Te

SCIENTIFIC ARTICLE

Management of Radial Nerve Palsy AssociatedWith

Humeral Shaft Fracture: A Decision Analysis Model

Julius Bishop, MD, David Ring, MD, PhD

Purpose When managing radial nerve palsy associated with a humerus fracture, both surgeonand patient must balance the risks and benefits of performing an invasive surgical procedureto address a functional deficit that is likely, but not certain, to recover with nonsurgicalmanagement. The purpose of this study was to better understand the determinants of optimalmanagement strategy using expected-value decision analysis.

Methods Probabilities for the occurrences of the potential outcomes after initial observationor early surgery were determined from a systematic review of the literature. Scores for theseoutcomes were obtained from a questionnaire on patient preferences completed by 82subjects without a history of humerus fracture and radial nerve palsy and used in the modelas a measure of utility. A decision tree was constructed, fold-back analysis was performedto determine optimal treatment, and sensitivity analyses were used to determine the effect ondecision making of varying outcome probabilities and utilities.

Results Observation was associated with a value of 8.4 and early surgery a value of 6.7 given theoutcome probabilities and utilities studied in this model, making observation the optimal man-agement strategy. When parameters were varied in sensitivity analysis, it was noted that when therate of recovery after initial observation falls below 40% or when the utility value for successfulearly surgery rises above 9.4, early surgery is the preferred management strategy.

Conclusions Initial observation was the preferred strategy. In clinical settings in which the likelihoodof spontaneous recovery of nerve function is low or when an informed patient has a strong preferencefor surgery, early surgery may optimize outcome. (J Hand Surg 2009;34A:991–996. © 2009Published by Elsevier Inc. on behalf of the American Society for Surgery of the Hand.)

Type of study/level of evidence Economic and Decision Analysis II.Key words Decision analysis, fracture, humerus, nerve palsy, radial nerve.

ensoempar

ctd

HE OPTIMAL MANAGEMENT strategy for radialnerve palsy in the setting of a humeral shaftfracture remains controversial.1,2 Proponents of

arly surgical exploration of the nerve cite a technically

From the Department of Orthopaedic Surgery, Massachusetts General Hospital, Boston, MA.

Received for publication July 20, 2008; accepted in revised form December 24, 2008.

Support was received from Small Bone Innovations, Smith and Nephew, Wright Medical, Tornier,and Acumed.

Corresponding author: David Ring, MD, PhD, Department of Orthopaedic Surgery, Massachu-setts General Hospital, Yawkey Center, Suite 2100, 55 Fruit Street, Boston, MA 02114; e-mail:dring@partners.org.

0363-5023/09/34A06-0003$36.00/0

vdoi:10.1016/j.jhsa.2008.12.029

asier procedure, the ability to immediately classify theerve injury, and a chance to stabilize the fracture andoft tissues as important advantages.3–5 Advocates ofbservation herald the high rate of spontaneous recov-ry and caution against the morbidity of surgical treat-ent of an injury that would have recovered with ex-

ectant management.6–13 Definitive high-quality datare lacking, resulting in substantial variation in opinionegarding the optimal management strategy.

Expected-value decision analysis is a methodologi-al tool that is based in gaming theory and allows forhe quantitative analysis of decision making under con-itions of uncertainty.14–16 The process of expected-

alue decision analysis involves the creation of a deci-

© Published by Elsevier, Inc. on behalf of the ASSH. � 991

992 HUMERUS FRACTURE AND RADIAL NERVE DECISION ANALYSIS

sion tree to structure the decision problem, thedetermination of outcome probabilities and utilities,fold-back analysis to determine the optimal decision-making strategy, and sensitivity analysis to determinethe effect of varying outcome probabilities and utilitieson decision making. In this way, decision analysis mayallow the clinician and the patient to optimize decisionmaking based on best-available evidence and patientpreferences. It also helps to determine the most impor-tant factors affecting management strategies and thedecision-making process, which may not always beintuitive and can help determine the best questions toaddress in future research.

Decision analysis requires simplification of a com-plex clinical situation into a simple model. It also re-quires that the clinical scenarios represented in themodel can be communicated to the general public withsufficient clarity that they can assign utilities to variousoutcomes. Our approach was based on an interpretationof the existing literature that concludes that nearly allintact nerves recover, thereby reserving nerve graftingfor lacerated nerves and tendon transfers for any nervethat does not recover no matter the prior treatment.

The purpose of this study was to use expected-valuedecision analysis to determine the optimal managementstrategy—initial observation or early surgical explora-tion—for a humerus fracture with an associated radialnerve palsy according to the best available data andpatient preferences based on survey results while alsodetermining the factors with greatest influence on themodel.

MATERIALS AND METHODSInstitutional review board approval was obtained for thisstudy. Analysis was performed using statistical software(the Treeage Pro software package; Treeage Software,Williamstown, MA).

Outcome probabilities

Outcome probabilities were determined by review ofthe literature for articles regarding humeral shaft frac-ture and radial nerve palsy. This body of literature wasrecently summarized in a review by Shao et al.1 Theseauthors systematically searched the literature and iden-tified 391 citations relevant to humerus fracture andradial nerve palsy. Of these, 35 studies involving 1045patients met inclusion criteria and were used to deter-mine the likelihood of recovery of radial nerve functionafter initial observation, early surgery, and late surgeryfor failed initial observation. These authors synthesized

data published as early as 1967 and as recently as 2004.

JHS �Vol A, July

We updated this search using the same inclusion criteriaand identified no additional studies meeting inclusioncriteria. It was assumed that the probability of bonyunion was the same in the surgical and nonsurgicaltreatment groups.

Outcome utilities

Utilities represent patient preferences for various dis-ease states. Outcome utility values were obtained from82 adult subjects at a major orthopedic clinic. Thesewere not patients but rather healthy escorts accompa-nying patients at the time of a clinic visit. Subject agesranged in age from 18 to 82 years, and any subject witha history of “arm fracture” was excluded. The samesurvey was also collected from 20 orthopedic residentswho served as an “expert panel” of people knowledge-able about the various outcomes. The questionnaireposed scenarios for the different outcomes and askedsubjects to rate these outcomes on a scale from 0 to 10,where 0 represented the worst possible outcome and 10represented the best possible outcome (Appendix; thisappendix can be viewed at the Journal’s web site atwww.jhandsurg.org). The scale was anchored at 0 bythe outcome of late surgical intervention followed by norecovery of nerve function and was anchored at 10 bythe outcome of nonsurgical treatment followed by re-covery of nerve function.

Decision tree and fold-back analysis

A decision tree was constructed with 1 decision node, 3chance nodes, and 5 terminal nodes (Fig. 1). The deci-sion tree demonstrates 2 differing management strate-gies for radial nerve palsy in the setting of ipsilateralhumeral shaft fracture. The decision node divides into 2branches: initial observation and early surgical explo-ration. Both branches are followed by various chancenodes, each one terminating in a discrete clinical out-come. Per convention, utility data were placed to theright of the terminal nodes, and probability data wereplaced under the terminal nodes.

Fold-back analysis was performed to identify theoptimal strategy. This involves multiplying each out-come utility by its associated probability, providing an“expected value” for each clinical end point. The ex-pected values for each end point can then be summedfor a given management strategy and the ultimate ex-pected values of the different strategies compared. Themanagement strategy associated with the highest ex-pected value is optimal for the given outcome utilities

and probabilities.

–August

HUMERUS FRACTURE AND RADIAL NERVE DECISION ANALYSIS 993

Sensitivity analysis

One-way sensitivity analysis was performed to modelthe effect on decision making of varying the utility ofrecovery after early surgery (Fig. 2) and the probabilityof recovery after initial observation (Fig. 3). These werethe 2 variables found to affect the decision-makingstrategy within their clinically plausible ranges.

Two-way sensitivity analysis was performed to modelthe effect on decision making of simultaneouslyvarying both the utility of recovery after early sur-gery and the probability of recovery after watchfulwaiting (Fig. 4).

Finally, 1-way sensitivity analysis was also performedto model the prospect of early tendon transfer surgeryafter failed initial observation. This was done bysetting the probability of “recovery” at 100% and

FIGURE 1: The decision tree for observation versus earlyDecision nodes are represented by a square, chance nodes aretriangle. Mean outcome utilities (0–10) are listed to the rightthe terminal node title.

FIGURE 2: One-way sensitivity analysis for the utility of successfulearly surgery (“uSuccessfulEarlySurgery”). The probability of suc-cessful early surgery is varied on the x axis. The lines represent theexpected value (EV) for the decision to manage with observation andthe decision to perform early surgery. Above the threshold value (autility of early surgery of 9.4), early surgery is favored.

then varying the utility of this outcome (Fig. 5).

JHS �Vol A, July

RESULTS

Outcome probabilities and utilities

For initial nonsurgical treatment, the probability ofrecovery after observation was 72% (315 of 435).For patients who failed to recover after initialobservation, there was a 69% (72 of 104) like-lihood of recovery with late nerve surgery, leavinga 31% (32 of 104) probability of no recovery at theend of treatment. After early surgery, the proba-bility of recovery was 85% (188 of 222), leaving a15% (34 of 222) probability of no recovery at theend of treatment.1

Recovery after late nerve surgery was found tohave a mean patient-derived utility of 6.4 (SD 2.5).Recovery after early nerve surgery had a meanutility of 7.4 (SD 2.3), and failure to recover after

ry for humerus fracture associated with radial nerve palsy.resented by a circle, and terminal nodes are represented by ae terminal node. Outcome probabilities (0–1) are listed under

FIGURE 3: One-way sensitivity analysis for the probability ofrecovery with initial observation. The probability of recoveryis varied on the x axis. The lines represent the expected value(EV) for the decision to manage with observation and thedecision to perform early surgery. Below the threshold value(a probability of recovery of with observation of 40%), earlysurgery is favored.

surgerep

of th

early nerve surgery had a mean utility value of 2.7

–August

994 HUMERUS FRACTURE AND RADIAL NERVE DECISION ANALYSIS

(SD 2.6) (Fig. 1). These utilities were not signifi-cantly different (p � .05) from the values derivedfrom the survey of orthopedic residents.

Decision analysis

Fold-back analysis identified initial observation as theoptimal decision strategy. The expected value for ob-servation was 8.4, and the expected value of earlysurgery was 6.7 (Fig. 1).

One-way sensitivity analysis was performed tomodel the effect on decision making of varying theutility of recovery after early surgery (Fig. 2) and theprobability of recovery after observation (Fig. 3). Earlysurgery was favored when the utility of recovery afterearly surgery was greater than 9.4 or when the proba-bility of recovery after initial observation was less than40%.

Two-way sensitivity analysis was performed tomodel the effect on decision making of simultaneouslyvarying both the utility of recovery after early surgeryand the probability of recovery after observation. Thepreferred decision path for the range of values of both“utility of recovery after early surgery” and “probabilityof no recovery after initial observation” is demonstratedgraphically by the shaded areas in Figure 4.

The model exploring early tendon transfer afterfailed initial observation showed that when utility fortendon transfer is greater than 4.4, this becomes thepreferred strategy compared with that of late nerve

FIGURE 4: Two-way sensitivity analysis for the probability ofrecovery after initial observation (x axis) versus the utility ofsuccessful early surgery (“uSuccessfulEarlySurgery”; y axis).The labeled areas on the plot represent the preferred decisionpath for the range of values of both the probability of recoverywith observation (x axis) and the utility of successful earlysurgery (y axis).

surgery (Fig. 5).

JHS �Vol A, July

DISCUSSIONThe management of radial nerve injury associated with aclosed fracture of the humeral diaphysis is debated.3–13

The decision between initial surgical or nonsurgicaltreatment of the radial nerve injury lends itself wellto expected-value decision analysis because of well-defined treatment options and relatively discrete out-comes. The principal advantage of observation is thefact that most nerves are intact and recover, and theprincipal advantage of early surgery is that it mayprovide the best opportunity for nerve recovery whena nerve laceration is identified and grafted in a timelyfashion.

Our decision analysis determined that initial obser-vation is the optimal decision path, given the outcomeprobabilities derived from the literature and the utilitiesobtained from surveys. This finding is in accordancewith a number of previous studies addressing this issue.Perhaps more importantly, our analysis identified theimportant variables in the decision-making process.

FIGURE 5: A The decision tree for nerve surgery versustendon transfer for persistent radial nerve palsy after initialobservation. Decision nodes are represented by a square,chance nodes are represented by a circle, and terminalnodes are represented by a triangle. Mean outcome utilities(0 –10) are listed to the right of the terminal node. Outcomeprobabilities (0 –1) are listed under the terminal node title.B One-way sensitivity analysis for the utility of tendontransfer (“uTendonTransfer”) for persistent radial nerve palsyafter initial observation. The utility of tendon transfer is variedon the x axis. The lines represent the expected value (EV) forthe decision to perform nerve surgery and the decision toperform tendon transfer. Above the threshold value (a utilityfor tendon transfer of 4.4), tendon transfer surgery is favored.

Sensitivity analysis determined that early exploration

–August

HUMERUS FRACTURE AND RADIAL NERVE DECISION ANALYSIS 995

was favored when the probability of recovery afterinitial observation was less than 40%. In certain scenar-ios, such as open humerus fracture, the probability ofnerve laceration exceeds 40%, and early explorationwould be the optimal decision path.17,18 In addition,other scenarios that fall outside the purview of thisdecision analysis might favor early exploration of theradial nerve: for instance, when surgical treatment ofthe humerus fracture is elected for other reasons (eg,concomitant forearm injury or “floating elbow”), thenadvantage should be taken of this opportunity to exam-ine the dysfunctional radial nerve.2

Sensitivity analysis also determined that early explo-ration was favored when the utility for recovery afterearly surgery was greater than 9.4. There are certainlysome “risk-taking” patients in clinical practice who optfor plate fixation over closed treatment of uncompli-cated humerus fractures, and these same patients wouldlikely favor exploration in the setting of radial nervepalsy.

Finally, we also used our model to explore the role oftendon transfers rather than nerve surgery in caseswhere radial nerve function fails to recover after obser-vation. Tendon transfers for radial nerve dysfunctionprovide rapid and reliable return of hand opening butcannot restore sophisticated extensor tendon func-tion.19–21 Sensitivity analysis in this case indicated thatassuming a 100% recovery rate, tendon transfer is thepreferred management strategy when utility for tendontransfer type function is greater than 4.3. A strategy ofinitial observation followed by tendon transfer if nervefunction fails to return may be particularly appealing tolower-demand patients who value a rapid and reliablerecovery and do not need sophisticated extensor func-tion for their daily activities. Appropriate patient coun-seling is critical to ensure that tendon transfer is seen asa treatment alternative rather than a failure in this set-ting.

The limitations of decision analysis involve themethods by which probabilities and utilities were ob-tained. In general, the most accurate, stable, and robustestimates of outcome probabilities are derived from ameta-analytic synthesis of randomized clinical trials,the highest level of clinical evidence. In this analysis,we obtained outcome probabilities from a quantitativereview of the literature. No randomized trials or evencomparative studies were available for analysis. Somestudies were more than 40 years old and probably donot reflect modern management strategies.

Determination of outcome utility also has limita-tions. Utility is a subjective value that an individual

places on the specific outcome. This can be very diffi-

JHS �Vol A, July

cult to operationalize and to quantify. In general, themost robust estimates of patient-derived utilities arederived from complex qualitative methods, such as thestandard reference gamble or time trade-offs, in whichpatients are asked to gamble or choose between healthstates usually referenced to death. In this study, wedetermined patient-derived utility values from a directscaling method using a Likert scale because of thecomplexity of the standard reference gamble and thedifficulty of referencing to death for humerus fracturewith radial nerve palsy. Although use of a directscale for the determination of utility values is lessrigorous than the standard reference gamble, thistechnique has been corroborated methodologically22

and has distinct advantages in terms of feasibility andreliability.23 Adult escorts of patients attending anoutpatient clinic are a sample of convenience that isprobably a good representation of the general public,but this is debatable. Although specific scenarioswere posed, the individuals completing the question-naires may not have had deep insight into the sub-tleties and implications of the various disease statesand treatments. Recognizing this, we also distributedthe utility survey to 20 orthopedic residents. Therewere no significant differences (p � .05) between theresponses of the residents and those of the subjects,suggesting that our scenarios provided adequate in-sight into the various described clinical conditions.Regardless of how outcome probabilities and utilitiesare determined, they are considered point estimatesin decision analysis, and sensitivity analyses aretherefore performed to assess how decision makingchanges over a range of these values, and our ap-proach is reasonable.

The results of this study may help us to better un-derstand the decision between initial observation andearly surgical exploration that is faced by a patient witha humerus fracture and associated radial nerve palsy.

For a given individual patient, the optimal strategydepends not only on the probabilities of the variousoutcomes but also on personal preference. Thus, theremay not be one right answer for all patients. Risk-takingpatients with a high utility for successful early surgerywould likely choose early exploration, whereas risk-averse patients with a lower utility even for successfulsurgery would not. Some low-demand patients whovalue quick and reliable recovery at the expense ofsophisticated hand function may even favor tendontransfer when there is no recovery after a period ofinitial observation. We therefore advocate a model ofshared decision making in which the physician and the

patient are jointly involved in the medical decision-

–August

996 HUMERUS FRACTURE AND RADIAL NERVE DECISION ANALYSIS

making process, considering both outcome probabilitiesand patient preferences.

Ongoing research efforts should focus on methods ofmaking an early distinction between nerves that willrecover spontaneously and those that will not in thisclinical scenario.

REFERENCES1. Shao YC, Harwood P, Grotz MR, Limb D, Giannoudis PV. Radial

nerve palsy associated with fractures of the shaft of the humerus:a systematic review. J Bone Joint Surg 2005;87B:1647–1652.

2. DeFranco MJ, Lawton JN. Radial nerve injuries associated withhumeral fractures. J Hand Surg 2006;31A:655–663.

3. Holstein A, Lewis GM. Fractures of the humerus with radial-nerveparalysis. J Bone Joint Surg 1963;45A:1382–1388.

4. Packer JW, Foster RR, Garcia A, Grantham SA. The humeral frac-ture with radial nerve palsy: is exploration warranted? Clin OrthopRelat Res 1972;88:34–38.

5. Dabezies EJ, Banta CJ II, Murphy CP, d’Ambrosia RD. Plate fixa-tion of the humeral shaft for acute fractures, with and without radialnerve injuries. J Orthop Trauma 1992;6:10–13.

6. Kettelkamp DB, Alexander H. Clinical review of radial nerve injury.J Trauma 1967;7:424–432.

7. Shaw JL, Sakellarides H. Radial-nerve paralysis associated withfractures of the humerus. A review of forty-five cases. J Bone JointSurg 1967;49A:899–902.

8. Pollock FH, Drake D, Bovill EG, Day L, Trafton PG. Treatment ofradial neuropathy associated with fractures of the humerus. J BoneJoint Surg 1981;63A:239–243.

9. Shah JJ, Bhatti NA. Radial nerve paralysis associated with fracturesof the humerus. A review of 62 cases. Clin Orthop Relat Res

1983;172:171–176.

10. Sonneveld GJ, Patka P, van Mourik JC, Broere G. Treatment of

JHS �Vol A, July

fractures of the shaft of the humerus accompanied by paralysis of theradial nerve. Injury 1987;18:404–406.

11. Samardzic M, Grujicic D, Milinkovic ZB. Radial nerve lesions associ-ated with fractures of the humeral shaft. Injury 1990;21:220–222.

12. Amillo S, Barrios RH, Martinez-Peric R, Losada JI. Surgical treat-ment of the radial nerve lesions associated with fractures of thehumerus. J Orthop Trauma 1993;7:211–215.

13. Larsen LB, Barfred T. Radial nerve palsy after simple fracture of thehumerus. Scand J Plast Reconstr Surg Hand Surg 2000;34:363–366.

14. Bernstein J. Decision analysis. J Bone Joint Surg 1997;79A:1404–1414.15. Kucey DS. Decision analysis for the surgeon. World J Surg 1999;

23:1227–1231.16. Kocher MS, Henley MB. It is money that matters: decision analysis

and cost-effectiveness analysis. Clin Orthop Relat Res 2003;413:106–116.

17. Foster RJ, Swiontkowski MF, Bach AW, Sack JT. Radial nervepalsy caused by open humeral shaft fractures. J Hand Surg 1993;18A:121–124.

18. Ring D, Chin K, Jupiter JB. Radial nerve palsy associated with high-energy humeral shaft fractures. J Hand Surg 2004;29A:144–147.

19. Raskin KB, Wilgis EF. Flexor carpi ulnaris transfer for radial nervepalsy: functional testing of long-term results. J Hand Surg 1995;20A:737–742.

20. Gousheh J, Arasteh E. Transfer of a single flexor carpi ulnaris tendonfor treatment of radial nerve palsy. J Hand Surg 2006;31B:542–546.

21. Ropars M, Dreano T, Siret P, Belot N, Langlais F. Long-term resultsof tendon transfers in radial and posterior interosseous nerve paral-ysis. J Hand Surg 2006;31B:502–506.

22. Stiggelbout AM, Eijkemans MJ, Kiebert GM, Kievit J, Leer JW, DeHaes HJ. The ‘utility’ of the visual analog scale in medical decisionmaking and technology assessment. Is it an alternative to the timetrade-off? Int J Technol Assess Health Care 1996;12:291–298.

23. Parkin D, Devlin N. Is there a case for using visual analogue scale

valuations in cost-utility analysis? Healthy Econ 2006:15:653–664.

–August

HUMERUS FRACTURE AND RADIAL NERVE DECISION ANALYSIS 996.e1

APPENDIXPatient Utility Questionnaire: Humerus Fracture/Radial Nerve Injury

Date: __________ Age: ____ Gender (M/F): ____With your help, this survey will define the relative

desirability of various results of treatment of afracture of the upper arm (broken humerus bone)combined with an injury to the radial nerve. For thepurposes of this survey, you should imagine that youhave this injury.

The humerus bone usually heals in about 2 to 3months without surgery and also does well if surgeryis indicated. For the purposes of this survey, pleaseassume that the fracture heals and causes you noproblems.

The radial nerve controls the muscles that openyour hand and lift your wrist. If your radial nervewas injured, you would develop a “wrist drop,”meaning that you would not be able to lift your wristor open your hand. There would also be a numb areaon the back of your hand and wrist, but yourfingertips would all feel normal.

The nerve usually recovers on its own over 6 to 12months, but sometimes surgery is needed. Surgeryleaves a long scar on your arm, and there is a smallrisk of infection, medical problems from anesthesia,and additional nerve injury. Two types of surgery arepossible:1. Nerve surgery. The surgeon looks at the nerve to

see if it is damaged. When the surgeon finds nervedamage, it is repaired by taking another nervefrom your lower leg, which leaves you perma-nently numb on the side of your foot. You canwalk and play sports fine, it just feels a littlestrange. The radial nerve usually takes a year ormore to recover after this surgery.

2. Tendon transfer surgery. Surgery can make yourhand and wrist open again by moving some ofyour arm muscles to different positions. While thehand will open and close after this surgery, itwon’t allow you to play piano and do other so-phisticated hand functions. Typing will be possi-ble, but slower.For each possibility/scenario, please choose a

number on the scale from 0, representing the worst

JHS �Vol A, July

possible result or outcome of treatment of the nerveinjury, to 10, representing the best possible result oftreatment of the nerve injury. Please make all of yourratings with respect to the best and worst possibleresults as we have defined them in items 1 and 5.1. Your nerve recovers completely on its own

within 6 to 12 months of the injury, without anytreatment or surgery.

0 1 2 3 4 5 6 7 8 9 10

worst best

2. Four weeks after the injury, the nerve has notrecovered. You have nerve surgery. The nerverecovers over the next year.

0 1 2 3 4 5 6 7 8 9 10

worst best

3. Four weeks after the injury, the nerve has notrecovered. You have nerve surgery. Over ayear later there is no recovery and you have asecond, tendon transfer surgery.

0 1 2 3 4 5 6 7 8 9 10

worst best

4. Six months after the injury, the nerve has notrecovered. You have nerve surgery. The nerverecovers over the next year.

0 1 2 3 4 5 6 7 8 9 10

worst best

5. Six months after the injury, the nerve has notrecovered. You have nerve surgery. Over ayear later there is still no nerve recovery andyou have a second, tendon transfer surgery.

0 1 2 3 4 5 6 7 8 9 10

worst best

–August