The Spine Journal 4 (2004) 669–674

The effects of doxorubicin (adriamycin) on spinal fusion: anexperimental model of posterolateral lumbar spinal arthrodesis

P. Justin Tortolani, MDa, Andrew E. Park, MDb, John Louis-Ugbo, MDc,Emad S. Attallah-Wasef, MDc, Chaiwat Kraiwattanapong, MDc, John G. Heller, MDc,

Scott D. Boden, MDc, S. Tim Yoon, MD, PhDc,*aScoliosis and Spine Center of Maryland St. Joseph Medical Center, O′Dea Medical Arts Bldg,

7505 Osler Drive, Suite 104, Towson, MD 21204-7737, USAbBaylor University Medical Center, 3600 Gaston Avenue, Suite 651, Dallas, TX 75246, USA

cEmory Spine Center, Department of Orthopaedics, Emory University School of Medicine, and the Veteran Affairs Medical Center,2165 North Decatur Road, Atlanta, GA 30033, USA

Received 22 August 2003; accepted 24 May 2004

Abstract BACKGROUND CONTEXT: Malignant spinal lesions may require surgical excision and segmen-tal stabilization. The decision to perform a concomitant fusion procedure is influenced in partby the need for adjunctive chemotherapy as well as the patient’s anticipated survival. Althoughsome evidence exists that suggests that chemotherapy may inhibit bony healing, no informationexists regarding the effect of chemotherapy on spinal fusion healing.PURPOSE: To determine the effect of a frequently used chemotherapeutic agent, doxorubicin, onposterolateral spinal fusion rates.STUDY DESIGN/SETTING: Prospective animal model of posterolateral lumbar fusion.OUTCOME MEASURES: Determination of spinal fusion by manual palpation of excised spines.Plain radiographic evaluation of denuded spines to evaluate intertransverse bone formation.METHODS: Thirty-two New Zealand White rabbits underwent posterior intertransverse processfusion at L5–L6 with the use of iliac autograft bone. Rabbits randomly received either intrave-nous doxorubicin (2.5 mg/kg) by means of the central vein of the ear at the time of surgery (16animals) or no treatment (16 animals; the control group). The animals were euthanized at 5 weeks, andthe lumbar spines were excised. Spine fusion was assessed by manually palpating (by observersblinded to the treatment group) at the level of arthrodesis, and at the adjacent levels proximal anddistal. This provided similar information to surgical fusion assessment by palpation in humans. Fusionwas defined as the absence of palpable motion. Posteroanterior radiographs of the excised spines weregraded in a blinded fashion using a five-point scoring system (0 to 4) devised to describe the amountof bone observed between the L5–L6 transverse processes. Power analysis conducted before initiation ofthe study indicated that an allocation of 16 animals to each group would permit detection of at leasta 20% difference in fusion rates with statistical significance at p�.05.RESULTS: Eleven of the 16 spines (69%) in the control group and 6 of the 16 spines (38%) inthe doxorubicin group fused. This difference was statistically significant (�.038). There was nosignificant correlation (p�.05) between the radiographic grade of bone formation (0 to 4) and fusionas determined by palpation. There were four wound infections in the control group and four in thedoxorubicin group. However, solid fusions were palpated in three of these four spines in boththe control and treatment groups.CONCLUSIONS: No significant differences in wound complications were noted with doxorubicinadministration. A single dose of doxorubicin administered intravenously at the time of surgery appearsto play a significant inhibitory role in the process of spinal fusion. If similar effects occur in

FDA device/drug status: approved for this indication (doxorubicin).Support in whole or in part was received from The North American

Spine Society under Grant 6-37-468. Nothing of value received from acommercial entity related to this research.

1529-9430/04/$ – see front matter � 2004 Elsevier Inc. All rights reserved.doi:10.1016/j.spinee.2004.05.254

* Corresponding author. The Emory Spine Center, 2165 North DecaturRoad, Decatur, GA 30033, USA. Tel.: (404) 778-7155; fax: (404) 778-7117.

E-mail address: tim_yoon@emory.org (S.T. Yoon)

P.J. Tortolani et al. / The Spine Journal 4 (2004) 669–674670

humans, these data suggest that doxorubicin may be harmful to bone healing in a spine fusion ifgiven during the perioperative period. Further investigation will be necessary to determine the effectof time to aid at determining whether doxorubicin administered several weeks pre- or postoperativelyresults in improved fusion rate, and whether bone morphogenetic proteins can overcome theseinhibitory effects. � 2004 Elsevier Inc. All rights reserved.

Keywords: Doxorubicin; Spinal fusion; Inhibition; Chemotherapy; Adriamycin

Introduction

Malignant spinal lesions may require surgical excisionand segmental stabilization. The decision to perform aconcomitant fusion procedure is influenced in part by thepatient’s anticipated survival as well as the need for adjuvantor neo-adjuvant chemotherapy and radiotherapy [1]. Litera-ture regarding the effects of chemotherapy on spinal fusionrate remains scant, and therefore it has been difficult forsurgeons to make rational decisions when faced with thedecision to fuse or not to fuse.

Although the negative effect of steroids, nicotine andnonsteroidal anti-inflammatory drugs in spinal fusion is doc-umented, there is no prior study addressing the effect of acancer chemotherapeutic agent on the rate of spinal fusion[2–4]. Other basic science studies on the effect of certainchemotherapeutic agents on bone biology have been re-ported. Friedlaender et al. [5] studied the short-term effectsof doxorubicin and methotrexate on bone biology in the tailvertebrae of rats. They reported a nearly 60% diminishedbone-formation rate and an 11.5% decrease in net trabecularbone volume in the doxorubicin group. Investigation of low-dose, long-term administration of doxorubicin to rabbits sug-gests a cytotoxic action on osteoblasts and osteoclasts inthe epiphyseal-metaphyseal regions of growing bones [6].Burchardt et al. [7] reported an increased nonunion rateand decreased bone formation in the healing of segmentalcortical autografts in dogs treated with adriamyicin. Further-more, Prevot et al. [8] have demonstrated that prolongedchemotherapy with methotrexate and doxorubicin de-creased bone consolidation during distraction osteogenesisof limb segments. Finally, in a rabbit femoral fracture model,Khoo [9] has reported significant impairment of wound andbone healing in rabbits administered adriamycin 4 or 7 dayspreoperatively compared with controls.

Despite experimental evidence demonstrating an inhibi-tory effect on long-bone growth and healing, doxorubicinremains one of the most commonly used agents in the treat-ment of primary and metastatic spinal neoplasms. Based onour review of the literature, there are no data regarding theeffect of doxorubicin (or any other cancer chemotherapeuticagent) on the rate of spine fusion. The rabbit model presentedherein has been shown to correlate well with the effectof various agents on posterolateral spinal fusion in higheranimals, such as nonhuman primates [10,11]. Furthermore,the pharmacokinetics and tissue penetration of doxorubicinare well defined in the rabbit, because this animal has

been used in earlier investigations on the cardiotoxic andmyelosuppressive effects of doxorubicin, effects that areknown to occur in humans undergoing adriamycin treatment[12–15]. The purpose of this study, therefore, was to deter-mine the effect of a single perioperative dose of doxorubicinon posterolateral spinal fusion in a well-validated animalmodel.

Materials and methods

Rabbit spinal fusion

Thirty-two skeletally mature (1 year of age) female NewZealand White rabbits weighing 4.5 to 5.0 kg were used inthis study. The rabbits were anesthetized by subcutaneousinjection with a mixture of ketamine (35 mg/kg) and xylazine(18 mg/kg) and then maintained with isoflorane 2% by meansof a cone-shaped mask that fit snuggly around the noseand mouth. Each rabbit underwent bilateral intertransverseprocess arthrodesis at L5–L6 and harvesting of bilateral iliaccrests, as previously described [16]. In this technique, oneskin incision is made, which is followed by separate fascialincisions for exposure of the intertransverse regions as well asthe iliac crest. Autologous bone graft, 2.5 cm3, was used oneach side.

Postoperative analgesia was maintained with buprenor-phine for 48 to 72 hours. All animals received perioperativeantimicrobial prophylaxis with one 5-mg/kg dose of Naxcel(Pharmacia and Upjohn Company, Kalamazoo, MI), admin-istered subcutaneously immediately after surgery. The rab-bits were housed one per cage and allowed ad libitum dietof rabbit chow and water. They were monitored daily forgeneral health and neurologic function.

Study groups

The rabbits were randomly divided into two study groups.The doxorubicin group (16 animals) were administered intra-venous doxorubicin (2.5 mg/kg; DOXOrubicin Hydrochlo-ride Injection, Ben Venue Laboratories, Inc., Bedford, OH)by means of the central vein of the ear immediately aftersurgical wound closure. The control group (16 animals)received no intravenous doxorubicin. The surgeons wereblinded to the study groups. The dose of adriamycin (2.5 mg/kg)was chosen because it has been shown in other animal studiesto approximate the therapeutic human interval-dosing

P.J. Tortolani et al. / The Spine Journal 4 (2004) 669–674 671

scheme (60 to 90 mg/m2) with regard to pharmacokinetics,metabolism and distribution in tissues [11–14]. In order tocreate a “worst case scenario” with a single administrationof doxorubicin, a test-dose of 5.0 mg/kg was initially ad-ministered in two rabbits. This dose was chosen because itis considered on the high end of a human equivalent dose(Dr. Barry Cusack, personal communication, 2002.). This5.0-mg/kg dose was fatal to both animals within 24 hours. Wethen administered 2.5 mg/kg to two additional animals whosurvived for 8 weeks without any physical characteristics,such as hair loss, failure to thrive, reduced food intake orgeneral appearance.

Assessment of spine fusions

All of the animals (n=32) were euthanized at 5 weeksafter surgery. This endpoint was chosen based on earlierwork on this model, which demonstrated that fusions reachedtheir peak strength by 5 weeks and that the rate of successfulfusion was no greater at longer periods [16]. The spineswere harvested, denuded of soft tissue, and fusions wereevaluated using manual palpation and posteroanteriorradiographs.

Manual palpationAt the time of harvest, the lumbar spines were manually

palpated (by observers blinded to the treatment group) atthe level of the arthrodesis, and at the adjacent levels proxi-mal and distal. This provides similar information to surgicalfusion assessment by palpation in humans [17] and has beenvalidated biomechanically in the rabbit model [18]. Eachmotion segment was graded as solid, if no motion was pres-ent, or not solid. Only levels graded as solid were consideredfused. We believe that there were no identifiable physicalcharacteristics, such as hair loss, failure to thrive, weight,general appearance or food intake of the specimens in eitherthe control or the treatment group that could have biased themasked observers. On the day of harvest, the animalsevaluated in this study were mixed with animals from otherstudies in our laboratory, such that identification of anysubset of specimens was virtually impossible.

Radiographic analysisPosteroanterior radiographs were made of all specimens

after they were euthanized. The tube-to-plate distance was 90cm. The radiographs were graded in a blinded fashion usingthe following scale devised to assess the amount of inter-transverse bone graft: Grade 0, no bone present betweenthe transverse processes; Grade 1, small islands of bonebetween the transverse processes; Grade 2, bridging bonewith two or more radiolucent lines or one large gap withinthe fusion mass; Grade 3, bridging bone with one radiolucentline in the fusion mass; Grade 4, bridging bone with no gapsor lucent lines.

Statistical methods

Power analysis conducted before initiation of the studyindicated that an allocation of 16 animals to each groupwould permit detection of at least a 20% difference in fusionrates with statistical significance set at p=.05. The Fisherexact test was used for comparisons of proportions betweengroups, with statistical significance set at p=.05.

Results

Mortality and complications

There were no complications or deaths related to theanesthesia. At the dose of doxorubicin used in the study,there was no evidence of systemic toxicity (hair loss orfailure to thrive). One animal that had been administereddoxorubicin was found dead in its cage on postoperative day7 and was eliminated from the study. The cause of death inthis animal could not be identified by postmortem dissec-tion and evaluation of postsurgical history of this animal.There were no cases of postoperative hind leg palsy, aspreviously described [16]. At the time the animals wereeuthanized, four animals in the control group and four ani-mals in the doxorubicin group were found to have subclinicalunilateral wound infections.

Manual palpation and radiographic results

Eleven of the 16 spines (69%) in the control group and6 of the 16 spines (38%) in the doxorubicin group fused (Fig.1). This difference was statistically significant (p�.038).Solid fusions were palpated in three of the four spines com-plicated by postoperative wound infections in both the con-trol and treatment groups.

Based on radiographic analysis, there was a trend towardreduced bone formation in the doxorubicin group com-pared with controls. Fifteen of the 32 intertransverse regions(47%) demonstrated Grade 3 or 4 bone formation in thedoxorubicin group compared with 20 of 32 (63%) in the con-trol animals. This difference was not statistically significant(p�.05). There was a trend toward greater observed bone

Fig. 1. Effect of doxorubicin on spinal fusion rates.

P.J. Tortolani et al. / The Spine Journal 4 (2004) 669–674672

formation in the solidly fused specimens compared withthe nonfused specimens: 21 of the 34 intertransverse regions(62%) demonstrated Grade 3 or 4 bone formation fusedsegments (Fig. 2) compared with 14 of the 30 (47%) Grade3 and 4 scores in the nonfused segments (Fig. 3). This trendtoward reduced bone formation in the nonunion group wasnot statistically significant (p�.05).

Discussion

The main finding in our study is that the rate of posterolat-eral spine fusion in the rabbit animal model is significantlyreduced by the administration of a single dose of doxorubi-cin. To our knowledge, this is the first study to examine theeffect of doxorubicin (adriamycin) on the rate of spinalfusion. A few authors have previously shown that doxorubi-cin alone, or in combination with methotrexate, can be det-rimental to fracture healing or bone remodeling. However,animals in these studies were given multiple doses of doxoru-bicin over the course of the study period [5]. Furthermore,these studies are not necessarily predictive of the effect of

Fig. 2. Radiograph of a rabbit lumbar spine from the control group.This specimen was judged to be solidly fused by manual palpation.Note the presence of continuous bridging bone in the intertransverseregion with peripheral cortication (arrows). This specimen was givena radiographic grade of 4 bilaterally.

doxorubicin on spinal fusion because the biology of spinalfusion is different from fracture healing. The model pre-sented herein has been shown to be predictive of the effectof various agents on posterolateral spinal fusion in higheranimals, such as nonhuman primates [10,11]. Our currentstudy therefore provides key data to confirm the inhibitoryeffect of doxorubicin on posterolateral spinal fusion. Al-though it is likely that our results with posterolateral spinalfusion may be generalizable to other forms of spinal fusion,this has not been clearly established by our current study be-cause the biology of interbody fusion is quite different fromposterolateral fusion.

The mechanism by which doxorubicin inhibits posterolat-eral spine fusion is not known. However, one can speculatethat the known myelosuppressive characteristics of doxoru-bicin may have an effect on spinal fusion biology. Boden etal. [16] detailed three phases of spinal fusion healing in therabbit model. The first phase, which lasts 3 to 4 weeks, ischaracterized histologically by the presence of inflammatorycells and hematoma formation, and later stages are character-ized by endochondral ossification, woven bone and ulti-mately lamellar bone formation. Agents that are toxic to theinflammatory cascade either by reducing the numbers ofcells or disrupting the functions of these cells would beexpected, therefore, to suppress bone graft healing. As isthe case with many chemotherapeutic agents, doxorubicinhas dose-limiting myelosuppressive properties [19,20]. Inhumans, the nadir of leukopenia occurs 14 days after drugadministration and recovery is complete by day 19 [20]. Thetime course for myelosuppression in rabbits is not known.However, other small mammals provide some insights. Four

Fig. 3. Radiograph of a rabbit lumbar spine from the doxorubicingroup. This specimen was not fused based on manual palpation. Notethe presence of at least two radiolucent lines in the intertransverse regionbilaterally (arrows). This specimen was given a radiographic grade of2 bilaterally.

P.J. Tortolani et al. / The Spine Journal 4 (2004) 669–674 673

days after a single intravenous doxorubicin dose in mice(analogous to human doses), a greater than 90% reduction inthe amount of neutrophils has been reported [21]. Similarly,neutrophil count nadirs of 0.8×103/mm3 (lower limit ofnormal�1.4×103) occur in ferrets 7 days after one-time intra-venous doxorubicin dosing [21]. Taken together, this sug-gests one potential mechanism whereby a single dose ofdoxorubicin given perioperatively rather potently inhibitsthe earliest stages of the inflammatory phase of spinal fusion.Whether doxorubicin administered several weeks pre- orpostoperatively results in increased fusion potential will bethe subject of further investigation.

A potential limitation of this study is that we examinedonly the effects of a single chemotherapeutic agent, doxoru-bicin. Despite its widespread use in the treatment of primaryand metastatic spinal malignancies, the findings reportedhere should not be extrapolated to antineoplastic drugs out-side the anthracycline class. Another potential limitation isthat we did not correlate fusions with biomechanical orhistological data. Manual palpation alone was used to gradefusions, because previous studies have shown an excellentcorrelation between the results of biomechanical testing andmanual palpation [16,18]. Furthermore, we did not attemptto classify spinal fusion based on radiographs, because thesehave been shown to be inaccurate up to 54% of the time whencompared with manual palpation [16]. Instead, a scoringsystem was devised for the amount of bone present in theintertransverse region (Grades 0 to 4), and this was correlatedwith fusion status (by manual palpation) or with treatment(doxorubicin versus controls). Interestingly, 20 of 32 (63%)of the control intertransverse regions exhibited Grade 3 or4 bone formation, whereas only 15 of 32 (47%) of theadriamycin treated regions exhibited Grade 3 or 4 bone for-mation. Further, 21 of 34 (62%) of the intertransverse regionsjudged to be fused by palpation exhibited Grade 3 or 4 boneformation, whereas only 14 of 30 (47%) of the intertrans-verse regions judged to be fused by palpation exhibitedGrade 3 or 4 bone formation. Although these results suggesta trend toward reduced bone formation in the doxorubicin-treated or nonfused specimens, they did not achieve statisti-cal significance (p�.05) with the sample sizes in this study.

In conclusion, a single dose of doxorubicin administeredintravenously at the time of surgery appears to play a signifi-cant inhibitory role in the process of spinal fusion. Ourdata suggest that if possible, doxorubicin should not beadministered in the immediate perioperative period to pa-tients in whom a posterolateral spinal fusion has beenperformed.

Acknowledgments

The authors wish to acknowledge the expert veterinarycare and valuable insights of Colleen Oliver, DVM, andSandy Yurevich in executing these experiments. Correspon-dence with Dr. Barry J. Cusack was invaluable regarding

dosing, administration and toxicity of doxorubicin in therabbit model. We thank Thelma Snider for her editorialassistance and the North American Spine Society for grantsupport of this research.

References

[1] Gokaslan ZL. Spine surgery for cancer. Curr Opin Oncol 1996;8:178–81.

[2] Sawin PD, Dickman CA, Crawford NR, Melton MS, Bichard WD,Sonntag VK. The effects of dexamethasone on bone fusion in an exper-imental model of posterolateral lumbar spinal arthrodesis. J Neurosurg2001;94(1 suppl):76–81.

[3] Martin GJ, Boden SD, Titus L. Recombinant human bone morpho-genetic protein-2 overcomes the inhibitory effect of ketorolac, a non-steroidal anti-inflammatory drug (NSAID), on posterolateral lumbarintertransverse process spine fusion. Spine 1999;24(21):2188–94.

[4] Schimandle JH, Boden SD, Hutton WC. Experimental spinal fusionwith recombinant human bone morphogenetic protein-2. Spine 1995;20(12):1326–37.

[5] Friedlaender GE, Tross RB, Doganis AC, Kirkwood JM, Baron R.Effects of chemotherapeutic agents on bone. I. Short-term methotrex-ate and doxorubicin (adriamycin) treatment in a rat model. J BoneJoint Surg Am 1984;66:602–7.

[6] Young DM, Fioravanti JL, Olson HM, Prieur DJ. Chemical and mor-phological alterations of rabbit bone induced by adriamycin. CalcTissue Res 1975;18:47–63.

[7] Burchardt H, Glowczewskie FP, Enneking WF. The effect of adriamy-cin and methotrexate on the repair of segmental cortical autograftsin dogs. J Bone Joint Surg Am 1983;65:103–8.

[8] Prevot J, Poncelet T, Lascombes P, et al. [Osteogenesis in distractionin animals submitted to anti-cancer chemotherapy. Experimental studyin rabbits]. Rev Chir Orthop Reparatrice Appar Mot 1988;74(suppl2):98–100.

[9] Khoo DB. The effect of chemotherapy on soft tissue and bone healingin the rabbit model. Ann Acad Med Singapore 1992;21:217–21.

[10] Suh DY, Boden SD, Louis-Ugbo J, et al. Delivery of recombinanthuman bone morphogenetic protein-2 using a compression-resistantmatrix in posterolateral spine fusion in the rabbit and in the non-human primate. Spine 2002;27:353–60.

[11] Boden SD, Schimandle JH, Hutton WC. 1995 Volvo Award in basicsciences. The use of an osteoinductive growth factor for lumbar spinalfusion. Part II: study of dose, carrier, and species. Spine 1995;20:2633–44.

[12] Cusack BJ, Young SP, Driskell J, Olson RD. Doxorubicin and doxoru-bicinol pharmacokinetics and tissue concentrations following bolusinjection and continuous infusion of doxorubicin in the rabbit. CancerChemother Pharmacol 1993;32:53–8.

[13] Long HJ, Diamond SS, Raflo CP, Burningham RA. Dose-responserelationships of chronic adriamycin toxicity in rabbits. Eur J CancerClin Oncol 1984;20:129–35.

[14] Sturgill MG, August DA, Brenner DE. Hepatic enzyme inductionwith phenobarbital and doxorubicin metabolism and myelotoxicityin the rabbit. Cancer Invest 2000;18:197–205.

[15] Herman EH, Ferrans VJ. Preclinical animal models of cardiac protec-tion from anthracycline-induced cardiotoxicity. Semin Oncol 1998;25:15–21.

[16] Boden SD, Schimandle JH, Hutton WC. An experimental lumbarintertransverse process spinal fusion model. Radiographic, histologic,and biomechanical healing characteristics. Spine 1995;20:412–20.

[17] Silcox DH, Boden SD, Schimandle JH, Johnson P, Whitesides TE,Hutton WC. Reversing the inhibitory effect of nicotine on spinalfusion using an osteoinductive protein extract. Spine 1998;23:291–6.

P.J. Tortolani et al. / The Spine Journal 4 (2004) 669–674674

[18] Erulkar JS, Grauer JN, Patel TC, Panjabi MM. Flexibility analysisof posterolateral fusions in a New Zealand white rabbit model. Spine2001;26:1125–30.

[19] Piscitelli SC, Rodvold KA, Rushing DA, Tewksbury DA. Pharmacoki-netics and pharmacodynamics of doxorubicin in patients with smallcell lung cancer. Clin Pharmacol Ther 1993;53:555–61.

Seventy-fiveYears Agoin Spine

In 1929, Alexander Fleming (nowknown as Sir Alexander Fleming) [1] published hisobservations of the antibacterial action of cultures of a

[20] Benjamin RS. A practical approach to adriamycin (NSC-123127)toxicology. Part 3. Cancer Chemother Rep 1975;6:191–4.

[21] Schurig JE, Schlein A, Florczyk AP, Farwell AR, Bradner WT. Animalmodels for evaluating the myelosuppressive effects of cancer chemo-therapeutic agents. Exp Hematol 1985;13(suppl 16):101–5.

penicillium. Eleven years later, Ernst Boris Chain et al. [2]published their account of the therapeutic effectivenessof penicillin. Fleming, Chain and H.W. Florey shared theNobel Prize in 1945.

References

[1] Fleming A. On the antibacterial action of cultures of a penicillium,with special reference to their use in the isolation of B influenzae.Br J Exp Path 1929;10:226–36.

[2] Chain EB, Florey HW, Gardner AD, et al. Penicillin as a chemo-therapeutic agent. Lancet 1940;2:226–8.