Skip navigation
Link to Back issues listing | Back Issue Listing with content Index | Subject Index

PPIs and fractures

PI and hip fracture [1]
PPI and fracture risk [2]
Reflux oesophagitis and vertebral fractures [3]
And finally

Proton pump inhibitors are among the safest drugs we know, and have been used increasingly in older people for conditions like reflux oesophagitis, especially since omeprazole came off patent and prices fell. In recent years the advent of very large databases has increased our ability to examine for rare adverse events, and such an eventuality now impacts on acid suppression, and particularly on proton pump inhibitors (PPIs). Several pieces of information now implicate long term use of PPIs, especially at high dose, on increased risk of osteoporotic fractures, especially hip fracture.

PI and hip fracture [1]

The link between PPI use and increased risk of hip fracture comes from a large case-control study using the UK General Practice Database (UKGPRD). The study cohort was 1.8 million people registered for a year or more in the years 1988-2003, 50 years or older, and without hip fracture before the study started. There were 192,000 users of PPI and 188,000 users of histamine antagonists (H2A) receiving at least one prescription, with 1.4 million non-users of acid suppression medicines.

Cases were people with a first occurrence of hip fracture at least one year after the start of follow up. Up to 10 controls were matched for each case, based on age, sex, and some other relevant factors. Considerably information was collected on cases and controls about demographics, medication use, and health conditions. The primary exposure of interest was PPI therapy of more than one year before the date of fracture.


There were 13,500 hip fractures and 135,000 controls. The average age at enrolment was 77 years, and 80% were women. The crude incidence rate of hip fracture was 4.0 per 1000 person years in those with more than one year of PPI therapy, and 1.8 per 1000 person years in controls. Those with a hip fracture were more likely to have received medicines of have health conditions associated either with osteoporosis of risk of falling, and these criteria were used to adjust the comparisons to remove the effects of these known confounders.

Overall, patients taking PPI or H2A for one year or longer had an increased risk of hip fracture (Table 1). The effect was significantly larger for those taking PPI alone than users of H2A alone. The effect was significantly larger in men than women.

Table 1: Adjusted odds ratios for hip fracture with PPI and H2A

Comparison with non-use
Adjusted odds ratio
(95% CI)
Patients taking only PP for 1 year or more
1.6 (1.4 to 1.9)
Including patients taking both PPI and H2A
1.4 (1.3 to 1.6)
Patients taking H2A for 1 year or more
1.2 (1.1 to 1.4)
Men taking PPI for one year or more
1.8 (1.4 to 2.2)
Women taking PPI for one year or more
1.4 (1.2 to 1.5)

For PPIs especially the effect on hip fracture was related to dose (Figure 1). The average dose was 1.75 doses per day. Those with higher doses (≥1.75 a day) were mostly taking medicines twice a day, whilst 90% of those at lower dose were taking it once a day. For PPIs the effect on hip fracture was related to duration of PPI use (Figure 2).

Figure 1: Effect of PPI and H2A on hip fracture compared with non use by dose

Figure 2: Effect of PPI on hip fracture compared with non use by duration of PPI use

PPI and fracture risk [2]

A large Danish study [2] was one report of several relating drug therapy of various types to fracture risk. It used as cases all those people in Denmark (population 5.3 million) who sustained any fracture during the year 2000, with three controls matched by age and sex chosen randomly. Denmark has an almost unique ability to link various databases together, so information could be obtained from hospital discharge registers and medication use.


There were 125,000 people with fractures and 374,000 controls. The average age was 43 years, about equally split between women and men. There were many differences between people sustaining a fracture and those who did not, in demographics, use of medical services, and drug therapies.

Recent use of PPI (within one year) was associated with increased risk of hip or spine, and any fracture. No increased risk was seen with H2A, with some suggestion of a reduced risk. For PPIs, use within the last year the odds ratios were 1.2 (1.1 to 1.4) for overall fracture risk, 1.5 (1.3 to 1.7) for hip fracture, and 1.6 (1.3 to 2.0) for spine fractures. There was little in the way of dose response, though the effects were not apparent for PPI use more than one year previously.

Reflux oesophagitis and vertebral fractures [3]

A paper from Japan reported a small survey of 75 postmenopausal Japanese women aged 52-97 years of whom 18 had refractory reflux oesophagitis with use of PPI for at least six months for symptom relief. The use of PPI in other women was not reported, but the implication was that PPI was not used because they did not have reflux oesophagitis.

Women with reflux had significantly higher rates of multiple vertebral fractures and hiatus hernia than those without, with an average of 2.5 vertebral fractures (range 0-11) with reflux and 1.1 without. Age adjusted odds ratios were up to 10 for increased frequency of both multiple vertebral fractures and hiatus hernia.

No direct link was made between use of PPI and vertebral fractures in the paper. Rather the link was made between vertebral fractures and the need for PPI.


This is all a bit new, and frankly unexpected. As best Bandolier can find, there is no other evidence from clinical studies, though there is a limited experimental literature. We have two large and well-done observational studies providing much the same answer. Moreover, there is evidence that there is a dose response and a duration response for PPIs.

Nor is the effect trivial. Hip fracture is an important cause of death and morbidity in the over 65s, especially in women. In the mid-1990s in the UK 57,000 people had hip fracture every year, accounting for 20% of all orthopaedic beds, with an average cost of stay of £5,000 and a total cost to the NHS of hospital care alone of about £280 million [4]. We know that mortality immediately after fracture is about 20%. An East Anglian audit [5] found that:

• Fewer than a quarter of patients both survived and returned to their pre-fracture level of function by 90 days.

• Of survivors, less than one-third returned to their pre-fracture level of function.

• Of survivors, 42% were receiving extra help with at least half of their daily living activities.

• Of survivors, 21% required an increased level of residential or hospital care.

• Of patients who returned home 35% required additional community health and social service visits.

The rate of hip fracture in over 65s is higher in women than men, but overall runs at about 1,100 per million. That means that the annual risk of hip fracture in over 65s runs at roughly 1 in 900 (higher in women and at higher age, of course). The risk of dying is about 1 in 4,500. So anything that doubles the risk produces an additional risk of 1 in 900 of a hip fracture and 1 in 4,500 of dying as a result of a hip fracture. This additional risk can be portrayed on a Paling Perspective Scale (Figure 3) for long-term use of a PPI.

Figure 3: Hip fracture risk on Paling Perspective Scale

These are not trivial risks, and are of the same sort of order as others we worry about (and some we don't, because we do not have a particularly coherent view of risk as yet).

This is a time for thought. We have no clear idea about any mechanism, but it is likely to be a combination of reduced calcium absorption and inhibition of ion transport in osteoclasts. Perhaps the main message is that high dose PPIs are not a good idea over the long term, and that we should be more careful about their use, especially in people with risk factors for osteoporosis. With a rapidly growing population of what might be called the older old, this is a topic that needs careful watching.

And finally

Just to make sure that we all realise just how complicated this business of drug therapy and fracture risk is going to be, the Danish group have published two other analyses from their large database [6,7].

On balance, it looks as if certain NSAIDs may be associated with increased risk, and current use of ibuprofen, diclofenac and naproxen (where there were large numbers of users) are implicated as having increased risk (Table 2 shows risk of any fracture). Neither rofecoxib nor celecoxib was associated with any different risk from nonuse, but the bad news was that other commonly used analgesics like paracetamol, tramadol, and morphine agonists were also associated with higher fracture risk. Paracetamol, diclofenac and ibuprofen were consistently associated with increased risk of hip fracture.

Table 2: Association between analgesic drug use and fracture at any site

Number of users
Adjusted odds ratio
(95% CI)
1.76 (1.82 to 1.81)
1.39 (1.35 to 1.44)
1.37 (1.29 to 1.46)
1.02 (0.96 to 1.07)
0.94 (0.84 to 1.04)
1.45 (1.41 to 1.49)
Morphine and morphine agonists
2.20 (2.08 to 2.32)
3.08 (2.99 to 3.17)

There is a really important reminder here. If you don't look, you won't find. For rare but serious adverse events, ignorance is bliss. The reality is that almost all drugs and interventions will have some rare but serious adverse event, but because they are rare we might never make the connection. Over the next decade or so the power of large databases will grow, making it possible to investigate this area more thoroughly than ever before.

Be prepared for lots of bad news like this. It makes our inability to explain and use risk information much more of a handicap than we would have thought a few short years ago.


  1. YX Yang et al. Long-term proton pump inhibitor therapy and risk of hip fracture. JAMA 2006 296:2947-2953.
  2. P Vestergaard et al. Proton pump inhibitors, histamine H2 receptor antagonists, and other antacid medications and risk of fracture. Calcified Tissue International 2006 79: 76-83.
  3. T Yamaguchi et al. Multiple vertebral fractures are associated with refractory reflux oesophagitis in postmenopausal women. Journal of Bone and Mineral Metabolism 2005 23: 36-40.
  4. W Hollingworth et al. The cost of treating hip fractures in the twenty first century. Journal of Public Health Medicine 1995 17:269-276.
  5. C Laxton et al. Morbidity at 3 months after hip fracture: data from the East Anglian audit. Health Trends 1997 29: 55-60.
  6. P Vestergaard et al. Fracture risk associated with use of nonsteroidal anti-inflammatory drugs, acetylsalicylic acid, and acetaminophen and the effects of rheumatoid arthritis and osteoarthritis. Calcified Tissue International 2006 79: 84-94.
  7. P Vestergaard et al. Fracture risk associated with the use of morphine and opiates. Journal of Internal Medicine 2006 260: 76-87.

previous or next story