Design
The Department of General Practice of Erasmus Medical Center, University Medical Center, Rotterdam, conducted a large prospective cohort study on knee complaints in primary care. A total of 40 GPs from five municipalities in the southwest region of the Netherlands asked patients with new knee complaints to participate in the general cohort study, with a follow-up of 1 year. This network represented a total patient population of around 84 000 patients. Detailed information about the study design has been published previously.22
How this fits in
Prognosis of persistent complaints after a knee injury is based on secondary care populations. In a primary care setting, however, no studies have addressed this issue. Being aged >40 years was found to be a predictor for persistent complaints 1 year after a knee injury, whereas physical examination and MRI findings showed no association for persistent complaints. At present, use of physical examination or MRI for prognosis of persistent complaints after knee injury is not recommended.
Based on this study, only history taking shows some prognostic value and the study cannot recommend physical examination or MRI examination for prognostic use.
Patients with knee complaints were informed about the general cohort study during the initial consult with the GP and were invited to participate. Those aged 18–65 years with a traumatic onset of knee complaints, and who visited their GP within 5 weeks after initial injury, were also asked to participate in an additional MRI study. After registration of the patient at the research centre of the Department of General Practice, an appointment was made, informed consent was acquired, and patients completed the baseline questionnaire If patients gave consent for the additional MRI study, an appointment for the MRI was scheduled; following this MRI, a physical examination was performed. Patients with MRI contraindications — such as pregnancy, metal implants, or a pacemaker — were excluded from this MRI study.
Data collection
At baseline, the participating patients completed a self-report questionnaire. The questionnaire collected data on age, sex, socioeconomic status, history of previous knee injuries and/or surgery, present symptoms, mechanism of injury, and the level of activity during work or sport. The therapeutic management initiated by the GP at baseline was also recorded.
The severity of pain was obtained using an 11-point numeric rating scale ranging from 0 (no pain) to 10 (unbearable pain).23 The Lysholm knee score was used to provide relevant information regarding pain, swelling, instability, and functional limitations such as walking and stair climbing.24 The Lysholm score ranged from 0 (worst possible score) to 100 (excellent score). The Tegner knee function score (range: 0–10) was used to determine the level of activity in work and sports prior to the knee injury; 0: on sick leave/disability,10: participation in competitive sports.24 The Tegner score was obtained with the ‘level of work’ and ‘sports activities’ determinants from the questionnaires. In the final questionnaire, at 1 year after baseline, patients were asked to report their perceived recovery using a seven-item Likert scale categorised as ‘full recovery’, ‘major improvement’, ‘minor improvement’, ‘about the same’, ‘minor deterioration’, ‘major deterioration’, or ‘worse than ever’.22,25
Persistent complaints were defined as those in which, 1 year after the initial knee injury, patients reported ‘minor improvement’, ‘about the same’, ‘minor deterioration’, ‘major deterioration’, or ‘worse than ever’. Patients’ reports of either ‘full recovery’ or ‘major improvement’ were considered as signifying a clinically relevant recovery.
The MRI was scheduled 2–6 weeks after the initial trauma and used a 1.0 Tesla General Electric device. The results of the MRI were determined by two radiologists independently, based on a standardised classification form. In cases of disagreement, the findings were discussed until consensus was reached. Meniscal tears were classified as horizontal, longitudinal, radial, or complex. Anterior cruciate ligament, posterior cruciate ligament, medial cruciate ligament, and lateral collateral ligament lesions were classified as either partial or complete. At the time of data collection for this study there was no validated scoring system for osteoarthritic features seen on MRI, so the study adapted the most frequently used method — the Kellgren and Lawrence method26 — to score the severity of osteoarthritis on X-rays. Detailed information about the MRI procedure is reported elsewhere.27
A physical examination was performed immediately after the MRI, according to a standardised protocol.22 A trained physical therapist with ≥15 years’ experience in performing physical examination in patients with knee injuries and with ≥10 years’ experience in diagnostic research performed the examination.28 Physical examination of both knees consisted of inspection of alignment and joint effusion, palpation of temperature, collateral ligaments and joint line tenderness, assessment of effusion, and passive range of motion in flexion and extension.29,30 Cruciate and collateral ligament integrity were assessed by means of the anterior and posterior drawer tests,31 the Lachman test,32 the pivot shift,33 and the valgus and varus stress tests.34
To avoid influencing the behaviour of the patient or the management by the GP according to clinical guidelines for traumatic knee complaints,35 the patients and their GPs were not informed about the findings of the MRI or physical examination during the 1-year follow-up.
Statistical analysis
Descriptive statistics using SPSS (version 15.0) were used to present the baseline characteristics of the study population and the findings on the MRI. Binary logistic regression analysis was used to determine the association of isolated determinants from history taking, physical examination, and MRI with the presence of persistent knee complaints after 1 year; these were expressed as odds ratios (ORs) with a 95% CI. To assess the independent prognostic value of related determinants with persistent complaints, multivariate backward logistic regression analysis (P entry 0.10, P removal 0.20) was used, using determinants with an isolated association (P<0.10) to eliminate redundant variables. With the results of this multivariate analysis, a prognostic model was built and the area under the curve (AUC) of the receiver operating characteristic (ROC) curve was calculated. Complete case analysis was used.
To determine the additive predictive value of MRI for the prognosis of persistent complaints, the associated determinants from MRI (P<0.10) were added in the prognostic model using the Enter method. Again, the AUC was calculated. The additive predictive value of MRI was determined by the difference in the AUC in both models.
