Osteoporosis International

, Volume 16, Issue 8, pp 861–870

Impact of incident vertebral fractures on health related quality of life (HRQOL) in postmenopausal women with prevalent vertebral fractures

Authors

  • Anna M. Oleksik
    • Department of EndocrinologyVU University Medical Center
  • Susan Ewing
    • Prevention Sciences GroupUCSF
  • Wei Shen
    • Eli Lilly & Co.
  • Natasja M. van Schoor
    • EMGO InstituteVU University Medical Center
    • Department of EndocrinologyVU University Medical Center
    • EMGO InstituteVU University Medical Center
Original Article

DOI: 10.1007/s00198-004-1774-3

Cite this article as:
Oleksik, A.M., Ewing, S., Shen, W. et al. Osteoporos Int (2005) 16: 861. doi:10.1007/s00198-004-1774-3

Abstract

Vertebral fractures are a common complication of osteoporosis and may cause a decrease of health-related quality of life (HRQOL). This study was designed to determine the impact of incident vertebral fractures on HRQOL. The Multiple Outcomes of Raloxifene Evaluation (MORE), a multicenter, randomized, double blind trial, in which women were taking raloxifene or placebo. This study was done in European centers only in a subset of 361 women from seven European countries, all with prevalent vertebral fractures. A specific questionnaire for osteoporosis developed by the International Osteoporosis Foundation was used for assessment at baseline, 1, 2 and 3 years. This questionnaire, Qualeffo, contains 41 questions in the domains pain, physical function, social function, general health perception and mental function. Domain and total scores are expressed on a 100-point scale with 0 corresponding to the best HRQOL. Standardized lateral spinal radiographs were made at baseline, 2 and 3 years and evaluated in a central facility. Sixty-seven patients sustained a fracture in a vertebra that was not fractured at baseline (incident vertebral fractures). Twenty of these were accompanied by signs and symptoms necessitating immediate doctor’s attention (clinical vertebral fractures) and 47 vertebral fractures were only diagnosed on radiographs (subclinical vertebral fractures). Incident vertebral fractures (clinical and subclinical) were associated with an increase of back pain (mean score change 6.4; 95% CI 2.1–10.7), deterioration of physical function (mean score change 2.4; 95% CI 0.1–4.8), and worse general health perception (mean score change 3.8; 95% CI 0.1–7.5). Score changes for patients with subclinical vertebral fractures were intermediate between those for patients with clinical vertebral fractures and patients without incident vertebral fracture. Clinical and subclinical incident vertebral fractures both have an adverse impact on HRQOL.

Keywords

Incident vertebral fracturesHealth-related quality of lifeOsteoporosisQuality of life change

Introduction

Vertebral fractures are a common complication of osteoporosis, and unlike hip fractures, they may occur early in the natural history of this disease [1]. In women, the lifetime risk of a vertebral fracture from age 50 years onwards is 16% [2,3]. A vertebral fracture may have long lasting consequences, but it often is difficult to exactly specify the onset of fracture. Osteoporosis has often been referred to as a “silent thief” [4,5], and vertebral fractures may be “silent” also. According to recent estimates, less than one-third of vertebral fractures cause so many symptoms that they come immediately to the attention of a doctor (clinical vertebral fractures) [6,7]. Other fractures may only cause vague complaints leading to the diagnosis at a later time (subclinical vertebral fractures). Pain caused by a vertebral fracture may last for 3 years or longer [8–11]. Vertebral fractures may result in spinal deformity (kyphosis) [12,13], height loss [14], impaired physical function [15–19], immobility [17], decreased pulmonary function [20,21], and gastroesophageal reflux. The decline in physical function and change in appearance contribute to social isolation, loss of self-esteem and depression [22,23].

To evaluate the effects of osteoporosis on a patient’s well-being, direct questioning is required [24], since there is only a moderate correlation between quality of life and findings on radiography or densitometry [17,24]. Health-related quality of life (HRQOL) plays an increasingly important role in intervention studies in patients with osteoporosis [25–27]. HRQOL is defined as “the value assigned to life as modified by the impairments, functional states, perceptions, and social opportunities that are influenced by disease, injury, and treatment” [28].

According to population-based and cohort studies, women with prevalent radiological vertebral fractures have decreased HRQOL [29–31]. Cross-sectional studies on radiological vertebral fractures include both symptomatic and silent fractures, since it is difficult to determine retrospectively which symptoms and signs accompanied the fracture. Very recent longitudinal data suggest that new vertebral fractures, even those not coming to clinical attention, may be associated with substantial increase in back pain and functional limitations [32,33].

We investigated HRQOL changes due to incident vertebral fractures (IVFX) in European participants of the Multiple Outcomes of Raloxifene Evaluation (MORE) study. Our analysis focused on women with prevalent vertebral fractures at baseline. HRQOL was assessed using QUALEFFO, the quality of life questionnaire of the International Osteoporosis Foundation designed for use in clinical trials [34,35]. The main purpose of our study was to evaluate the impact of IVFX on HRQOL during the trial.

Materials and methods

Study population

The study was performed in the framework of the Multiple Outcomes of Raloxifene Evaluation (MORE) Study, a multicenter randomized double blind trial. The patients in this trial were randomized to placebo, raloxifene 60 mg or raloxifene 120 mg per day. The main outcome criterion was the number of incident (new) vertebral fractures (IVFX). HRQOL was a secondary outcome criterion. For this study, HRQOL follow-up was assessed in 361 ambulant women from seven countries in Europe (Belgium, France, Germany, Great Britain, Italy, Netherlands and Sweden). These countries were selected because an authorized translation of Qualeffo was available. All women had prevalent vertebral fractures and HRQOL assessment at baseline, and at least one HRQOL assessment during the 3 years of follow-up. At the beginning of the study, the women were aged from 53 to 81 years [29] and were at least 2 years postmenopausal, or for those women who had a hysterectomy, had serum follicle-stimulating hormone (FSH) >30 IU/l and serum estradiol <73 pmol/l. A further inclusion criterion was a low bone mineral density (T-score ≤−2.5) either at the lumbar spine or at the femoral neck or at least two vertebrae with more than 25% height loss (moderate fracture) [36]. Women with a history of metabolic bone disease (including hyperparathyroidism, osteomalacia, Paget’s disease), malignancy, or recent treatment for osteoporosis (with the exception of calcium and vitamin D supplements and/or hormonal replacement up to 6 months prior to randomization) were excluded. Spinal deformities that impaired radiographic morphometry were also an exclusion criterion (for example scoliosis >15°). All subjects gave informed consent.

Before randomization, standardized baseline spinal radiographs were assessed (Kiel, Germany) using a semiquantitative scale for vertebral shape (Th4–L4) [37,38]. Prevalent vertebral deformities were diagnosed by anterior, central or posterior height loss in comparison with the adjacent vertebra. A mild vertebral fracture corresponded to 20–25% reduction from expected vertebral height, moderate vertebral fracture to 25–40% height reduction, and severe vertebral fracture to >40% height loss.

Prior to randomization, MORE participants were stratified to one of two study groups at the time of radiographic screening. Patients were assigned to study group 1 if they had low mineral bone density (T-score ≤−2.5) and no more than one mild prevalent vertebral fracture. Study group 2 included women with either at least one moderate vertebral fracture or at least two mild vertebral fractures in the presence of low BMD, or at least two moderate fractures, regardless of BMD. By study design, HRQOL follow-up as reported in this study was obtained in women from study group 2 only.

Evaluation of vertebral fractures

Vertebral radiographs were made at baseline, and after 2 and 3 years of follow-up. When symptoms of vertebral fracture occurred, vertebral radiographs were made at interim 6-month visits or at any time according to the severity of complaints. Radiographs were also performed in women who terminated the study early.

After 3 years of follow-up, a radiologist blinded to the treatment assignment graded the baseline and endpoint radiographs using the same semiquantitative scale: 0 for absent, 1 for mild, 2 for moderate, and 3 for severe fracture [37,38]. An IVFX was defined as a grade change of at least 1 from baseline to endpoint. If no fractures were detected upon review of a patient’s baseline and endpoint radiographs, then no additional radiographic analyses were performed. For fractures observed at baseline or endpoint, a second radiologist determined whether a fracture was present for each vertebra and also performed quantitative morphometry. According to the latter method, an IVFX was defined as vertebral height loss of at least 20% and at least 4 mm. Vertebral fractures (prevalent and incident) were reported if they were confirmed by two of the three types of assessment, i.e. from two independent semiquantitative readings and one quantitative assessment.

At each visit (at baseline, and every 6 months thereafter), all patients were asked if they experienced symptoms of vertebral fractures, including acute pain, stiffness and physical impairments. Incident vertebral fractures in vertebrae that were not fractured at baseline, and were accompanied by symptoms were defined as clinical IVFX. Incident vertebral fractures in vertebrae that were not fractured at baseline, and without symptoms were defined as subclinical IVFX.

Health-Related Quality of Life questionnaire

HRQOL was assessed using a disease-targeted instrument of the International Osteoporosis Foundation (QUALEFFO) containing 41 questions in five domains: pain, physical function, social function, general health perception and mental function. The questionnaire was administered to MORE study participants at clinics in countries where an officially translated version was available, i.e. Belgium, France, Germany, Italy, The Netherlands, Sweden and the United Kingdom. Subjects completed the questionnaires after a short instruction period by a study nurse. At baseline, the questionnaire was administered after screening and enrollment procedures, but before randomization. QUALEFFO follow-up was obtained after 1, 2 and 3 years from baseline, or at study discontinuation.

QUALEFFO was scored according to the following algorithm [29]. First, the response options for each item were standardized, so that option 1 corresponded to the best and option 5 to the worst quality of life. Next, if at least one-half of the items in the domain were answered, the domain scores for each person were calculated by averaging the non-missing scores for individual questions within a domain. By doing this, we necessarily assumed that the mean of the missing scores would be equal to the mean of the non-missing scores (imputation). At baseline [29], in the general health perception domain 1% of the scores had to be imputed. In the domains of pain, physical function and mental function 11% were imputed. This percentage was rather high (66%) in the social function domain because some questions had a “non applicable” answer option. Similar percentages were seen during the follow-up. The domain scores were linearly transformed to a 100-point scale with 0 corresponding to the best HRQOL and 100 to the worst HRQOL. Finally, we calculated two QUALEFFO “total” scores: the average of the transformed domain scores (total 1) and the average of scores for all individual questions (total 2). The QUALEFFO total score 1 and 2 were calculated only if at least 50% of all items were answered. Both total scores were expressed on a 100-point scale.

Other variables

The presence of adverse events and new non-vertebral fractures was recorded at each visit [36]. Adverse events that resulted in death, hospitalization, cancer, permanent disability, or threat to life were classified as serious. The Coding Symbol and Thesaurus of Adverse Events (COSTAR) (USA FDA 1970) was used to categorize reported adverse events.

Treatment

Within each study group, each woman was randomly assigned to one of three treatment groups: placebo, raloxifene 60 mg or raloxifene 120 mg per day. Randomization was performed by the Eli Lilly Clinical Trials Materials Group (Indiana, USA). Upon entry into the study, all women received daily supplements of 500 mg calcium and 400–600 IU cholecalciferol per day.

Study discontinuation

Participants were required to discontinue the study if they had experienced a bone mineral density decrease at 1 year of more than 7% in the lumbar spine or more than 10% in the femoral neck; or a decrease at 2 years of more than 11% in the lumbar spine or more than 14% in the femoral neck; if in 2 years during the study they had experienced more than two incident vertebral fractures; or if any exclusion criterion was met. The criteria on BMD loss and incident vertebral fractures caused the exclusion of 46 patients during the study.

Statistical methods

Baseline characteristics and the occurrence of adverse events were compared between patients experiencing incident vertebral fractures and those who did not, using the two-independent-sample t-test for continuous variables (or in the case of non-normal distribution the Wilcoxon rank sum test) and the chi-square test for categorical variables.

The significance of the impact of incident vertebral fractures on HRQOL was tested using two types of analyses (SPSS 8.0 software package), which are described below. These analyses were adjusted for age, country of origin, baseline scores, number of prevalent fractures (1, 2, or >2) and occurrence of serious adverse events (yes/no). The covariates that did not confound the analyses and had no apparent effect on HRQOL were excluded from the final models.

First, HRQOL changes from baseline to years 1, 2, and 3 of follow-up and/or discontinuation were calculated for each patient and averaged. For each domain, the mean averaged QUALEFFO change was compared according to the clinical severity of incident vertebral fracture (none, subclinical, or clinical) using analysis of variance (general factorial design). Patients with incident vertebral fractures in vertebrae that were fractured at baseline (worsening of prevalent vertebral fracture) were excluded, since they could misinterpret persistent or worsened complaints as related to older fracture. Besides covariates listed above, the models were adjusted for number of follow-up QUALEFFO observations (with those patients who completed all MORE visits as the reference category). The assumptions of linearity, constancy of variance and normality of error terms were verified using residual plots.

Second, the proportions of the averaged QUALEFFO changes beyond a threshold value (1/3 and 1 baseline SD for respective score in the whole study population) were compared, according to the clinical severity of incident vertebral fracture. Patients with worsening of a prevalent vertebral fracture were excluded from the analysis. Logistic regression models were adjusted for covariates listed above and number of QUALEFFO observations. To determine whether age and baseline scores were linear in the logit, each variable was coded into two dummy variables based on tertiles of age and baseline scores. These dummy variables temporarily replaced the continuous version of the variables in the multivariate model, and the coefficient for each tertile was obtained. Since the estimated coefficients for age increased from one tertile to the next, and the estimated coefficients for baseline score decreased from one tertile to the next, the continuous versions of these variables were used in the final models. For each model, the goodness of fit was verified using the Hosmer–Lemeshow test.

Results

Of 361 patients with prevalent vertebral fractures at baseline and at least one QUALEFFO follow-up visit (Fig. 1), 71 patients suffered an incident vertebral fracture (IVFX) during 3 years of follow-up. Four patients experienced worsening of a prevalent fracture. The remaining 67 patients had incident vertebral fractures in a vertebra that was not fractured at baseline; 20 fractures were accompanied by symptoms and signs, i.e. the patient complained and the radiograph confirmed the fracture (clinical IVFX) and 47 fractures were diagnosed on scheduled follow-up vertebral radiographs (subclinical IVFX). Non-responders (i.e. no Qualeffo follow-up; n=88) were not different from responders (n=361) with regard to age, time after menopause and number of prevalent vertebral fractures. Baseline characteristics are summarized in Table 1. In comparison to those who did not experience incident vertebral fractures, patients with IVFX had a higher mean number of prevalent vertebral fractures at baseline [mean (SD): 3.0 (2.2) versus 2.3 (1.8), P=0.003]. Interestingly, the patients who suffered an IVFX during follow-up tended to have higher baseline QUALEFFO scores than the patients who did not have IVFX, suggesting the importance of adjusting for baseline scores in the analyses.
Fig. 1

Schematic representation of study population. This study (within solid line) concerns HRQOL assessment in 361 women with prevalent vertebral fractures at baseline of Multiple Outcomes of Raloxifene Evaluation (MORE) study. Of these, 290 had no incident vertebral fractures, four had progression of prevalent vertebral fractures and 67 had fractures in vertebrae that were not fractured at baseline. Of these 67 incident vertebral fractures, 20 were accompanied with symptoms or signs (clinical vertebral fractures), and 47 were diagnosed only on radiographs (subclinical vertebral fractures). The MORE study comprised 7705 postmenopausal women with osteoporosis [36]. At baseline, HRQOL was assessed in 751 women from seven European countries (within dashed line) [29]

Table 1

Baseline characteristics in postmenopausal women with prevalent vertebral fractures and without (no IVFX) or with incident vertebral fractures (IVFX)

No IVFX (n=290)

IVFX (n=71)

P-value

Mean

(SD)

Mean

(SD)

Age

68.2

(6.3)

69.7

(5.7)

0.062T

Time after menopause

20.2

(8.0)

22.2

(7.5)

0.051T

Prevalent thoracic fractures

1.5

(1.4)

1.9

(1.6)

0.053W

Prevalent lumbar fractures

0.8

(1.0)

1.1

(1.2)

0.017W

Prevalent vertebral fractures

2.3

(1.8)

3.0

(2.2)

0.003W

n

(%)

n

(%)

Prevalent thoracic fractures

236

(81%)

56

(79%)

0.74X

Prevalent lumbar fractures

153

(53%)

46

(65%)

0.068X

Osteoporotic non vertebral fractures

132

(46%)

31

(47%)

0.79X

Arthralgia

74

(26%)

19

(27%)

0.83X

Musculoskeletal pain

28

(10%)

7

(10%)

0.96X

Headache

20

(7%)

7

(10%)

0.35X

Insomnia

39

(13%)

9

(13%)

0.86X

Hypertension

69

(24%)

18

(25%)

0.78X

Depression

16

(6%)

4

(6%)

0.96X

Gastric complaints

20

(7%)

7

(10%)

0.22X

Mean

(SD)

Mean

(SD)

QUALEFFO paina

37.5

(25.8)

42.7

(27.8)

0.13T

QUALEFFO physical functiona

19.1

(15.6)

23.1

(14.3)

0.006W

QUALEFFO social functiona

40.5

(20.3)

42.8

(20.7)

0.34W

QUALEFFO health perceptiona

47.3

(20.1)

54.6

(20.1)

0.006T

QUALEFFO mental functiona

32.2

(16.0)

36.0

(18.3)

0.088T

QUALEFFO total 1a

35.4

(15.3)

39.9

(14.9)

0.008W

QUALEFFO total 2a

29.8

(14.5)

33.7

(13.2)

0.006W

aQUALEFFO scores range between 0 and 100. QUALEFFO total 1 was calculated as mean of domain scores, QUALEFFO total 2 as mean of all individual questions

P-value: Tt-test for independent samples; WWilcoxon rank sum test; Xchi-square. Osteoporotic non-vertebral fractures: fractures of hip and wrist

During the follow-up, patients with IVFX were significantly more likely to experience an additional serious adverse event (38% versus 23%, P=0.012), but occurrence of non-vertebral fractures was similar in the two groups (8% versus 6%, NS). Occurrence of IVFX was similar in the different countries (13–19%, data not shown).

The impact of clinical and subclinical IVFX on average change in HRQOL from baseline was assessed using all available QUALEFFO data . First, the average QUALEFFO changes during follow-up were used as dependent variables in analysis of variance (Table 2, Fig. 2). New IVFX were associated with greater average QUALEFFO change for scores: pain (P<0.01), physical function (P=0.04), general health perception (P=0.04), and both total scores (both P<0.03). Patients with clinical IVFX had greater average QUALEFFO changes than patients without IVFX, while the changes in patients with subclinical IVFX were intermediate. Interestingly, patients with subclinical IVFX differed significantly from those without IVFX for the pain score.
Table 2

Impact of new incident vertebral fractures on average changes in HRQOL from baseline. Analysis was performed using data from all patients with at least one QUALEFFO follow-up visit. Patients who had an incident vertebral fracture in a vertebra that was fractured at baseline were excluded. Incident vertebral fractures (IVFX) of intact vertebrae and serious adverse events were tested for fixed effects on HRQOL using a general factorial model with age and baseline score as covariates. Country of origin and number of follow-up visits were random factors. The number of prevalent vertebral fractures had no effect, did not confound the analysis and was excluded from the final models. In all models, baseline score was a confounder. Serious events were included in all models, had a significant additional impact on HRQOL scores, and were a confounder in models with physical function scores

No IVFX (n=290)

Subclinical IVFX (n=47)

Clinical IVFX (n=20)

β (95% CI) for all IVFX vs no IVFX

β (95% CI) for clinical IVFX vs no IVFX

Mean (SD)

Mean (SD)

Mean (SD)

Pain

Baseline score

37.5 (25.8)

41.2 (27.7)

43. 6 (26.1)

6.4 (2.1; 10.7)

8.2 (0.9; 15.6)

Average change from baseline

−3.6 (18.0)

0.8 (20.5)*

3.6 (24.1)*

Percentage change

−9.6%

1.9%

8.3%

Physical function

Baseline score

19.1 (15.6)

22.7 (14.8)

23.8 (14.3)

2.4 (0.1; 4.8)

3.5 (-0.5; 7.6)

Average change from baseline

1.3 (9.0)

2.7 (11.8)

4.8 (12.5)$

Percentage change

6.8%

11.9%

20.2%

Social function

Baseline score

40.5 (20.3)

41.0 (20.5)

50.2 (19.2)

2.8 (−1.1; 6.9)

3.6 (−3.2; 10.4)

Average change from baseline

1.3 (14.7)

4.0 (18.6)

3.6 (15.40)

Percentage change

3.2%

9.8%

7.2%

Health perception

Baseline score

47.3 (20.1)

55.3 (18.5)

52.1 (22.3)

3.8 (0.1; 7.5)

7.1 (0.8; 13.3)

Average change from baseline

−6.8 (15.8)

−0.4 (13.8)

6.0 (18.5)*

Percentage change

−14.4%

−0.7%

11.5%

Mental function

Baseline score

32.2 (16.0)

36.9 (17.8)

32.4 (17.6)

−0.4 (-3.2; 2.4)

0.9 (−3.8; 5.6)

Average change from baseline

1.5 (11.0)

−0.1 (12.3)

2.6 (8.9)

Percentage change

4.7%

-0.3%

8.0%

QUALEFFO total 1

Baseline score

35.4 (15.3)

39.7 (14.9)

39.9 (13.4)

2.7 (0.4; 5.1)

5.0 (1.1; 9.0)

Average change from baseline

0 (8.9)

0.7 (11.2)

4.6 (11.5)*

Percentage change

0%

1.8%

11.5%

QUALEFFO total 2

Baseline score

29.8 (14.5)

33.5 (13.8)

34.1 (11.8)

2.4 (0.3; 4.5)

4.0 (0.4; 7.6)

Average change from baseline

0.5 (7.9)

1.5 (10.6)

4.3 (9.9)*

Percentage change

1.7%

4.5%

12.6%

Difference from no IVFX: $P<0.10; *P<0.05

Fig. 2

Impact of incident vertebral fractures on average changes in Qualeffo scores from baseline. The fractures were clinical (i.e. accompanied by symptoms or signs) or subclinical (only diagnosed on radiographs). *P<0.05 vs no incident vertebral fractures

In all models presented in Table 2, age, country of origin, and serious adverse events had independent additional effects on average changes of QUALEFFO scores. In particular, older patients had larger increases in total scores 1 and 2 than younger patients (0.17 point per year, 95% CI: 0.01–0.32). Serious adverse events increased mean changes of QUALEFFO total scores [mean (95% CI): total score 1, 2.2 (0.02–4.3); total score 2, 2.6 (0.6–4.5)]. In contrast, the number of prevalent vertebral fractures had no effect on average QUALEFFO changes. Baseline scores were the strongest predictors of average QUALEFFO changes (for each score, P<0.05) Patients with higher baseline scores were likely to show smaller increases in QUALEFFO scores than patients with lower scores.

The proportion of patients whose HRQOL loss exceeded a threshold value was related to the clinical significance of IVFX (Table 3, Fig. 3). For each score, two threshold values were tested: 1/3 SD and 1 SD of the baseline score. For the pain and physical function domains and both total scores, patients with IVFX (subclinical and clinical) were more likely to experience HRQOL decline beyond the threshold value than patients without IVFX. In addition, patients with IVFX had a significantly higher frequency of HRQOL declines beyond the 1 baseline SD threshold in the social function domain and beyond the 1/3 baseline SD threshold in the general health perception domain. In the mental function domain, there were no significant differences between the three groups. The effect of treatment on Qualeffo scores was also tested. Qualeffo scores were somewhat lower in patients on raloxifene than in patients on placebo, but these differences were not significant.
Table 3

Proportion of patients whose HRQOL loss exceeded a threshold value. Impact of incident vertebral fractures (IVFX) of intact vertebrae on HRQOL losses (average QUALEFFO change) beyond a threshold value was tested in logistic regression models. Threshold values are listed in brackets (second column) and were 1/3 baseline SD and 1 SD for respective QUALEFFO scores in the whole study population. Age, country of origin, baseline score, and serious adverse events and number of QUALEFFO assessments were covariates. The number of prevalent vertebral fractures had no additional effect and did not confound the analysis, and was therefore excluded from the final models

Threshold

No IVFX (n=290)

Subclinical IVFX (n=47)

Clinical IVFX (n=20)

IVFX vs no IVFX

Pain

1/3 baseline SD (8.7)

62/289 (21.5%)

17/47 (36.2%)**

9/20 (45.0%)**

<0.001

1 baseline SD (26.1)

11/289 (3.8%)

4/47 (8.5%)$

4/20 (20.0%)**

0.002

Physical function

1/3 baseline SD (5.2)

67/290 (23.1%)

15/47 (31.9%)

11/20 (55.0%)**

0.005

1 baseline SD (15.5)

13/290 (4.5%)

7/47 (14.9%)**

2/20 (25.0%)***

<0.001

Social function

1/3 baseline SD (6.7)

85/268 (31.7%)

15/40 (37.5%)

6/18 (33.3%)

0.41

1 baseline SD (20.3)

25/268 (9.3%)

9/40 (22.5%) *

2/18 (11.1%)

0.040

Health perception

1/3 baseline SD (6.7)

88/286 (30.8%)

13/46 (28.3%)

10/20 (50.0%)*

0.048

1 baseline SD (20.2)

23/286 (8.0%)

4/46 (8.7%)

3/20 (15.0%)

0.192

Mental function

1/3 baseline SD (5.5)

96/289 (33.2%)

16/47 (34.0%)

8/20 (40.0%)

0.77

1 baseline SD (16.8)

24/289 (8.3%)

3/47 (6.4%)

1/20 (5.0%)

0.60

QUALEFFO total 1

1/3 baseline SD (5.1)

69/290 (23.8%)

16/47 (34.0%)*

9/20 (45.0%)*

0.004

1 baseline SD (15.2)

12/290 (4.1%)

8/47 (10.6%)*

4/20 (20.0%)**

0.002

QUALEFFO total 2

1/3 baseline SD (4.8)

73/290 (25.2%)

17/47 (36.2%)*

9/20 (45.0%)*

0.006

1 baseline SD (14.4)

11/290 (4.1%)

8/47 (17.0%)***

3/20 (15.0%)**

<0.001

Comparison with no IVFX: $P=0.065, *P<0.05, **P<0.01, ***P<0.001

Fig. 3

Quality of life loss following subclinical or clinical incident vertebral fractures: proportion of patients with Qualeffo total score increases >1 SD. *P<0.05 vs no incident vertebral fractures

Discussion

Our results underscore the clinical importance of incident vertebral fractures. To date, there have been only a few longitudinal studies addressing quality of life components in patients with IVFX. In these studies, incident vertebral fractures were associated with worsening of back pain and back pain-related disability [32], with decreased physical function [33] and increased emotional problems [33].

Patients who suffer clinical vertebral fractures experience an abrupt descent into disease and disability [39]. In our study, incident vertebral fractures were analyzed in patients with prevalent vertebral fractures, and these patients were aware of their illness. This could be why we did not find any impact of IVFX on mental function. In contrast, in the non-European MORE population, incident vertebral fractures had a negative impact on emotional function [33].

An important issue is the interpretation of HRQOL changes. Relatively small changes in quality of life may be statistically significant, especially in large studies [40]. Effect size may be a useful parameter, and one may consider a change in scores to be clinically meaningful if it exceeds a threshold of 1 baseline SD within the group or, for randomized groups, 1/3 baseline SD [41]. Interestingly, in our study, the mean of HRQOL changes for those who experienced a clinical IVFX were close to the 1/3 baseline SD threshold value for the pain and general health perception domains, and total scores.

On the other hand, some researchers [40,42–44] believe that Fletcher’s approach and similar distribution-based interpretations are measures of responsiveness rather than importance and that they underestimate clinically important differences. Therefore, the so-called anchored approach is preferred. HRQOL changes could be anchored, or attributed to the effects of age or other well-described conditions. In our linear regression models, the estimated changes in QUALEFFO total scores for clinical IVFX were similar to the estimated changes due to ten years’ difference in age. Furthermore, the impact of serious adverse events on QUALEFFO total scores was intermediate between the impact of subclinical and clinical IVFX.

The impact of subclinical IVFX on HRQOL could be somewhat overestimated in our study, if the new fracture is located close to a previous fracture. By the study design, our results cannot be extrapolated to patients without prevalent vertebral fractures who subsequently experience an IVFX. Recent cross-sectional data indicate that patients with a prevalent vertebral fracture had lower HRQOL than those without vertebral fractures [29,30,45]. However, without adjusting for BMD or comorbidities, such cross-sectional studies could overestimate the impact of vertebral fractures on HRQOL: low BMD is associated with significantly higher levels of depressive symptoms [46] and fears [47], and vertebral fractures are associated with excess mortality due to comorbidities [21,48,49]. Patients with IVFX suffered more adverse events than the other patients, which probably is also due to comorbidity.

The described changes may represent short-term, rather than long-term, consequences of vertebral fractures, since the maximum follow-up time for our study was 3 years. Whether these changes continue to persist over time needs further study. We were not able to investigate secondary changes that occur some time after the fracture. A cross-sectional study showed that the passage of time since fracture might modify the perception of pain, social function and well-being, while its effect on physical function is less evident [50].

In our models, the baseline QUALEFFO score did inversely predict the average score change. This could be explained by the fact that scores are organized on a 100-point scale. High scores, close to the upper range of the scale, are less likely to increase than low scores. Our finding suggests that in future research, baseline score should be taken into account when evaluating a treatment effect.

A limitation of the study is that QUALEFFO data could not be obtained in all patients at visits when QUALEFFO data were scheduled to be collected. We tried to avoid possible bias by analyzing data originating from all patients with at least one QUALEFFO follow-up visit by using their average changes in QUALEFFO scores, since baseline as an outcome variable, and by adjusting for number of QUALEFFO follow-up visits in the models. When making a distinction between clinical and subclinical IVFX, we excluded patients who had worsening of a previous fracture. This limited the number of new incident vertebral fractures to 67. Therefore, we were not able to investigate the effect of fracture location or adjust the analysis for other variables such as BMI, smoking status, alcohol use, and co-medication.

The effect of treatment with raloxifene versus placebo on quality of life was not significant. Raloxifene has no analgesic properties. The main effect of raloxifene in this trial was a decrease in the number of incident vertebral fractures [36].As such, one may postulate that quality of life would be better in the raloxifene group than in the placebo group, but the sample size was too small to detect a significant effect of raloxifene on HRQOL. Other potential benefits such as reduction of breast cancer risk, cholesterol and blood pressure [51] could improve HRQOL in women using raloxifene but these could be partially counteracted by raloxifene-related adverse events such as hot flushes, leg cramps, leg edema or thrombo-embolic events [51]. The sample size to detect an effect on quality of life might be in the range of the complete MORE study.

In conclusion, in patients with previous vertebral fractures, incident vertebral fractures are associated with a significant decline in HRQOL for most of the domains that were analyzed. In our study, patients with subclinical incident vertebral fractures also had a decrease in HRQOL. The study stresses the importance of vertebral fracture prevention. In clinical trials, any incident radiological fracture, not just clinical fractures, should be considered as a primary end-point with regard to health-related quality of life.

Acknowledgements

We thank all investigators and study nurses involved in this project. We are grateful to Nicolette Pliester, Greetje Asma, Elisabeth Moseley, Sunny Xie, Michele Ennis and Emily Kenyon for data management. We thank Michael Minshall, Ronald Jansen, Alison Dawson and Tu Duong for advice on quality of life assessment and data analysis. Natasja van Schoor was supported by an unconditional grant from Wyeth. This study was funded by E. Lilly and Company. We also thank the Working Party for Quality of Life of the International Osteoporosis Foundation for comments and advice in the planning and reporting of this study. Members include D. Agnusdei (Siena), F. Caulin (Paris), C. Cooper, J. Kanis, A. Leplege (Paris), O. Johnell (Malmö), U.A. Liberman (Petah Tiqva), H Minne (Bad Pyrmont), J. Reeve (Cambridge), J.Y. Reginster (Liège), C. Todd (Cambridge), M.C. de Vernejoul (Paris) and I. Wiklund (Mölndal).

Copyright information

© International Osteoporosis Foundation and National Osteoporosis Foundation 2004