We have recently observed an increased risk for vertebral fractures (VF) in a randomized controlled trial comparing the analgesic effect of vertebroplasty (VP) versus conservative treatment in symptomatic VF. The aim of the present study was to evaluate the risk factors related to the development of VF after VP in these patients. We evaluated risk factors including age, gender, bone mineral density, the number, type, and severity of vertebral deformities at baseline, the number of vertebral bodies treated, the presence and location of disk cement leakage, bone remodeling (determining bone turnover markers) and 25 hydroxyvitamin D [25(OH)D] levels at baseline in all patients. Twenty-nine radiologically new VF were observed in 17 of 57 patients undergoing VP, 72% adjacent to the VP. Patients developing VF after VP showed an increased prevalence of 25(OH)D deficiency (<20 ng/mL) and higher P1NP values. The principal factor related to the development of VF after VP in multivariate analysis was 25(OH)D levels < 20 ng/mL (RR, 15.47; 95% CI, 2.99–79.86, p < 0.0001), whereas age >80 years (RR, 3.20; 95% CI, 1.70–6.03, p = 0.0007) and glucocorticoid therapy (RR, 3.64; 95% CI, 1.61–8.26, p = 0.0055) constituted the principal factors in the overall study population. Increased risk of VF after VP was also associated with cement leakage into the inferior disk (RR, 6.14; 95% CI, 1.65–22.78, p = 0.044) and more than one vertebral body treated during VP (RR, 4.19; 95% CI, 1.03–34.3, p = 0.044). In conclusion, nearly 30% of patients with osteoporotic VF treated with VP had a new VF after the procedure. Age, especially >80 years, the presence of inferior disk cement leakage after the procedure, the number of cemented vertebrae, and low 25(OH)D serum levels were related to the development of new VF in these patients, with the latter indicating the need to correct vitamin D deficiency prior to performing VP.
Vertebroplasty (VP) is a widely used method for the management of painful vertebral compression fractures.[1-11] However, the clinical indications of this procedure are still under debate, as is its possible relationship with the development of additional vertebral fractures (VF).[12-15] Thus, although several observational and “in vitro” studies have suggested an increased risk of fracture in the vertebrae located adjacent to the cemented vertebral body,[16-19] most prospective randomized clinical trials (RCT) comparing VP with either a sham procedure or conservative approach have not confirmed this finding.[12, 13] Nevertheless, in a recent RCT our group showed than the VP procedure was 2.78-fold more frequently related to the development of new radiological VF compared with a conservative approach. Most of these fractures were observed within 3 months after the procedure, especially during the 44 days post-VP and adjacent to the treated vertebral body. In addition, as Lin and colleagues suggested, we observed that cement leakage into the inferior disk during the procedure was associated with an increased risk for developing new VF.
Therefore, the aim of the present study was to analyze the risk factors related to the development of VF after VP in these patients, including not only clinical and radiological factors but also densitometric and biochemical parameters.
Patients and Methods
The detailed study design has been published. Briefly, we performed a RCT comparing VP with conservative treatment in improving pain and quality of life in individuals with painful osteoporotic VF over a 1-year follow-up period. The study was conducted at the Neurointerventional Radiology Department of the Imaging Diagnostic Centre, in conjunction with the Rheumatology Department of the Hospital Clinic of Barcelona, Spain. Inclusion criteria were the following: acute, painful osteoporotic vertebral fractures from T4 to L5 with clinical onset <12 months, as confirmed by spine X-ray and by the presence of edema on short T1 inversion recovery (STIR) MRI or activity on bone scan, and with a visual analogue scale (VAS) score ≥4 on a scale from 0 to 10. Exclusion criteria were the following: untreatable coagulopathy, active local or systemic infection, current malignancy, vertebral canal occupation by a fragment of the vertebral body or nonosteoporotic vertebral fracture, non-informed consent, or active associated disorders (ie, fibromyalgia or spondyloarthropathies) that may interfere with correct assessment of quality of life and pain. Ethical approval was obtained from our hospital Ethics Committee (Project number 2907; Clinical Trials number ID NCT00994032 Quality of Life After Vertebroplasty Versus Conservative Treatment in Patients With Painful Osteoporotic Vertebral Fractures; http://clinicaltrials.gov/show/NCT00994032) and all participants provided written informed consent.
Spinal X-rays, bone mass measurements, blood analysis including bone metabolism parameters, and MRI were performed in all patients at baseline. In addition, all individuals were clinically assessed at baseline and thereafter at 2 weeks and at 2, 6, and 12 months. Medication details, including antiosteoporotic and analgesic therapy, were recorded throughout the study as was the use of glucocorticoids or any other drug that may affect bone mineral metabolism. Most VP procedures were carried out with bilateral transpedicular 10-gauge or 13-gauge needle injection of poly(methyl methacrylate) (PMMA) cement. Dyna-CT (AXIOM Artis FD Biplane Angiosuite with DynaCT; Siemens Medical Solutions, Erlangen, Germany) immediately after VP or standard CT 24 hours after the procedure was undertaken in order to check cement distribution or leakage. The procedure was undertaken in symptomatic vertebral compression fractures with associated edema in MRI, with a maximum of four vertebral levels being treated in a single session. Treatment included calcitonin during the first month and analgesics when necessary (with standardized format). Conservative therapy consisted of analgesics with standardized format and nasal calcitonin (first month). After 1 month of treatment, both groups began/or continued treatment with bisphosphonates (BP), except for those with intolerance to these compounds, who continued therapy with teriparatide or strontium ranelate, according to the attending physician.
Standard X-rays of the thoracic and lumbar spine were obtained to evaluate VF at baseline, and at 6 and 12 months. The evaluation of the VF was performed by a semiquantitative approach by an experienced rheumatologist of the Unit of Metabolic Bone Diseases. VF was defined as a reduction of 20% or more in the anterior, middle, or posterior height of the vertebral body compared with adjacent, undeformed vertebrae.[21, 22] Doubtful cases of vertebral deformities were discussed in a consensus meeting. According to the Genant criteria, confirmed VF were classified as wedge, biconcave, or compression fractures and the severity of the deformity was graded as mild (20% to 25% height reduction), moderate (25% to 40%), or severe (>40%).
MRI with STIR-weighted images was performed at baseline to confirm bone marrow edema of the painful vertebra and repeated on suspicion of a new fracture during follow-up. In cases of MRI contraindication, a bone scan was used to assess fulfillment of the inclusion criteria.
Bone mineral density (BMD) of the lumbar spine (L2–L4) and femur (femoral neck and total hip) was performed in all patients by dual X-ray absorptiometry (Lunar Prodigy) before randomization and at the last follow-up visit at 12 months. The results of BMD are expressed as T-scores.
Serum calcium and phosphate levels, liver and renal function tests and a complete blood count were evaluated in all patients by standard procedures between early in the morning after an overnight fast. In addition, 25 hydroxyvitamin D [25(OH)D] serum levels (using the Liason DiaSorin chemiluminescent immunoassay system), procollagen type 1 N-propeptide (P1NP) (Roche Elecsys) as a marker of bone formation, and the urinary N-terminal cross-linked telopeptide of type I collagen (NTx) determined by ELISA (Ostex, Seattle, WA, USA) as a marker of bone resorption were analyzed. Urine determinations were expressed in relation to creatinine excretion. Serum values of 25(OH)D less than 20 ng/mL were considered indicative of vitamin D deficiency; serum values of P1NP greater than 55 ng/mL and urinary NTx >65 nmol/mmol were both considered above the reference values.
Statistical analysis was carried out using a per protocol strategy. Quantitative variables were summarized using means and SDs whereas counts and percentages were reported for qualitative variables. Clinical, radiological, densitometric, and biochemical characteristics of patients with and without new VF were compared among therapeutic approaches using analysis of variance (ANOVA), logistic regression or Poisson regression depending upon whether the outcome was continuous, binary, or a count. A Poisson regression model was fitted to analyze the association between the incidence of VF and covariates. The association between VF and adjacency was assessed by fitting a logistic regression via GEE approach accounting for the subject cluster effect. Significance of variables was assessed using the F Wald test. The overall type-I error rate was set at 5%. Statistical analyses were carried out using SAS, v9.2 (SAS Institute, Inc., Cary, NC, USA).
Of the initial 125 patients randomized 95 (47/64 in the VP arm and 48/61 in the conservative treatment arm) completed the 12-month follow-up (Fig. 1). Of the initial 64 patients randomized to VP, 57 underwent VP on 140 vertebrae (2.46 ± 1.56 vertebrae/procedure). Twenty-nine new radiological VF were observed in 17 of the 57 patients treated with VP, mainly being found adjacent to the VP procedures (72%) (RR = 9.42, 95% CI, 4.24–20.88, p < 0.0001). Conversely, 11 of the 61 patients treated with conservative approach (7/61 were excluded due to rescue treatment) developed 11 new VF, 27% being adjacent to previous VF (RR = 1.49, 95% CI, 0.40–5.55, p = 0.47). Table 1 shows the clinical characteristics of the patients.
|Overall (n = 111)||Subjects with VP without new VF (n = 40)||Subjects with VP with new VF (n = 17)||Subjects with CT without new VF (n = 43)||Subjects with CT with new VF (n = 11)||Interaction pa||New VF p|
|Age (years)||73.59 ± 9.32||71.45 ± 10.29||72.39 ± 8.90||74.06 ± 8.63||80.49 ± 5.28||0.1683||0.1015|
|Gender (F %)||77.5||75.0||70.6||81.0||90.9||0.3788||0.8554|
|VF||3.28 ± 2.45||3.15 ± 2.62||3.94 ± 2.19||3.02 ± 2.11||3.00 ± 2.76||0.6379||0.5825|
|VP||2.40 ± 1.50||2.18 ± 1.38||2.94 ± 1.68||–||–||–||0.0771|
|Mild VF||1.05 ± 1.41||1.15 ± 1.41||1.88 ± 1.73||0.57 ± 0.97||1.18 ± 1.78||0.6638||0.0264|
|Moderate VF||1.39 ± 1.61||1.38 ± 1.55||2.29 ± 1.96||1.07 ± 1.47||1.27 ± 1.49||0.4866||0.0864|
|Severe VF||1.24 ± 1.29||0.90 ± 1.37||1.35 ± 1.00||1.43 ± 1.38||1.55 ± 0.82||0.4587||0.2784|
|Wedge VF||1.56 ± 1.45||1.25 ± 1.13||2.41 ± 1.80||1.55 ± 1.47||1.45 ± 1.51||0.0782||0.0649|
|Biconcave VF||1.15 ± 1.63||1.40 ± 1.78||1.35 ± 1.80||0.74 ± 1.15||1.55 ± 2.21||0.1606||0.3599|
|Crush VF||0.95 ± 1.43||0.78 ± 1.07||1.76 ± 2.61||0.79 ± 1.07||1.00 ± 0.89||0.2936||0.0248|
|FN T-score||−2.19 ± 0.94||−2.17 ± 1.10||−2.09 ± 0.56||−2.23 ± 1.00||−2.29 ± 0.60||0.7824||0.9379|
|Lumbar T-score||−2.62 ± 1.59||−2.43 ± 1.75||−2.61 ± 1.92||−2.89 ± 1.42||−2.41 ± 1.12||0.4351||0.7928|
|25(OH)D (ng/mL)b||26.48 ± 24.10||28.62 ± 21.00||18.62 ± 21.01||27.12 ± 29.46||28.18 ± 11.91||0.0120||–|
|25(OH)D < 20 (ng/mL)b||46.7||35.1||81.3||53.5||9.1||<0.0001||–|
|NTx (nmol/mmol)||71.55 ± 62.45||60.77 ± 47.37||110.16 ± 110.14||62.74 ± 38.19||58.50 ± 14.85||0.3610||0.0536|
|P1NP (ng/mL)||63.41 ± 64.07||58.46 ± 69.77||101.62 ± 97.83||57.03 ± 38.76||40.43 ± 26.63||0.0270||–|
|Glucocorticoid treatment (%)||12.8||14.3||14.3||4.8||36.4||0.0572||0.0749|
On comparing the clinical, radiological, densitometric, and biochemical characteristics of patients with and without new VF, depending on the therapeutic approach (ie, VP versus conservative treatment at baseline) and accounting for the VF-therapeutic approach interaction effect, patients who developed VF after VP showed lower mean vitamin D serum levels, an increased prevalence of 25(OH)D deficiency (<20 ng/mL), and higher P1NP serum values, findings that were not observed in patients treated with the conservative approach (Table 1). In addition, these subjects also tended to have a higher number of wedged vertebral deformities at baseline and a higher number of vertebral bodies treated per patient (Table 1). Conversely, glucocorticoid treatment tended to be more frequently associated with new VF in patients treated with conservative approach (Table 1).
Cement leakage into disks occurred in 15.4% of the VP procedures. Leakages into the inferior disk were observed in 8.4% of the procedures, with 18.2% of these cases being found in adjacent VF (Fig. 2). Leakages into the superior disk were observed with a similar frequency (8.8%); however, none was related to a new VF.
Antiosteoporotic treatment was similar in the two groups of patients with and without new VF (p = 0.26), with most patients being treated with oral BP (59% versus 63% of patients with and without VF in VP groups, respectively, and 82% versus 70% of patients with and without VF in the conservative groups).
In the univariate analysis (Table 2), adjusted by gender and age, the risk factors for the development of new VF after VP were as follows: deficient 25(OH)D serum levels (<20 ng/mL) (RR, 6.82; 95% CI, 2.06–22.59, p = 0.0017) with a significant distinct effect to that observed in the conservative group (p < 0.0001), cement leakage into the inferior disk after the procedure (RR, 6.14; 95% CI, 1.65–22.78, p = 0.044) and the number of VP procedures carried out per patient, especially in those undergoing more than one procedure (RR, 4.19; 95% CI, 1.03–34.30, p = 0.044). Univariate analysis also showed an increased risk of VF in the overall study population related to elderly age (>80 years) (RR, 3.77; 95% CI, 1.83–7.78, p = 0.0007). Having increased P1NP values (>75 ng/mL; a value that was observed to be associated with a higher risk for VF) and more than two initial VF deformities (mild or moderate or wedged) were also associated with increased risk of VF in the overall study population. Nevertheless, this increased risk was especially observed among patients treated with VP (Table 2), in whom a marginally significant interaction was observed (p < 0.1). Conversely, glucocorticoid treatment tended to be associated with an increased risk for VF in the conservative group (Table 2).
|Groups in comparison||Overall||Vertebroplasty||Conservative||Interaction pa||Effect p|
|RR||95% CI||RR||95% CI||RR||95% CI|
|Age (years)||>80/ ≤ 80||3.77||1.83–7.78||4.31||1.86–10.00||2.67||0.69–10.37||0.5632||0.0007|
|VF at baseline||>2/2–1||1.47||0.69–3.12||2.23||0.85–5.88||0.60||0.15–2.41||0.1208||0.3135|
|Lumbar T-score||≥ –3/< –3||1.45||0.57–3.70||1.06||0.34–3.28||4.44||0.32–62.76||0.2735||0.4383|
|P1NP (ng/mL)||>75/ ≤ 75||3.83||1.53–9.57||5.81||2.00–17.22||0.71||0.05–9.65||0.0929||0.0064|
|25(OH)D (ng/mL)b||<20/ ≥ 20||1.73||0.81–3.72||6.82||2.06–22.59||0.12||0.02–0.93||<0.0001||0.1522|
|Moderate VF||≥2/ < 2||2.06||1.02–4.17||2.35||0.99–5.58||1.51||0.39–5.80||0.5852||0.0436|
|Wedge VF||>2/ ≤ 2||3.04||1.55–5.97||4.64||2.07–10.37||0.89||0.18–4.43||0.0929||0.0064|
|Crush VF||>2/ ≤ 1||2.05||0.86–4.89||2.32||0.83–6.48||1.47||0.28–7.71||0.9012||0.2646|
|2/ ≤ 1||1.59||0.61–4.14||1.70||0.54–5.30||1.35||0.23–7.99|
Multivariate analysis including: type and severity of the VF, age, gender, 25(OH)D, and P1NP serum levels and glucocorticoid therapy showed that 25(OH)D insufficiency was the principal factor related to the development of VF after VP, whereas age >80 years and glucocorticoid therapy constituted the principal factors in the overall study population (Table 3).
|Groups in comparison||Treatment group||RR||95% CI||p|
|Age||>80/ ≤ 80||Overall||3.20||1.70–6.03||0.0007|
|25(OH)D||<20/ ≥ 20||Vertebroplasty||15.47||2.99–79.86||<0.0001|
Figure 3 shows the distribution of VF by vertebral levels at baseline as well as the treated vertebral levels and the new VF after VP (Fig. 3A), and the distribution of baseline and new VF in the group treated with conservative approach (Fig. 3B). As shown, most VF were located in the thoracolumbar junction within the T10 to L2 levels, with L2 being the single most affected vertebra at baseline in the VP treated group and T12 the most affected in the group treated with conservative approach. These locations were also the most frequent sites for VP procedures and new VF, the latter being more frequently observed at the L1, T12, T11, T10, and T9 levels. The most frequent locations for new VF in conservatively treated patients were L3, T10, and T12. Figure 4 shows the distance between the VF level at baseline and the new VF fractures after VP and conservative treatment, indicating that most new VF fractures after VP were adjacent to the VP procedure.
This study shows that the increased risk for VF after VP was mainly associated with low 25(OH)D serum levels and age. In addition, the number of vertebral bodies treated and the presence of inferior disk leakage after the procedure were also related to the development of new fractures, with the highest risk being observed in patients >80 years of age, with 25(OH)D serum values <20 ng/mL, undergoing more than one VP procedure. Glucocorticoid therapy was also associated with the development of VF, but this effect was especially observed in patients treated with the conservative approach. It should be noted that most of these factors have previously been associated with increased risk of VF in patients with osteoporosis. Indeed, age is a well known risk factor related to fractures, independently of BMD. Thus, for any BMD fracture risk is much higher in the elderly than in the young population. Similarly, in our study all the patients over the age of 80 years showed more than a threefold increased risk for new VF regardless of the treatment received. The effect of age on the increased risk of skeletal fractures seems to not only be linked to decreased BMD but also to age-related changes in bone quality. Indeed, in our study incident VF were not related to low BMD values, a factor that has been associated with the development of VF after VP in several previous studies[19, 25, 26] and that could possibly play a role in cases with a marked decrease in BMD, such as in the study of Rho and colleagues in which patients with VF had a mean T-score of −4. In our study the low percentage of patients with such a marked decrease in BMD values (only 15.3% of the patients showed a T-score value < –4) could partly explain the absence of a relationship with the BMD values. Nevertheless, patients with new VF also showed more severe bone disease, with a higher number of vertebral deformities at baseline, thereby confirming a more severe alteration in the bone strength of the high-risk group. In this sense, there was a fivefold to twofold increased risk of a new VF in patients with more than two mild, moderate, or wedge VF at baseline. The presence of an incidental VF is a well known risk factor for developing a new VF, with nearly 20% of these individuals having a new fracture within the next year if they do not receive antiosteoporotic treatment. In addition, and similar to our results, the increment of risk increased not only with the number of VF but was also related to the type and severity of the deformity, being wedge deformities those with the highest risk.[28, 29] It seems that this type of deformity may induce a higher fracture risk into the adjacent vertebrae by modifying the spinal alignment.
Moreover, disk leakage into the inferior disk was also a contributory factor for developing a new VF after VP with a greater than sixfold increased risk for a new VF. Cement leakage into the disk has been previously associated with adjacent VF after VP.[16, 31] Such leaks might occur through fractured endplates or vacuum clefts or may result from iatrogenic endplate perforations with the needle tip. The hard cement into the disk probably alters the load transfer increasing the mechanical pressure of the endplate of the adjacent vertebral body, increasing the risk of a VF in osteoporotic vertebrae. In addition, the asymmetrical distribution of the cement contributes to the alteration in the biomechanics of the vertebra. Kyphoplasty has also been associated with cement leakage into the vertebral disks. Although increased risk of VF has not been reported in RCT, comparative studies with VP have shown a similar incidence of vertebral fractures with both therapeutic approaches,[33, 35, 36] and retrospective analyses have also suggested an increased risk of VF with kyphoplasty.[37, 38] Nevertheless, it should be pointed out that although cement leakage contributed to the development of a new VF after the procedure, in general leakage was an infrequent cause of VF. Thus, although cement leakage was observed in 15.4% of the procedures, only 8.4% of the procedures presented leakage into the inferior disk, with 18.2% of these resulting in VF.
Similar to previous studies, most of the VF at baseline were located at the thoracolumbar junction, within the T10 to L2 vertebral levels as were treated vertebrae, with T11 being the single most frequently treated vertebra. New VF after VP also showed a similar distribution, with L1, T12, T11, T10, and T9 being the vertebrae most frequently affected. The latter may be explained by the ninefold increased risk of VF by adjacency to the VP procedure (72% of the new VF) in the present study. Otherwise, only 27% of new VF in the conservatively treated patients were adjacent to a previous VF. Previous reports have also suggested an increased prevalence of VF ranging from 41% to 67%, in the vertebra adjacent to the procedure,[19, 40] especially in the first few months after VP.
Interestingly, we observed that 25(OH)D deficiency was associated with an increased risk of VF after the VP procedure. Thus, 25(OH)D insufficiency was the principal independent factor in the multivariate analysis related to the development of VF after VP. Patients with levels lower than 20 ng/mL showed a greater than 15-fold increased risk for a new VF, a finding that was not observed after conservative treatment. The reason for these differences is not known and can not be related to differences in baseline 25(OH)D serum values. Thus 49% of patients undergoing VP had 25(OH)D insufficiency compared to 44% of patients in the conservative treatment group. Although several studies have shown a relationship between the presence of low vitamin D serum levels and the increased risk for fragility fractures in osteoporotic individuals, to our knowledge there are currently no data relating this parameter to an increased risk of VF post-VP. However, a recent observational study has reported lower vitamin D serum levels in patients with recurrent VF after kyphoplasty. There are several possible explanations for this increased risk among these patients, such as the secondary increase in bone turnover observed in these individuals.[23, 44, 45] Thus, patients with VF also showed increased values of serum P1NP. Although we can not totally rule out the increased values of bone turnover markers being due to the fracture itself, the significantly increased serum values of P1NP in patients with new VF after VP further suggests a contributory effect of the increased bone turnover. Nonetheless, low vitamin D serum levels probably indicate a high risk for developing fractures in the frail population; ie, older individuals with more severe bone disease. Although there are no data indicating a decrease in the incidence of VF after VP in patients treated with vitamin D, the present results suggest that the the vitamin D deficiency should be corrected before performing this therapeutic approach. However, this factor should be analyzed in future RCT and follow-up studies.
In addition, the number of vertebrae treated per patient also seemed to be associated with the development of VF. Thus, we observed a fourfold increased risk of VF when more than one vertebral body was treated (RR, 4.19; 95% CI, 1.03–34.30, p = 0.044). This data may also partly explain the differences in the increased risk of fractures after VP when compared with other RCT. Thus, in our study the mean number of vertebrae treated was greater than in previous series, with 61% of patients having two or more vertebrae treated per procedure, in contrast to previous randomized studies in which one vertebrae was treated in most patients.[12, 14] All of the above indicate that VP induces a greater modification in the biomechanics of the spine when a higher number of fractures and more severe deformities are present. Clearly, altered biomechanics due to the stiffness of PMMA, as well as the natural course of the underlying osteoporosis, are both contributory factors for new VF after VP.
In the multivariate analysis glucocorticoid therapy was a factor related to VF in the overall population (RR, 3.64; 95% CI, 1.61–8.26, p = 0.0055). This effect was especially notable in patients receiving conservative treatment and has also been reported in previous studies. Conversely, factors such as the type of antiosteoporotic treatment were not associated with new VF. Nevertheless, the low number of patients included in the present study may have influenced these results.
One limitation of this study was the low number of patients included in the analysis. Nonetheless, the precise evaluation of these patients, all included in a RCT, with an extensive analysis of biochemical, clinical, and radiological factors related to the development of new fractures, is one of the strengths of this study, being the only RCT to date showing an increased risk of VF after VP further enhancing the clinical significance of these results.
In conclusion, nearly 30% of patients with osteoporotic VF treated with VP presented a new VF after the procedure. Age, especially >80 years, and low vitamin D serum levels were related to the development of new VF in these patients, as were the number of vertebrae treated per patient and the presence of inferior disk cement leakage after the procedure. All these factors should be taken into account when selecting the patients most likely to benefit from VP with the lowest risk of complications.
All authors state that they have no conflicts of interest.
This study was funded by a grant from the Fundació La Marató de TV3, the Spanish Society of Radiology and Catalan Society of Rheumatology. We thank J. Puig Martínez for contributing to data collection and analysis.
Authors' roles: AMF, JB, and PP contributed to study design, literature research, data interpretation, data collection, and writing. AM and NG contributed to data collection and study design. JM, LS, and ALR contributed to data collection. JLC contributed to data analysis and database design.