J Neurol Neurosurg Psychiatry: first published as 10.1136/jnnp.68.1.75 on 1 January 2000. Downloaded from J Neurol Neurosurg Psychiatry 2000;68:75–79 75 When does the patient with a disc herniation undergo lumbosacral discectomy? PatrickCAJVroomen, MarcCTFMdeKrom, J Andre Knottnerus Abstract and treating physician. Finally, the complaints Objectives—To design a model that could of nerve root compression should have existed accurately predict eventual lumbar disc forasuYcient time period before surgery is surgery in the patient initially presenting considered. What is a suYcient time? In view of with clinical findings of nerve root com- the favourable natural course of clinical nerve pression. root compression,17–20 and the possible compli- Methods—Prospective study in 183 pa- cations of surgery, most physicians will prob- tients with nerve root compression sam- ably prefer to continue conservative treatment pled from a primary care population. All for at least 6 weeks. If there were a way of pre- patients underwent a standardised his- dicting an eventual poor outcome of conserva- tory, physical examination, and MRI. Sur- tive therapy with the subsequent need for sur- gery carried out in the next 6 months was gery at the time that the patient initially recorded. Models were constructed to presents, time intervals until surgery might be predict whether patients eventually re- shortened in a subset of patients. ceived surgery. In a prospective study, patients with clinical Results—Two models were constructed. findings indicating nerve root compression Reduced model A was based on baseline were investigated with regard to findings that findings, only, and contained six variables. might predict the future need for operation. Model B incorporated change over time as This prediction was evaluated by logistic well and contained 10 variables. The area regression at two stages: (1) how could future under the curve (in a receiver operating surgery be predicted at the first consultation characteristic) for these models was 0.86 (model A), and (2) how could it be predicted and 0.92, respectively. It was shown that at at a second consultation after 2 weeks taking a probability cut oV of 0.60, model B pre- into account the disease course over 2 weeks dicted eventual surgery with a sensitivity (model B). As the benefits of these predictions of 57% and a specificity of 100%. might expediate MRI studies and eventual Conclusions—Given the requirement of a surgery, we assumed that the model used to high specificity, eventual operation could expediate MRI studies should have a specifi- be adequately predicted in a sample of 183 city exceeding 95% (limiting the rate of MRI patients with clinical nerve root compres- studies done in patients that would eventually sion. The application of the model in not be operated to 5%) and the model used to patients with nerve root compression expediate surgery should have a specificity of http://jnnp.bmj.com/ might lead to earlier operation in a subset nearly 100% (so that no patients would be of patients resulting in a reduction of operated on as a result of the model’s predici- duration of illness and associated indirect tion that would otherwise not be operated on). costs. An important prerequisite for fu- The value of the predictive models was ture application would be the validation of estimated to indicate whether future validation the prediction rule in another population. of the models in additional research would be Department of (J Neurol Neurosurg Psychiatry 2000;68:75–79) useful. Neurology, Maastricht on September 24, 2021 by guest. Protected copyright. University Hospital, Keywords: nerve root compression; disc displacement; Maastricht, The surgical therapy; discectomy; prediction Patients and methods Netherlands STUDY POPULATION PCAJVroomen Between February 1995 and December 1996, MCTFMdeKrom Several papers support the eVectiveness of 50 general practitioners (GPs) in Maastricht lumbosacral discectomy in patients with clini- and surrounding villages asked patients to par- Department of 1–4 General Practice, cal findings of nerve root compression. ticipate in a study on sciatica. Patients were Maastricht University, Those clinical findings as highlighted in referred to the neurology department of the Maastricht, The various papers5–8 should then be supported by Maastricht University Hospital if they pre- Netherlands radiological images of nerve root com- sented for the first time with the current J A Knottnerus pression.9–15 The issue raised by these papers is episode of pain radiating into the leg below the Correspondence to: how high the likelihood of clinical and gluteal fold of suYcient intensity. They were Dr Patrick C A J Vroomen, radiological nerve root compression should be excluded for previous spinal surgery, preg- Maastricht University to warrant operation.16 The optimal likelihood nancy, confinement in prison, unavailability for Hospital, P Debyelaan 25, PO Box 5800, 6202 AZ depends on the situation and the patients’ and follow up visits (moving plans, severe disease), Maastricht, The Netherlands doctors’ preferences. Even patients with near and severe comorbidity. The researchers re- certain nerve root compression are not auto- viewed the selection criteria and applied two Received 8 September 1998 matically operated on, however. The nerve root further criteria: patients were to have symp- and in final form 18 June 1999 compression should be suYciently bothersome toms of and signs of lumbar nerve root involve- Accepted 28 June 1999 for surgery to be considered by both patient ment (at least two of the following signs and J Neurol Neurosurg Psychiatry: first published as 10.1136/jnnp.68.1.75 on 1 January 2000. Downloaded from 76 Vroomen, de Krom, Knottnerus symptoms: typically radicular pain distribu- STAGE 1: HISTORY/QUESTIONNAIRE ITEMS tion, pain in the leg increasing with coughing, These were age, sex, high education level, liv- sneezing, or straining, decreased muscle ing alone, mentally demanding job type, previ- strength, sensory loss, reflex loss, positive nerve ous episodes of pain in the leg, previous tension test). They were excluded if there was episodes of low back pain, family history, an indication for direct surgical intervention comorbidity, smoking, sports activities, exer- (either morphine dependent, intractable pain, cised back or abdominal muscles, more pain in progressive, severe decreased paresis of limited leg than back, sudden onset of pain, cause of duration, or cauda equina syndrome). pain known, pain worse in the evening or night, paroxysmal pain, pain already improv- BASELINE EXAMINATIONS ing, typically dermatomal pain distribution, All patients selected were examined within 2 more pain on coughing, sneezing, or straining, days of referral by the GPs and again 2 weeks more pain on sitting, numbness in the leg, later. At stage 1, two classes of independent paraesthesias in the leg, cold feeling in the leg, variables were investigated for their predictive worrying about health, intensity of pain in the value: leg scored by the patient on a visual analogue scale (VAS), complaints considered severe by 1.0 the physician, pain rating index of McGill pain questionnaire21 for sensory and aVective di- mensions (PRI-S and PRI-A), questionnaire items diYculty putting on socks/stockings, and 0.8 pain while dressing. STAGE 2: PHYSICAL EXAMINATION ITEMS These were finger-floor distance, paresis, hyp- aesthesia, hypalgesia, ankle tendon reflex dif- 0.6 ference, knee tendon reflex diVerence, straight leg raising (SLR; the test was positive if there was typically dermatomal pain on straight leg raising at an angle less than 45o), reversed SLR, Sensitivity 0.4 crossed straight leg raising, Kemp sign, NaVziger sign. Prediction after Additionally, it was investigated how much 2 week consultation predictive value would be gained by re- Prediction after evaluating the patient after 2 weeks, thereby 0.2 first consultation taking the time course into account. The second stage analysis incorporated results of the 2 week follow up visit (instead of baseline values) for typically dermatomal pain distribu- 0.0 tion, pain worse on sitting, SLR, and crossed 0.0 0.2 0.4 0.6 0.8 1.0 SLR. Also, for pain intensity in the leg the 1 – Specificity change of VAS over 2 weeks was assessed http://jnnp.bmj.com/ Figure 1 ROC Curves for the prediction of eventual surgery by model A and B. instead of the VAS at baseline. The PRI-A of ROC=receiver operating characteristic. the McGill pain questionnaire after 2 weeks was investigated instead of the PRI-A at 183 Sciatica patients baseline. STATISTICAL ANALYSIS Backward stepwise logistic regression analysis22 with the likelihood ratio test on SPSS software on September 24, 2021 by guest. Protected copyright. Positive Model B Negative was used to predict surgical intervention in the 18 patients 165 Patients next 6 months. At the first stage, set I was modelled. Then, set II was modelled also incorporating the predictive variables from the MRI model of set I. The resulting reduced models were based on data collected at first consulta- tion and predicted the likelihood of surgery in MRI positive MRI negative 132 Patients 13 Patients 17 Patients 1 patient treated treated surgically the next 6 months. The probability of surgery treated surgically conservatively eventually for each patient can be estimated using the for- in week 3 mula below table 2. At any given probability cut oV, some patients will be below the cut oV and some above. These numbers can be 133 Patients treated conservatively compared with the numbers actually operated ina2×2 table. In this way, sensitivity and spe- cificity values can be calculated for any 30 Patients treated probability cut oV value. The (1-specificity) surgically versus sensitivity values for the various prob- 17 Early/13 Eventually ability cut oVs of both models are depicted in the receiver operating characteristic23 24 in Figure 2 An example of model B application. figure 1. The area under the curve was J Neurol Neurosurg Psychiatry: first published as 10.1136/jnnp.68.1.75 on 1 January 2000.