

Step 8 — Examine dechallenge/rechallenge

Does risk change after stopping or restarting exposure?

Step 9 — Assess biological plausibility

Does the proposed mechanism make sense?

Step 10 — Assess alternative explanations

What else could cause the event?

Step 11 — Review regulatory evolution

What did regulators conclude, and when?

Step 12 — Assess risk management

What actions were taken?

Step 13 — Perform benefit-risk assessment

What harm is caused by the medicine, and what harm does it prevent?

Step 14 — State uncertainty explicitly

What remains unknown?

This framework should be reusable across the Signal Evaluation series.

50. Overall Signal Assessment

The totality of evidence supports a causal association between bisphosphonate exposure and atypical femoral fractures.

The evidence is stronger than a simple spontaneous-report signal because several independent lines of evidence converge:

At the same time, the magnitude of risk must not be exaggerated.

The event is rare.

The relative risk can be very high because the baseline incidence is low.

The absolute excess risk remains small compared with the number of osteoporotic fractures prevented by treatment in many populations.

The risk is also not static.

It increases with prolonged treatment and decreases after discontinuation.

Patient characteristics can modify the balance.

The most defensible overall conclusion is therefore:

Bisphosphonate treatment is causally associated with rare atypical femoral fractures, particularly with prolonged exposure. The association is supported by characteristic clinical and radiographic findings, epidemiological evidence, duration-response relationships and decreasing risk after discontinuation. The exact biological mechanism is not completely established, but suppression of bone remodelling provides a plausible explanation. Despite the increased relative risk, the absolute risk remains low compared with the fractures prevented by bisphosphonate treatment in appropriately selected patients. Consequently, the signal supports recognition, investigation and periodic reassessment of prolonged therapy rather than routine abandonment of treatment.

51. Final Perspective

The history of bisphosphonate-associated atypical femoral fractures demonstrates what a mature pharmacovigilance signal looks like.

It did not begin with a large randomised trial.

It began with clinicians noticing something unusual.

The unusual phenotype led to case reports.

Case reports led to epidemiological investigation.

Epidemiology established an association.

Duration-response strengthened the causal hypothesis.

Discontinuation provided additional supporting evidence.

Mechanistic reasoning provided biological plausibility.

Regulators assessed whether the finding represented a class effect.

Finally, benefit-risk analysis determined that the medicine's established fracture-prevention benefits continued to outweigh the rare adverse event for appropriate patients.

The lesson is therefore broader than bisphosphonates.

A good signal evaluation does not ask only:

"Is the drug associated with the event?"

It asks:

  1. What exactly is the event?
  2. How distinctive is the phenotype?
  3. What was the first credible observation?
  4. What does the epidemiology show?
  5. What is the absolute risk?
  6. Does risk change with duration?
  7. Does risk fall after discontinuation?
  8. What alternative explanations exist?
  9. What is established about the mechanism?
  10. What remains uncertain?
  11. How did regulators respond?
  12. How did product information evolve?
  13. What risk-minimisation measures followed?
  14. What benefits does the medicine provide?
  15. Does the totality of evidence change clinical management?

That is the difference between reporting a safety signal and evaluating one.

References

  1. Schilcher J, Michaëlsson K, Aspenberg P. Bisphosphonate Use and Atypical Fractures of the Femoral Shaft. N Engl J Med. 2011;364:1728-1737. doi:10.1056/NEJMoa1010650.

  2. Dell RM, Adams AL, Greene DF, et al. Incidence of atypical nontraumatic diaphyseal fractures of the femur. J Bone Miner Res. 2012;27:2544-2550.

  3. Shane E, Burr D, Ebeling PR, et al. Atypical subtrochanteric and diaphyseal femoral fractures: second report of a task force of the American Society for Bone and Mineral Research. J Bone Miner Res. 2014;29:1-23.

  4. Girgis CM, Seibel MJ. Atypical femoral fractures and bisphosphonate use. N Engl J Med. 2010;362:1848-1849.

  5. European Medicines Agency. Bisphosphonates - referral. Committee for Medicinal Products for Human Use review concerning atypical femoral fractures.

  6. European Medicines Agency. European Medicines Agency concludes class review of bisphosphonates and atypical fractures. 2011.

  7. Black DM, Geiger EJ, Eastell R, et al. Atypical femur fracture risk versus fragility fracture prevention with bisphosphonates. N Engl J Med. 2020;383:743-753. doi:10.1056/NEJMoa1916525.

  8. Kharazmi M, Hallberg P, Schilcher J, et al. Mortality after atypical femoral fractures: maturing evidence concerning a rare adverse effect.

  9. Shane E, Burr D, Abrahamsen B, et al. Atypical subtrochanteric and diaphyseal femoral fractures: report of a task force of the American Society for Bone and Mineral Research. J Bone Miner Res. 2010;25:2267-2294.

  10. European Medicines Agency. Guideline on good pharmacovigilance practices (GVP) Module IX — Signal Management.

  11. European Medicines Agency. Guideline on good pharmacovigilance practices (GVP) Module V — Risk Management Systems.

  12. European Medicines Agency. Guideline on good pharmacovigilance practices (GVP) Module VII — Periodic safety update report.

  13. European Medicines Agency. Relevant European product information and regulatory assessment materials for bisphosphonate-containing medicines concerning atypical femoral fractures.

  14. U.S. Food and Drug Administration. Bisphosphonates and atypical subtrochanteric femoral fractures: regulatory safety communication and product-labelling information.

Revision History

Related Articles

Evaluating the Fluoroquinolone–Aortic Aneurysm and Dissection Signal: From Signal Detection to Regulatory Action

A detailed evaluation of how the fluoroquinolone–aortic aneurysm and dissection signal evolved from an epidemiological concern into regulatory action, including the evidence that supported the signal and later studies that challenged its causal interpretation.

Signal Evaluation: Fluoroquinolones and Tendon Injury and Tendon Rupture

A historical signal-evaluation case study showing how fluoroquinolone-associated tendon injury evolved from clinical reports into a well-characterised safety concern involving timing, Achilles tendon injury, corticosteroid exposure, patient susceptibility and class-wide regulatory restrictions.

Signal Evaluation: GLP-1 Receptor Agonists and Acute Pancreatitis

A practical signal-evaluation case study showing how the pancreatitis signal associated with GLP-1 receptor agonists evolved from spontaneous reports and mechanistic concern to extensive clinical and epidemiological evaluation, and why the final interpretation requires careful separation of association, causality and class effect.