ERCUTANEOUS vertebroplasty has been regarded as aneffective management option for various vertebralcompression fractures, but the development of post-

PVP vertebral fracture is a not uncommon sequela and anadditional vertebroplasty is frequently required. The inci-dence of subsequent vertebral fracture is reported to varyfrom 12 to 52% according to the literature.10,17,25,28,29

It is uncertain whether the vertebroplasty itself is thecause of subsequent vertebral fractures. Although many au-thors have reported the condition as a new fracture aftervertebroplasty, there are also studies explaining the risk ofsubsequent fracture as a natural phenomenon of the dis-

ease.1,18,19 Unfortunately, no valid study has been conductedto contend with this issue.27

To prevent or reduce the occurrence of fracture afterPVP, knowledge of the risk factors for subsequent fractureis essential. It has been reported that adjacent vertebrae arethe most common site of subsequent fractures (incidence41–67%).8,12,28,29 Some biomechanical and clinical studieshave demonstrated the mechanism and causative factors ofadjacent-level vertebral fracture.3,4,6 The authors of thesestudies have named the phenomenon the “pillar” effect or“stress-riser” effect and have attributed this phenomenon tointradiscal cement leakage or maximum filling of cementas the predictive factors. However, these are only proce-dure-related factors, and the patient’s constitution or spinalgeometry has not been taken into consideration. Moreover,there was no adequate explanation regarding the mech-anism, nor were the prediction rules set for nonadjacentneighboring-segment fracture after vertebroplasty.

J. Neurosurg.: Spine / Volume 9 / August 2008

J Neurosurg Spine 9:129–136, 2008

Predictive factors for subsequent vertebral fracture afterpercutaneous vertebroplasty

YONG AHN, M.D., JUNE HO LEE, M.D., HO-YEON LEE, M.D., PH.D., SANG-HO LEE, M.D., PH.D., AND SANG-HYUN KEEM, M.D.

Department of Neurosurgery, Wooridul Spine Hospital, Seoul, Korea

Object. The purpose of this study was to evaluate the predictive factors for subsequent vertebral fracture occur-ring after percutaneous vertebroplasty (PVP) at the neighboring levels (adjacent vs nonadjacent levels).

Methods. The medical records of 508 consecutive patients treated with PVP between January 2000 and Decem-ber 2002 were retrospectively reviewed. A total of 45 patients with 49 painful vertebral fractures occurring afterPVP was identified based on clinical and radiological findings. New vertebral fractures, developing at any of the 3consecutive vertebral bodies (VBs) above or below the previously treated level, were the focus of the study. Thepatients were divided into 3 groups: an adjacent-level fracture group, nonadjacent-level fracture group, and a con-trol group composed of 50 randomly selected patients in whom there was no evidence of a new fracture. Clinical,imaging, and procedure-related factors for each group were statistically analyzed.

Results. In 31 patients 35 VBs were classified as adjacent-level fractures, and in 14 patients 14 VBs were classi-fied as nonadjacent-level fractures. After further vertebroplasty, the overall pain intensity and satisfaction rate in pa-tients with post-PVP fractures were similar to those in the control group. In cases involving adjacent fractures, lowerbody mass index and intradiscal cement leakage were the significant predictive factors of fracture. In contrast, lowermobility of the index segment was related to nonadjacent-level fracture.

Conclusions. According to the authors’ results, the mechanisms of subsequent fracture at adjacent and nonadja-cent vertebrae are different. A direct pillar effect (that is, the difference in strength caused by cement augmentation)may provoke an adjacent-level fracture, whereas a dynamic hammer effect (the difference in segmental mobility)may lead to a nonadjacent fracture. (DOI: 10.3171/SPI/2008/9/8/129)

KEY WORDS • osteoporosis • predictive factor • surgery-related vertebral fracture •vertebroplasty

P

129

Abbreviations used in this paper: BMD = bone mineral density;BMI = body mass index; PMMA = polymethylmethacrylate; PVP =percutaneous vertebroplasty; ROM = range of motion; SD = stan-dard deviation; VAS = visual analog scale; VB = vertebral body.

The objective of this study was to evaluate the predictivefactors for neighboring-segment fracture after vertebro-plasty, taking into account each patient’s constitution, spi-nal geometry, and the procedure-related factors. We alsotried to ascertain the possible mechanism of these subse-quent fractures and to explain how the mechanism differedbetween adjacent and nonadjacent fractures.

Methods

Study Design and Patient Selection

We examined our computerized database to identify allpatients who had undergone PVP at our hospital from Jan-uary 2000 to December 2002. Acute or subacute osteo-porotic vertebral compression fractures were the sole path-ological entity, and cases involving traumatic fractures ormalignancies were excluded from this study. The medicalrecords of 508 consecutively treated patients were retro-spectively reviewed. From the database, we selected casesin which the patients had undergone additional vertebro-plasty to treat a painful vertebral fracture after initial PVP.The inclusion criteria for the new fracture were as follows:1) relapse of pain after initial vertebroplasty, after an obvi-ous pain-free interval; 2) evidence of new VB fracture onboth MR imaging and radiography, occurring within any ofthe 3 consecutive VBs above or below the previously treat-ed level; and 3) additional vertebroplasty required to re-lieve painful symptoms due to a subsequent fracture. Casesinvolving a painless, minor fracture and/or same-segmentfracture were excluded from the study. The patients weredivided into 3 groups: those with adjacent-level fractures(Fig. 1), those with nonadjacent-level fractures (Fig. 2), andthose comprising a control group. We carefully reviewedthe clinical and imaging records to rule out any evidence ofvertebral fracture occurring after PVP during the follow-up

period. There were 50 control patients randomly selectedfrom a database of cases in which there was no evidence ofsubsequent vertebral fracture after initial PVP. The absenceof subsequent vertebral fracture was defined in 2 ways: 1)no pain relapse after initial PVP and 2) no findings of sub-sequent fracture on follow-up radiographs. These findingswere further confirmed by telephone interview or patient’svisit to the outpatient office.

Procedural Detail

The PVP was performed using the technique describedby Jensen et al.14 The patient was placed prone on the radio-lucent table. A fluoroscopically guided uni- or bilateraltranspedicular approach was performed with an 11-gaugebone biopsy needle (Hyun Medical). The ideal needleplacement was at the anterior third of the body. The PMMA(Simplex P; Stryker Howmedica) was mixed with bariumsulfate powder (Biotrace, Bryan) and tobramycin. Themixed bone cement was injected by using 1-ml syringesunder fluoroscopic guidance. The end point was when thecement reached the posterior quarter of the VB or whensignificant leakage occurred.

Comparison Parameters

Data regarding parameters such as age, sex, height, bodyweight, BMI, and BMD were obtained from clinical re-cords. The BMI is calculated as weight in kilograms divid-ed by the square of the height in meters (kg/m2); the lum-bar spine bone density and the femur bone density weremeasured by dual-energy x-ray absorptiometry.7 Spinalgeometric properties such as the location of the fracture,type of the fracture, local kyphosis, and sagittal-planeROM were measured using preoperative radiographs (Fig.3). Simple lateral radiographs with the patient in a neutral

Y. Ahn et al.

130 J. Neurosurg.: Spine / Volume 9 / August 2008

FIG. 1. Radiographic and MR imaging examples of an adjacent-segment fracture after initial PVP in a 68-year-oldwoman with an L-2 vertebral compression fracture. A: Prevertebroplasty images showing L-2 compression fracture(white arrow). B: Radiograph acquired immediately after L-2 vertebroplasty showing intradiscal cement leakage (whitearrow) with intact adjacent vertebrae. C: Images obtained 4 weeks after vertebroplasty (white arrow) demonstratingnew compression fracture at L-1 (black arrow).

position and during flexion and extension were available inall cases. Local kyphosis across the fractured vertebra wascalculated using the Cobb angle measurement on standinglateral radiographs.20 Kyphosis was defined as the anglebetween the lines parallel to the superior endplate of thevertebra 1 level above the fractured vertebra and the in-ferior endplate of the vertebra 1 level below the fracturedvertebra. The sagittal-plane ROM was measured on theflexion/extension lateral radiograph to evaluate segmentalmobility. The adjacent ROM was defined as the sagittal-plane rotation angle between the fractured vertebra and ad-jacent vertebra, and the overall ROM was defined as thetotal sagittal rotation angle of the adjacent 4 contiguous in-tervertebral segments. In most patients, the upper and/orlower 4 consecutive segments were measurable as the long-est vertebral segments from the index vertebra on the later-al radiographs. The mean relative mobility was defined asthe ratio between the adjacent ROM and the overall ROM(adjacent ROM/overall ROM). Procedure-related parame-ters were also analyzed. Any cement leakage beyond theendplate and into the disc was checked on the immediatepostoperative lateral radiograph. However, contact of theinjected cement with either VB endplate was not regardedas intradiscal leakage. The volume of PMMA injected intothe treated segment and the bilaterality of approach werealso documented.

Outcome Evaluation

Pain intensity was rated using a 10-point VAS. Satisfac-tion rate was also assessed at the final follow-up. Each patient was asked the following question: “What is yourlevel of satisfaction regarding the surgical procedure per-formed?” The patient could choose between 3 different as-sessments of satisfaction: very satisfied, moderately satis-fied, or not satisfied. The overall satisfaction rate included

patients who were very satisfied or moderately satisfied.Analysis of variance, the Kruskal–Wallis test, and the chi-square test were used to compare the study groups. Datawere presented as the mean 6 SD. A binary logistic regres-sion analysis was then used to determine if the individualfactors were independently associated with the subsequentvertebral fracture. The level of statistical significance wasp , 0.05.

Results

Demographics and Clinical Outcome

There were 50 patients (55 vertebrae) in the controlgroup in whom there was no evidence of subsequent verte-bral fracture, whereas there were 31 patients with 35 af-fected vertebrae in the adjacent fracture group and 14patients with 14 affected vertebrae in the nonadjacent frac-ture group. Of the 49 new VB fractures, 35 (71.4%) devel-oped at the adjacent level. The mean ages of patients in thecontrol, adjacent fracture, and nonadjacent fracture groupswere 69.6, 69.5, and 69.1 years, respectively. The meanVAS score decreased from 7.3 to 3.8 in the control group,from 7.8 to 3.5 in the adjacent fracture group, and from 8.9to 3.2 in the nonadjacent fracture group. The overall satis-faction rate in each group was 88, 83.9, and 85.7%, respec-tively. No statistically significant differences were found inage, sex, preoperative and final VAS, or the overall satis-faction rate among the groups (Table 1).

Constitutional Factors

Among the constitutional factors, the mean BMI and themean body weight were highest in patients with nonadja-cent-level fractures, followed by those in the control group;they were lowest in patients with adjacent-level fractures.

J. Neurosurg.: Spine / Volume 9 / August 2008

Subsequent fracture after vertebroplasty

131

FIG. 2. Radiographic and MR imaging examples of nonadjacent-segment fracture after initial PVP in a 63-year-oldwoman with a T-12 vertebral compression fracture. A: Prevertebroplasty images showing a T-12 compression fracture(white arrow). B: Radiograph acquired immediately after T-12 vertebroplasty (white arrow) demonstrating intactneighboring vertebrae. C: Images obtained 8 weeks after vertebroplasty (white arrow) showing new compression frac-ture at L-2 vertebra (black arrow).

In particular, the mean BMI of patients in the adjacent frac-ture group was 22.3 kg/m2, which was significantly lowerthan those in the other groups (p , 0.05). On the otherhand, the mean body weight of patients in the nonadjacentfracture group was 60 kg, significantly higher than that inthe other groups (p , 0.05). However, the BMDs were notsignificantly different among the groups. The mean spinalBMD (T- and Z-scores) calculated in the adjacent fracturegroup was relatively lower and that in the nonadjacent frac-ture group was higher than that in the control group, but thedifference was not statistically significant (Table 2).

Spinal Geometry

As for the spinal geometry, the degree of local kyphosisand the level of the fracture was not different among thegroups, although the degree of local kyphosis in the adja-cent-level fracture group was somewhat higher than that inthe other groups. The incidence of fracture occurrence atthe thoracolumbar junction (T11–L2) tended to be higherboth in the nonadjacent fracture group (71.4%) and theadjacent fracture group (67.6%) compared with the controlgroup (59.4%), but the difference was not significant. Themean adjacent ROM in the nonadjacent fracture group(1.9°) tended to be lower than that in the adjacent fracturegroup (3.5°) or the control group (3.6°), but the differencewas not significant (p = 0.22). However, the mean relative

mobility in the nonadjacent fracture group (0.2) was signif-icantly lower than that in the adjacent fracture (0.4) or ofthe control (0.6) group (p = 0.027) (Table 3).

Y. Ahn et al.

132 J. Neurosurg.: Spine / Volume 9 / August 2008

FIG. 3. Radiographs used to define local kyphosis and segmental ROM. A: Local kyphosis around the fractured ver-tebra (asterisk) was measured using the Cobb method on the standing lateral radiograph. The Cobb angle was measuredbetween the superior endplate of the upper vertebra (upper line) and the inferior endplate of the lower vertebra (lowerline). B: The segmental ROM around the fractured vertebra (asterisk) was measured on the flexion/extension lateralradiographs. The adjacent ROM was defined as the sagittal-plane rotation angle between the endplate of the fracturedvertebra (lower line) and the endplate of adjacent vertebra (center line). The overall ROM was defined as the total sagit-tal rotation angle of the adjacent 4 consecutive intervertebral segments (between the upper and lower lines).

TABLE 1Demographic and clinical data obtained in

patients who underwent PVP and sustained anadjacent-segment or remote-segment fracture*

Fracture Group

Variable Control Group Adjacent Nonadjacent p Value

no. of patients 50 31 14no. of vertebrae 55 35 14mean age (yrs) 69.6 69.5 69.1 0.949†M/F ratio 4:46 5:26 2:12 0.430‡mean VAS pain score§preop 7.3 6 2.1 7.8 6 2 8.9 6 1.1 0.1†postop 3.8 6 1.9 3.5 6 1.8 3.2 6 0.9 0.521†

overall satisfaction rate (%) 88 83.9 85.7 0.918‡very satisfied 9 4 1mod satisfied 35 22 11not satisfied 6 5 2

* mod = moderately.† Comparison made using analysis of variance (ANOVA).‡ Comparison made using the chi-square test.§ Mean values are presented 6 SD.

Procedure-Related Factors

As for the procedure-related factors, the mean volume ofPMMA injected per VB in the adjacent fracture group (6.4ml) was larger than that in the nonadjacent fracture group(5.1 ml) or control group (5.8 ml), but the variation was notsignificant (p = 0.96). However, intradiscal PMMA leakagewas strongly related to the development of an adjacent-seg-ment fracture. The rate of intradiscal cement leakage in theadjacent fracture group was 54.3%, which was significant-ly higher than that in the other groups (21.4 and 20.8% forthe nonadjacent and control groups, respectively) (p =0.004) (Table 4). The method of approach (bi- or unilater-al) showed no significant intergroup difference.

Multivariate Analysis for Subsequent Vertebral Fracture

In patients with an adjacent-level fracture, both lowerBMI (p = 0.013) and presence of intradiscal cement leak-age (p = 0.002) were significant. In patients with a nonad-jacent-segment fracture, relative mobility (p = 0.024) wasthe most important predictive factor. After multivariateanalysis, a higher body weight (p = 0.728) was not signifi-cantly associated with a nonadjacent-level fracture.

Discussion

For patients with a painful osteoporotic compression

fracture, vertebroplasty is widely accepted as a minimallyinvasive treatment modality, but an unexpected subsequentor additional postoperative fracture may ensue and requirefurther interventions. The causal relationship between theprocedure and the subsequent fracture is unclear, and thereare 2 issues that remain debatable. The first question iswhether the vertebroplasty itself causes subsequent frac-tures. Some authors have explained the occurrence of asubsequent fracture as a progression of the underlying dis-ease,1,18,19 whereas others have suggested that the cementaugmentation and increased physical activity after verte-broplasty may have played a role.5,10,16,17,28,29 The secondquestion concerns which level could be affected after initialtreatment. Thus far, the adjacent vertebra is regarded as themost common site where a subsequent fracture develops.However, there is no predictive rule for the nonadjacentvertebral fracture. Both of these issues, especially the latter,had to be thoroughly addressed in the current study, andtherefore we evaluated the risk factors for the nonadjacentfracture group as well as for the adjacent fracture group.

In general, it has been believed that the patients with alow BMI and low body weight are at higher risk of recur-rent fracture of the spine or the hip,9,11,26 but there have beenfew reports on the effect of constitutional factors on thedevelopment of subsequent spinal fracture after verte-broplasty. As the result of univariate analysis in our study,BMI and body weight were found to be the major constitu-tional factors affecting postprocedural vertebral fracture.Slenderly built patients with osteoporotic bone and lowerBMI ran the risk of sustaining an adjacent-level fractureafter vertebroplasty, whereas relatively heavier patientswith osteoporotic bone ran the risk of sustaining a nonad-jacent-level fracture. The multivariate analysis demonstrat-ed that the adjacent-segment fracture was closely related tolower BMI, whereas the nonadjacent-level fracture was notrelated to any constitutional factors. This finding suggeststhat the mechanism of the subsequent fracture might be dif-ferent according to contiguous or noncontiguous relation-ship of the lesion to the initially treated level. The reasonfor this phenomenon is unclear. We postulate that lowerBMI may be associated with weaker vertebra, which is thensubject to the direct pillar effect of adjacent vertebral aug-mentation. The degree of osteoporosis may be a potentialpredictor for subsequent fracture,21,24 but there was no inter-group difference in the mean BMD in our study population.The BMD may be affected by degenerative changes or

J. Neurosurg.: Spine / Volume 9 / August 2008

Subsequent fracture after vertebroplasty

133

TABLE 2Comparison of groups by constitutional factors*

Fracture Group

Variable Control Group Adjacent Nonadjacent p Value†

spinal BMDT-score –3.2 6 1.0 –3.3 6 1.1 –2.9 6 1.2 0.504Z-score –1.1 6 0.9 –1.4 6 1.0 –0.9 6 1.0 0.234

femur BMDT-score –3.5 6 0.9 –3.4 6 1.1 –3.4 6 0.8 0.877Z-score –1.5 6 0.9 –1.4 6 1 –1.3 6 0.7 0.786

weight (kg) 54.9 6 8.5 52.5 6 7.5 60.1 6 10.7 0.021height (cm) 151.6 6 7.5 153.3 6 7.9 156.2 6 8.3 0.145BMI (kg/m2) 23.8 6 3.1 22.3 6 2.6 24.6 6 3.3 0.023

* Values are presented as mean 6 SD.† Comparison made using ANOVA.

TABLE 3Comparison of groups by spinal geometry*

Fracture Group

Variable Control Group Adjacent Nonadjacent p Value

local kyphosis (°) 7.6 6 18.5 7.8 6 17.1 6.3 6 15.1 0.791†level (T11–L2) (%) 59.4 67.6 71.4 0.567‡ROMoverall (°) 10.8 6 9.6 9.6 6 5.6 11.8 6 5 0.416†adjacent (°) 3.6 6 3.1 3.5 6 3.1 1.9 6 1.6 0.22†

relative mobility§ 0.6 6 1 0.4 6 0.3 0.2 6 0.1 0.027†

* Values are presented as mean 6 SD, where appropriate.† Comparison made using the Kruskal–Wallis test.‡ Comparison made using the chi-square test.§ Relative mobility is the adjacent ROM/overall ROM.

TABLE 4Comparison of groups by procedure-related factors*

Fracture Group

Variable Control Group Adjacent Nonadjacent p Value

PMMA vol (ml)per vertebra 5.8 6 2.9 6.4 6 2 5.1 6 1.8 0.96†total 8.1 6 4.4 7.1 6 2.9 8.9 6 5.8 0.803†

intradiscal leakage (%) 20.8 54.3 21.4 0.004‡unilat approach (%) 87 81.8 78.6 0.646‡

* Values are presented as mean ± SD, where appropriate.† Comparison made using the Kruskal–Wallis test.‡ Comparison made using the chi-square test.

osteoarthropathy. Thus, when the weak vertebra is associ-ated with spinal degenerative osteoarthropathy, a highBMD does not necessarily indicate stronger bones, and therisk of vertebral fracture is not diminished.2,23 Shiraki et al.22

have reported that BMD was significantly related to BMI,and the incidence of vertebral fracture was negatively cor-related with BMI in elderly women. According to our re-sults, we also postulate that BMI is a better indicator of sub-sequent vertebral fracture than BMD. In our study, thecharacteristics of segmental mobility differed among thegroups. The mean relative mobility of the adjacent segmentwas significantly lower in patients with nonadjacent-levelfractures. In other words, if the adjacent segment is rela-tively rigid, the possibility of subsequent fracture is higherfor a remote segment than it is for the adjacent segment.When the adjacent segment is relatively rigid or less mo-bile, the axial loading from daily activities of living or theeffect of cement augmentation may affect the more mobileremote segment. The higher mobility gradient may producea hammerlike effect on the remote vertebra, causing a sub-sequent fracture. We postulate that increased physical ac-tivity after the vertebroplasty may promote this phenome-non.

Besides segmental mobility, another possible geometri-cal factor is local kyphosis. Theoretically, kyphotic align-ment may exaggerate the hammer effect, causing a subse-quent fracture, but in our study the degree of local kyphosisor the level of the vertebra was not significantly related tothe subsequent fracture.

Over the past few years, many convincing studies havebeen undertaken on the subject of cement volume andcement leakage. In previous studies the authors have re-ported that intradiscal cement leakage might increase thestress to the adjacent vertebra and cause a new fracture.17,24

These findings concur with results of our current study,because we also found that intradiscal PMMA leakage dur-ing the initial vertebroplasty was the main cause of adja-cent-level fracture. As for the injected cement volume, the-oretically, maximum filling with a large amount of cementmay increase the stress to the adjacent vertebra and occa-sionally may cause a subsequent fracture. There is somecontroversy about this theory according to various centersof research. Many authors have commented on this pheno-menon.3,6,24 Others, however, have reported that the cementvolume was not related to the outcome.13,15 In our series, thevolume of PMMA was not significantly related to a subse-quent vertebral fracture.

In summary, the significant predictive factors for thepost-PVP development of adjacent-level fracture were low-er BMI and the presence of intradiscal cement leakage. Thefactor associated with subsequent nonadjacent-level frac-ture was lower mobility of the index segment causing amobility gradient. Based on the present results, we postu-late that the mechanisms of subsequent fracture are dif-ferent for adjacent- and nonadjacent-segment fracture con-ditions. The mechanism for adjacent fractures can beexplained by the direct pillar effect (Fig. 4). Cement aug-mentation of a very weak bone with intradiscal PMMAleakage may increase the strength gradient, leading to anadjacent fracture. In contrast, the mechanism of a nonadja-cent fracture can be explained by the dynamic hammereffect (Fig. 5). If the adjacent segment is already rigid, thepillar effect is not prominent through the adjacent segment.In that case, the augmentation strength may affect a mobileremote segment. The mobility gradient between a rigidadjacent segment and a relatively mobile remote segmentmay cause a subsequent fracture.

There may be some contradictory factors, because our

Y. Ahn et al.

134 J. Neurosurg.: Spine / Volume 9 / August 2008

FIG. 4. Illustrations showing the mechanism of adjacent-segment fracture after initial PVP. A: The damaged verte-bra requiring initial vertebroplasty. B: Cement augmentation of a weak, fractured VB with intradiscal leak may increasethe strength gradient. C: Consequently an adjacent-level fracture ensues.

study was performed retrospectively. Moreover, the studyincluded only patients who underwent repeated interven-tion, not those patients with subsequent fracture only treat-ed with conservative management. Therefore, the trueprevalence of subsequent fracture could not be calculated.However, our study also has considerable validity. The pa-tients of each study group were compared under strict in-clusion criteria, and the cases were selected from a rela-tively large patient population (508 consecutive patients).In our study, postoperative factors such as the type of med-ical treatment or the level of physical activity after the pro-cedure were not considered. Increased physical activity af-ter the vertebroplasty may cause subsequent factures.16,24,29

However, we mainly focused on preoperative or procedure-related factors in our study. The risk of postoperative activ-ity-related subsequent fracture may be one of the interest-ing issues for future studies to consider. Finally, there is afurther question that needs to be addressed. The consisten-cy of the injected cement material may be important. Ba-roud et al.3 have suggested that the use of softer cement ma-terial may be helpful in preventing a subsequent vertebralfracture. However, we could not evaluate the effect ofcement softness because the same kind of bone cement wasused in all the patients. We believe it will be very importantfor a future study to examine the effects of bone cementswith varying strength and hardness.

Conclusions

The risk factors and mechanisms of subsequent fractureat adjacent and nonadjacent vertebrae are different. A

strength gradient caused by bone cement augmentationmay provoke a subsequent adjacent fracture (direct pil-lar effect), whereas a mobility gradient between neighbor-ing segments may cause a subsequent nonadjacent fracture(dynamic hammer effect).

Acknowledgments

We sincerely thank Y. C. Lee, I. S. Cho, M. S. Kim, J. M. Son,and J. H. Kim for their help in preparing the manuscript and figures.

References

1. Ananthakrishnan D, Berven S, Deviren V, Cheng K, Lotz JC, Xu Z, et al: The effect on anterior column loading due to differ-ent vertebral augmentation techniques. Clin Biomech (BristolAvon) 20:25–31, 2005

2. Arden NK, Griffiths GO, Hart DJ, Doyle DV, Spector TD: Theassociation between osteoarthritis and osteoporotic fracture: theChingford Study. Br J Rheumatol 35:1299–1304, 1996

3. Baroud G, Heini P, Nemes J, Bohner M, Ferguson S, Steffen T:Biomechanical explanation of adjacent fractures following verte-broplasty. Radiology 229:606–608, 2003

4. Baroud G, Nemes J, Heini P, Steffen T: Load shift of the inter-vertebral disc after a vertebroplasty: a finite-element study. EurSpine J 12:421–426, 2003

5. Barr JD, Barr MS, Lemley TJ, McCann RM: Percutaneous verte-broplasty for pain relief and spinal stabilization. Spine 25:923–928, 2000

6. Berlemann U, Ferguson SJ, Nolte LP, Heini PF: Adjacent verte-bral failure after vertebroplasty. A biomechanical investigation. JBone Joint Surg Br 84:748–752, 2002

7. Daniels ED, Pettifor JM, Schnitzler CM, Russell SW, Patel DN:Ethnic differences in bone density in female South African nurs-es. J Bone Miner Res 10:359–367, 1995

J. Neurosurg.: Spine / Volume 9 / August 2008

Subsequent fracture after vertebroplasty

135

FIG. 5. Illustrations demonstrating the mechanism of nonadjacent-segment fracture after initial PVP. A: The dam-aged vertebra requiring initial vertebroplasty. B: If the adjacent segment is rigid and less mobile, the pillar effectbetween the adjacent vertebrae is not prominent, while the augmentation strength may affect a more mobile remote seg-ment. C: Therefore, the mobility gradient between a rigid adjacent segment and a relatively mobile remote segmentmay cause a subsequent fracture.

8. Fribourg D, Tang C, Sra P, Delamarter R, Bae H: Incidence ofsubsequent vertebral fracture after kyphoplasty. Spine 29:2270–2277, 2004

9. Fujiwara S, Kasagi F, Yamada M, Kodama K: Risk factors for hipfracture in a Japanese cohort. J Bone Miner Res 12:998–1004,1997

10. Grados F, Depriester C, Cayrolle G, Hardy N, Deramond H,Fardellone P: Long-term observations of vertebral osteoporoticfractures treated by percutaneous vertebroplasty. Rheumatology(Oxford) 39:1410–1414, 2000

11. Gunnes M, Lehmann EH, Mellstrom D, Johnell O: The relation-ship between anthropometric measurements and fractures inwomen. Bone 19:407–413, 1996

12. Harrop JS, Prpa B, Reinhardt MK, Lieberman I: Primary and sec-ondary osteoporosis’ incidence of subsequent vertebral compres-sion fractures after kyphoplasty. Spine 29:2120–2125, 2004

13. Hodler J, Peck D, Gilula LA: Midterm outcome after vertebro-plasty: predictive value of technical and patient-related factors.Radiology 227:662–668, 2003

14. Jensen ME, Evans AJ, Mathis JM, Kallmes DF, Cloft HJ, Dion JE:Percutaneous polymethylmethacrylate vertebroplasty in the treat-ment of osteoporotic vertebral body compression fractures: tech-nical aspects. AJNR Am J Neuroradiol 18:1897–1904, 1997

15. Kaufmann TJ, Trout AT, Kallmes DF: The effects of cement vol-ume on clinical outcomes of percutaneous vertebroplasty. AJNRAm J Neuroradiol 27:1933–1937, 2006

16. Kim SH, Kang HS, Choi JA, Ahn JM: Risk factors of new com-pression fractures in adjacent vertebrae after percutaneous verte-broplasty. Acta Radiol 45:440–445, 2004

17. Lin EP, Ekholm S, Hiwatashi A, Westesson PL: Vertebroplasty:cement leakage into the disc increases the risk of new fracture ofadjacent vertebral body. AJNR Am J Neuroradiol 25:175–180,2004

18. Lindsay R, Burge RT, Strauss DM: One year outcomes and costsfollowing a vertebral fracture. Osteoporos Int 16:78–85, 2005

19. Lindsay R, Silverman SL, Cooper C, Hanley DA, Barton I, BroySB, et al: Risk of new vertebral fracture in the year following afracture. JAMA 285:320–323, 2001

20. Phillips FM, Ho E, Campbell-Hupp M, McNally T, Todd WetzelF, Gupta P: Early radiographic and clinical results of balloonkyphoplasty for the treatment of osteoporotic vertebral compres-sion fractures. Spine 28:2260–2267, 2003

21. Ross PD, Davis JW, Epstein RS, Wasnich RD: Pre-existing frac-

tures and bone mass predict vertebral fracture incidence in wom-en. Ann Intern Med 114:919–923, 1991

22. Shiraki M, Ito H, Fujimaki H, Higuchi T: Relation between bodysize and bone mineral density with special reference to sex hor-mones and calcium regulating hormones in elderly females.Endocrinol Jpn 38:343–349, 1991

23. Sornay-Rendu E, Munoz F, Duboeuf F, Delmas PD: Disc spacenarrowing is associated with an increased vertebral fracture risk in postmenopausal women: the OFELY Study. J Bone MinerRes 19:1994–1999, 2004

24. Syed MI, Patel NA, Jan S, Harron MS, Morar K, Shaikh A: Intra-diskal extravasation with low-volume cement filling in percuta-neous vertebroplasty. AJNR Am J Neuroradiol 26:2397–2401,2005

25. Syed MI, Patel NA, Jan S, Harron MS, Morar K, Shaikh A: Newsymptomatic vertebral compression fractures within a year fol-lowing vertebroplasty in osteoporotic women. AJNR Am J Neu-roradiol 26:1601–1604, 2005

26. Thomas T, Barou O, Vico L, Alexandre C, Lafage-Proust MH:Recurrence of vertebral fracture with cyclical etidronate therapyin osteoporosis: histomorphometry and x-ray microanalysis eval-uation. J Bone Miner Res 14:198–205, 1999

27. Trout AT, Kallmes DF: Does vertebroplasty cause incident verte-bral fractures? A review of available data. AJNR Am J Neuro-radiol 27:1397–1403, 2006

28. Trout AT, Kallmes DF, Kaufmann TJ: New fractures after verte-broplasty: adjacent fractures occur significantly sooner. AJNRAm J Neuroradiol 27:217–223, 2006

29. Uppin AA, Hirsch JA, Centenera LV, Pfiefer BA, Pazianos AG,Choi IS: Occurrence of new vertebral body fracture after percuta-neous vertebroplasty in patients with osteoporosis. Radiol-ogy 226:119–124, 2003

Manuscript submitted February 26, 2008.Accepted April 22, 2008.Sources of support: This study was supported by a grant from

Wooridul Spine Foundation.Address correspondence to: Yong Ahn, M.D., Department of

Neurosurgery, Wooridul Spine Hospital, 47-4 Chungdam-dong,Gangnam-gu, Seoul 135-100, South Korea. email: ns-ay@hanmail.net.

Y. Ahn et al.

136 J. Neurosurg.: Spine / Volume 9 / August 2008