SEARCH

SEARCH BY CITATION

Keywords:

  • NSAIDs;
  • Fracture healing;
  • Opioids;
  • Confounding

Abstract

  1. Top of page
  2. Abstract
  3. INTRODUCTION
  4. SUBJECTS AND METHODS
  5. RESULTS
  6. DISCUSSION
  7. REFERENCES

Objective

To analyze the relationship between nonunion of humeral shaft fractures and nonsteroidal antiinflammatory drug (NSAID) exposure in older adults.

Methods

A cohort of 9,995 patients with humeral shaft fractures was identified using diagnosis and procedure codes from a Medicare database of >500,000 patients. Prescription NSAID as well as prescription opioid use was assessed from pharmacy claims data for 3 30-day periods immediately after the initial fracture. Nonunion was defined by the presence of procedure codes for repair of nonunion 90–365 days after the index fracture. We examined the association between NSAIDs and nonunion using multivariate Cox proportional hazards models.

Results

Of the 9,995 humeral shaft fractures, 105 patients developed nonunions (1.1%), and 1,032 (10.3%) were exposed to NSAIDs in the 90 days after fracture. NSAID exposure within the first 90 days was significantly associated with nonunion (relative risk [RR] 3.7, 95% confidence interval [95% CI] 2.4–5.6). When indicators for exposure to NSAIDs during each of the 3 30-day windows were placed into the same multivariate model, only the period 61–90 days postfracture was significantly associated with nonunion (RR 3.9, 95% CI 2.0–6.2). We observed a similar association between opioids and nonunion, with exposure to opioids between 61 and 90 days associated with nonunion (RR 2.7, 95% CI 1.5–5.2), but exposure to opioids during neither of the 2 earlier 30-day periods significantly associated with nonunion.

Conclusion

We found that exposure to nonselective NSAIDs or opioids in the period 61–90 days after a humeral shaft fracture was associated with nonunion. Although these associations may be causal, they are more likely to reflect the use of analgesics by patients with painful nonhealing fractures.


INTRODUCTION

  1. Top of page
  2. Abstract
  3. INTRODUCTION
  4. SUBJECTS AND METHODS
  5. RESULTS
  6. DISCUSSION
  7. REFERENCES

Cyclooxygenase-2 enzyme activity is required for fracture healing (1). Several animal studies have documented that high-dose nonselective nonsteroidal antiinflammatory drugs (NSAIDs), medications that inhibit cyclooxygenase, impair fracture healing (2–4). However, NSAIDs are widely used by trauma patients and there have been no consistent findings suggesting that these drugs inhibit fracture healing in humans (5, 6). Prior studies have been limited by small sample sizes, referral populations, dependence on self report for NSAID exposure, and inconsistent definitions of nonunion.

Important clinical controversy exists regarding the safety of NSAIDs in the setting of healing fractures. The goal of the present study was to examine the relationship between nonselective NSAIDs and nonunion of fractures in older adults. To ensure adequate numbers of nonunions and unbiased reporting of NSAID exposure, we studied this question using a large community-based population where information on health care and pharmacy claims can be linked.

SUBJECTS AND METHODS

  1. Top of page
  2. Abstract
  3. INTRODUCTION
  4. SUBJECTS AND METHODS
  5. RESULTS
  6. DISCUSSION
  7. REFERENCES

Study participants and data sources.

Subjects were patients enrolled in Medicare and living in 2 US states between 1995 and 2000. All subjects must also have been enrolled in a state-run pharmacy benefit program during this period. From this group of over 500,000 patients, a cohort of 9,995 patients with an inpatient or outpatient diagnosis of humeral shaft fracture, the index fracture, was identified (International Classification of Diseases, Ninth Revision codes 812.20, 812.21, 812.30, and 812.31). The study database contains diagnosis and procedure codes, dates of service, and provider codes for inpatient (Part A) and ambulatory (Part B) Medicare. Subjects were required to demonstrate health care system use by filling a prescription or a physician or hospital visit within 6 months prior to the date of the index fracture.

Humeral shaft fractures and nonunions.

We chose humeral shaft fracture as a model because the treatment is relatively standard. Humeral shaft fracture is common among elderly individuals and generally treated without open procedures (7). Nonunions of humeral shaft fractures are treated surgically with only rare exception (8), and thus easily identified. Nonunion cases were identified by procedure codes for surgical fixation of the humerus occurring 90–365 days after the index fracture (Current Procedural Terminology codes 24430, 24435, 24515, 24516).

Medication exposure.

Use of prescription NSAIDs and opioid analgesics was tracked by prescription claims from the pharmacy benefits information. These claims data include the prescription names, doses, quantity dispensed, and day supply. Selective cyclooxygenase-2 inhibitors were not included in this analysis because too few patients with humerus fractures were exposed to these agents. Ninety days of exposure was selected as the study period because the risk of NSAID use is purported to occur in the early stages of fracture healing, and the majority of fractures are healed within 3 months (3, 5). The exposure period was broken down into 3 consecutive 30-day periods following the index fracture date: 1–30 days, 31–60 days, and 61–90 days. We required patients to have at least 10 days of medication in a given 30-day period to be considered exposed to a given medicine. For example, if a patient received a 30-day supply of an NSAID on day 20 after the index fracture, then the patient had NSAID exposure in both the 1–30-day period and the 31–60-day period. Exposure to opioids within these windows was defined in a similar manner. Antitussive agents were not considered part of the opioid analgesics.

Statistical analysis.

Nonunions of humeral shaft fractures were identified. Baseline characteristics of subjects with and without nonunions were compared. Exposure patterns to NSAIDs and opioids were assessed in both groups. We examined the relative risk associated with NSAID or opioid exposure and nonunion by calculating hazards ratios in Cox proportion hazard models. Initially, the 90-day period was considered a continuous period. The period was then broken into the 3 30-day periods that were considered individually in multivariate models. Then, all 3 were placed in the same model. Although many patients were exposed in multiple 30-day periods, not all were exposed in consecutive periods. By placing all 3 periods into the model as indicator terms, one can determine the independent effects of each 30-day window. Covariates included in multivariate models were sex, age, history of osteoporosis, diabetes, comorbid medical conditions, and nursing home residence. These were defined based on demographic, health care, and pharmacy claims data. All analyses were performed using SAS Statistical Software, version 8.2 (SAS Institute, Cary, NC).

RESULTS

  1. Top of page
  2. Abstract
  3. INTRODUCTION
  4. SUBJECTS AND METHODS
  5. RESULTS
  6. DISCUSSION
  7. REFERENCES

We identified 9,995 patients with an index humerus fracture and 105 (1.1%) had procedures for nonunion (see Table 1). The majority of patients were women with a mean age of 77 years. Most patients in the study sample were white, approximately one-quarter had diabetes mellitus, 10% had a diagnosis of osteoporosis, and the average number of comorbid conditions was ∼2.0. Nursing home residency was less common among patients who experienced nonunions.

Table 1. Characteristics of study cohort
 Nonunion n = 105Healed fracture n = 9,890P*
  • *

    Student's t-test for continuous variables and chi-square test for categorical variables.

Age, mean ± SD, years78 ± 6.377 ± 6.50.29
Female sex, no. (%)94 (89)8691 (87)0.60
Race, no. (%)   
 White98 (93)9,093 (91)0.60
 Other7 (7)797 (8) 
History of diabetes mellitus, no. (%)30 (28)2,548 (25)0.51
History of osteoporosis, no. (%)10 (9)1,297 (13)0.27
Nursing home residency, no. (%)9 (9)1,704 (17)0.02
Comorbidity score, mean ± SD1.8 ± 1.71.8 ± 1.30.29

NSAID use was common in the study cohort; 1,032 (10.3%) patients in both groups used NSAIDs within the 90 days after the index fracture. Many patients used NSAIDs in several of the 30-day periods immediately after their index humerus fracture (see Table 2). NSAID use in the 90 days after the index humeral fracture was more common in patients with nonunion than in patients with healed fractures.

Table 2. Patterns of prescription NSAID use*
 NonunionHealed fracture
  • *

    Data presented as no. (%). NSAID = nonsteroidal antiinflammatory drug.

Any NSAID use33 (31)999 (10)
0–30 days1 (1)157 (2)
31–60 days1 (1)131 (1)
61–90 days8 (8)205 (2)
0–60 days2 (2)105 (1)
0–30 and 61–90 days0 (0)21 (0.2)
31–90 days10 (10)195 (2)
0–90 days11 (11)185 (2)

Patients using NSAIDs within the 90 days after fracture had an unadjusted relative risk (RR) of nonunion of 3.7 (95% confidence interval [95% CI] 2.4–5.6) (see Table 3). Exposure to NSAIDs within any of the 3 30-day exposure periods was associated with an increased unadjusted risk of nonunion. Exposure during the period 61–90 days had the highest risk (RR 5.2, 95% CI 3.4–8.0).

Table 3. Unadjusted relative risk of humeral shaft nonunion in subjects using prescription NSAIDs or opioids*
Exposure after index humeral fractureRelative risk (95% CI)
  • *

    NSAID = nonsteroidal antiinflammatory drug; 95% CI = 95% confidence interval.

NSAID use within 1–90 days3.7 (2.4–5.6)
NSAID period, days 
 1–302.9 (1.6–5.1)
 31–604.0 (2.6–6.4)
 61–905.2 (3.4–8.0)
Opioid use within 1–90 days1.6 (1.1–2.5)
Opioid period, days 
 1–301.3 (0.8–2.1)
 31–601.7 (1.0–2.9)
 61–902.6 (1.6–4.2)

To control for the independent contribution of each exposure period, all 3 exposure periods were entered into the same multivariate model. When NSAID use during each of the 3 30-day exposure periods was included in the same adjusted model, only NSAID use during the period 61–90 days postfracture proved to be significant: RR 3.9, 95% CI 2.0–6.2 (Figure 1). The RR of nonunion for NSAID exposure during the periods 1–30 days and 31–60 days after the index fracture were no longer significantly elevated.

thumbnail image

Figure 1. Multivariate analysis of nonselective nonsteroidal antiinflammatory drugs (NSAIDs) and nonunion risk. Data are controlled for age, sex, nursing home status, diabetes, osteoporosis, and comorbidity index. Error bars represent 95% confidence intervals.

Download figure to PowerPoint

To assess whether NSAID use in the period 61–90 days postfracture was aimed to treat the pain of nonunion, we investigated the pattern of opioid use in similar 30-day periods. Opioid use in the period 61–90 days postfracture was also significantly associated with nonunion: RR 2.7, 95% CI 1.5–5.2 (see Figure 2). The 2 earlier opioid exposure periods were not associated with nonunion.

thumbnail image

Figure 2. Multivariate analysis of opioid use and nonunion risk. Data are controlled for age, sex, nursing home status, diabetes, osteoporosis, and comorbidity index. Error bars represent 95% confidence intervals.

Download figure to PowerPoint

DISCUSSION

  1. Top of page
  2. Abstract
  3. INTRODUCTION
  4. SUBJECTS AND METHODS
  5. RESULTS
  6. DISCUSSION
  7. REFERENCES

Because nonunion after fractures is a rare event, our study was designed to use a large health care claims database to investigate the relationship between NSAID use and nonunion. Because nonunion is defined as the absence of healing after a set duration, it is difficult to determine exactly when a nonunion began. This problem exists in our health care claims database as well as in the medical records. Moreover, clinical datasets that may be able to date nonunions more precisely are not large enough to study this rare outcome. This ambiguity with “dating” the endpoint makes estimating the association between a drug exposure and nonunion problematic.

In this study of older adults, we found no relationship between nonselective NSAID use in the first 60 days after humerus fracture and nonunion. The use of NSAIDs 61–90 days after fracture was associated with an increased risk of nonunion, and the increased risk was also observed for patients exposed to opioid analgesics in the same period 61–90 days after humerus fracture.

We doubt that this association between late use of NSAIDs and nonunion is causal. First, this effect is inconsistent with what is known about the biology of fracture healing. Cyclooxygenase activity is strongly upregulated in the first 14 days postfracture, and by 21 days returns to normal; even animals treated with high-dose NSAIDs will have fracture healing by 35 days (3). Second, there are no data suggesting that opioids retard fracture healing. For these 2 reasons, we strongly believe that the associations that we have described are due to protopathic bias (9), where NSAID and opioid analgesic use are markers for patients in pain from unstable fractures. Protopathic bias is essentially confusing the cause and effect, such that the drug-outcome relationship is reversed. In our example, a nonhealing fracture may prompt analgesia with an NSAID or an opioid and subsequently the patient undergoes a procedure to stabilize the nonunion. In our claims database, we cannot date the onset of the nonunion and thus late analgesic use might appear as the “cause” of nonunion.

Few human studies have investigated the relationship between NSAIDs and nonunions. Giannoudis et al retrospectively reviewed the medical records of 377 patients with femur fractures (5). Twenty of 32 patients with nonunion took NSAIDs at some point in their postoperative course (odds ratio 10.7). However the time course and dose of NSAID were not reported. It is possible that patients with nonunions experienced greater pain and may have been using NSAIDs to treat nonunions after they were beginning to develop. Bhandari and colleagues investigated the predictors of reoperation after tibia fractures and found that NSAIDs did not prove to be a significant risk factor for nonunion in the multivariate analysis (6). In a recent prospective randomized study of heterotopic ossification after acetabulum fractures, Burd and colleagues noted that the group receiving high-dose indomethacin for 6 weeks had an increased rate of long bone nonunion (10). However, this study was limited to patients at risk for heterotopic ossification and examined only one NSAID, high-dose indomethacin. Moreover, the previous clinical literature did not address the time course of NSAID use; thus, prior findings may be documenting the same phenomenon: NSAIDs are being used for the pain of nonhealing fractures, not that NSAIDs cause nonunions.

As noted by Einhorn, although the laboratory evidence of inhibition of fracture healing is strong, the widespread use of NSAIDs by patients with fractures has not manifested a clinical problem with nonunion (11). This may be due to differences in the biology of human fractures, or due to the relative doses of NSAIDs used in animals as compared with normal prescription doses. Because nonunion is a rare event (∼1%) with modern treatment regimens, adequately powered studies must be very large. We present the largest series to date with the greatest number of nonunions.

Our study has several limitations. First, we do not have detailed clinical data on the patients, such as fracture configuration or pain intensity. However, clinical databases that may contain such information usually rely on patient report to determine drug utilization. Because patient self report of medication use is unreliable, this poses its own set of problems. Second, we can only document prescription NSAID use. In low-income elderly patients with full drug benefits, prescription NSAID use (versus over-the-counter) is the presumption. However, if control patients were using over-the-counter NSAIDs, this would tend to bias the data toward the null. Third, our data do not allow us to know that patients actually took the medications, only that they filled a prescription. Fourth, there is little data on the biology of established nonunions. It may be that nonselective NSAID use 61–90 days postfracture does indeed have a strong detrimental effect on established nonunions. Finally, it may be that both NSAIDs and opioids are inhibitors of fracture healing through pain reduction and enhanced movement of the affected limb. Movement of a fractured humerus may potentiate nonunions. We cannot rule this out as a possible explanation of our findings.

In summary, the association between NSAIDs and nonunions is complex. Although crude analyses suggest that NSAID exposure is associated with nonunions, analysis of the time course suggests that it is use of NSAIDs late in fracture healing that is most strongly associated with nonunion. Because a similar association is observed with opioids, a drug category without any known effects on fracture healing, it is probable that NSAIDs are being used to treat painful impending nonunions, rather than the NSAIDs causing nonunions.

REFERENCES

  1. Top of page
  2. Abstract
  3. INTRODUCTION
  4. SUBJECTS AND METHODS
  5. RESULTS
  6. DISCUSSION
  7. REFERENCES