Objective. In this post hoc observational study, we investigated psychological predictors of outcome after radiofrequency and injection treatments, commonly performed in the management of chronic low back pain (CLBP).
Design & Setting. Data, comprising 161 patients (29 eventually lost to follow-up), were obtained from two randomized controlled trials on efficacy of radiofrequency treatment for back pain and sciatica. Subsequently patients were additionally treated in an open prospective follow-up period. Although all groups presented a significant visual analog scale reduction after 3 and 12 months, no additional pain relief after radiofrequency compared with injection treatment was found. Both trial populations showed sufficient similarities. A principal component (factor) analysis was performed on baseline psychometric tests, SF-36, and physical activity variables. We constructed five clinically relevant psychological profiles: “psychologically negative,” “adaptive manager,” “rigid qualities,” “supporting partner,” and “strong ego.” These were examined as possible predictors of significant pain relief using logistic regression analysis.
Results. The “psychologically negative” dimension showed a negative and the “adaptive manager” dimension a positive prognostic effect on outcome.
Conclusions. Minimally invasive treatment for CLBP leads to significant pain reduction, including potential placebo effects. However, psychologically vulnerable patients, characterized by, among others, reduced life control, disturbed mood, negative self-efficacy, catastrophizing, high anxiety levels, inadequacy, and poor mental health, tend not to respond to this treatment. Patients characterized by a.o. reduced pain and interference levels, positive expectations, and reasonable physical and social functioning, react more favorably. From both a clinical and a financial perspective, psychosocial evaluation and selection of patients seems appropriate, before applying minimally invasive procedures for CLBP.
Low Back Pain
In the past two decades, radiofrequency (RF) treatment procedures have been added to already employed steroid injection therapies in the minimally invasive treatment arsenal for chronic low back pain (CLBP) . Based on clinical impressions of effectiveness, these procedures are applied on a large scale in pain clinics in The Netherlands and other Western countries . Few controlled outcome studies have appeared on the efficacy of RF procedures in CLBP syndromes [3–6], including two systematic reviews [7,8]. Minimally invasive treatments for CLBP appear to significantly contribute to the rising costs associated with treatment of low back pain .
Recently two multicenter randomized controlled trials were finished regarding the efficacy of RF procedures in the treatment of CLBP. One trial evaluated the efficacy of lumbar RF facet denervation compared with sham treatment in patients with predominantly back pain . The other trial assessed the efficacy of lumbosacral RF treatment of the dorsal root ganglion vs sham treatment in patients with predominantly leg pain . Both studies reported no statistically or clinically significant differences after 3 months between RF and sham lesion outcomes on visual analog scale (VAS) scores, physical activities, and analgesics intake. However, in each trial, a comparable statistically significant VAS reduction was found in both RF and sham treatment groups. Furthermore, in the same trials in a post hoc inspection of the data, it was suggested that psychologically resilient patients would respond better to (any) treatment. This finding is supported by an earlier report on classification of pain patients based on combinations of physical, psychosocial, and behavioral measures . It should be noted that the sham (control) treatments used in these trials consisted of an injection with local anesthetic plus sham RF treatment.
Psychosocial issues play an important role in chronic pain [13–15]. Psychological predictors of disability and chronicity in patients with low back pain have been identified [16,17]. Recent systematic reviews reported on predictors of outcome after multidisciplinary rehabilitation programs and after lumbar disc surgery [18,19]. Psychological predictors for efficacy of laminectomy and/or spinal fusion procedures for CLBP and sciatica were identified [20–22]. Several studies on psychological prognostic factors of outcome after minimally invasive procedures like chemonucleolysis and spinal cord stimulation are available [23–27], but only one after (cervical) RF treatment . In nearly all above-mentioned studies, it was concluded that psychologically vulnerable patients tend not to respond favorably to treatment, while patients with reasonable physical functioning and adequate coping have a better prognosis.
The present study was designed as a post hoc observational analysis to evaluate the predictive value of the patient psychological profile on outcome after 3 and 12 months after RF and injection treatment procedures for the management of CLBP.
For this study, patient data were derived from two above-mentioned multicenter randomized controlled trials [10,11]. These were conducted at the pain clinics of four medical centers in The Netherlands (i.e., University Medical Center, Utrecht; Rijnstate Hospital, Arnhem; Juliana Hospital, Apeldoorn; Twenteborgh Hospital, Almelo). The medical ethical committees of these centers approved both studies. All patients provided a written informed consent form. Both trials were specifically designed to reflect common clinical practice and were identical in study design and standardized intake, except for the localization of the pain. The first trial included patients with predominantly back pain, and the second trial with predominantly leg pain. Exclusion criteria were actual indication for low back surgery, age less than 18 years, prior RF treatment, coagulopathy, allergies for radiopaque contrast dye or local anesthetics, malignancy, mental handicap or psychiatric condition precluding adequate communication, language problems, and pregnancy. Furthermore, in the back pain group, diagnostic intra-articular facet joint blocks were required to yield at least 50% pain reduction on a standard VAS, whereas in the leg pain group, selective spinal nerve blocks were required to yield at least 75% pain reduction on a standard VAS. Eventually 81 patients participated in the back pain trial and 80 in the leg pain trial. In both trials, sham (control) treatment was performed identical to the RF procedures, but no RF current was applied. All patients received local anesthetic infiltration of the target structures prior to RF or sham lesioning. In this manner, the double blinding was adequately ensured. Results were determined 3 months after the initial RF or control treatment. In both trials, the vast majority of patients failed to respond according to the primary combined outcome measure, comprising VAS, physical activity scores, and analgesics scores. However, in each trial, a comparable statistically significant VAS reduction was found in both RF and control lesion groups. It was concluded that in these selected groups of patients, both RF and sham (local anesthetic injection) treatment should be regarded as about equally effective. The patients in both trials were subsequently followed and treated as found necessary on clinical grounds in an open prospective follow-up period of 9 months. Additional treatment consisted of a variable combination of RF and/or steroid injection treatments (Table 1). The applied diagnostic and RF treatment procedures were consistent with the Dutch guidelines for pain management . During the study period patients did not receive any surgical interventions, nor was psychological or behavioral therapy instated.
2–4 ml bupivacain 0.25%+ 40 mg triamcinolonacetonid
Therapeutic epidural injection
Mechanical and/or radicular pain
1; 22G/10 cm
2–4 ml bupivacain 0.25%+ 40 mg triamcinolonacetonid
CLBP = chronic low back pain; RF = radiofrequency.
Before initial treatment and at 3 and 12 months after this procedure, median values of maximum pain intensity were determined. Median values were calculated from four successive scores based on diaries that were completed twice weekly by the patients during 2 weeks before, and during 12 months following the treatment. The pain diary consisted of two standard VASs (both recording maximum pain levels during the past 24 hours), specified for low back pain (VAS-back) and for radiating pain in the leg (VAS-leg). A number of psychometric tests were performed before initial treatment. The Dutch version of the Multidimensional Pain Inventory (MPI-DLV) was administered to register pain-related psychosocial and behavioral aspects [30,31]. Cognitive dimensions were registered with the Pain Cognition List (PCL) . Anxiety was measured with the Dutch version of the Spielberger State-Trait Anxiety Inventory (STAI-DV) . The Dutch Personality Questionnaire (DPQ) was applied to record personality aspects . Quality of life was measured at baseline with the SF-36 questionnaire . Daily activities were measured at baseline using a physical activities scale (derived from the Dutch Central Bureau of Statistics), in which the patient answered on 10 specific activities of increasing effort, if they could be performed “without difficulty,” “with difficulty,” “with help from others,” or “not possible.” The scores were respectively 3, 2, 1, or 0 points. Maximum score was 30 points.
In order to identify predictive factors for substantial back pain and/or leg pain reduction, patients were classified for VAS-back reduction ≥ 50% and VAS-leg reduction ≥ 50% compared with baseline values after 3 and 12 months. Outcomes were controlled for initial randomization, VAS at baseline, and success at 3 months in the analysis at 12 months.
Baseline demographic characteristics, psychosocial qualities, and mean values of the psychometric tests, quality of life, and daily activities variables were used to compare initial groups with primarily back pain and primarily leg pain. Both groups showed similar patient characteristics and baseline values of outcome parameters (Table 2), and psychosocial qualities (Table 3). Only for “MPI––pain intensity” (P = 0.0056) and “DPQ––conceitedness” (P = 0.0467) statistically significant differences were found between the back pain and leg pain group. It was concluded that there was sufficient matching between the groups to justify a merging for further analyses.
Factor analysis was performed on the measured psychosocial variables to reduce the number of independent variables, in order to create a limited number of prominent recognizable psychological profiles. A principal components analysis with varimax rotation was applied to the 36 × 36 interitem correlation matrix of the assimilated items of the five questionnaires: the MPI-DLV, the PCL, the STAI-DLV, the DPQ, and the SF-36, and of the physical activities scale. Principal-components analysis was chosen as an appropriate method for identification of the minimum number of factors needed to account for the maximum amount of variance in data . Different models regarding the percentage of VAS reduction were explored. A VAS reduction of 50% as definition of “success” turned out to produce a more stable and clinically relevant result compared with a VAS reduction of 25%. Factors were produced based on Eigen values greater than 1.0, Scree plots , and a clinically relevant interpretation of factors. The minimum loading of a variable for insertion in a factor was 0.45 (positive or negative), with a minimum intervariable difference of the loading of 0.15 (Table 4). Thus, five psychological profiles were constructed: “psychologically negative” (PSYNEG), “adaptive manager” (ADAPT), “inflexible qualities” (RIGID), “presence of a supporting partner” (SUPPART), and “strong ego” (STREGO) (Table 5). These five profiles explained a sufficient part (56.9%) of the variance in the original data pool. The internal consistencies of the scales were measured with Cronbach's alpha and were respectively 0.71, 0.60, 0.77, 0.83, and 0.63. Factor loadings were used to compute the scale scores. Sensitivity (value to predict a positive response to treatment) of this model is moderate (0.49), and specificity (value to predict a negative response) is high (0.93). Furthermore, in order to control for adequate matching between the original back pain and leg pain groups, separate factor analyses of both groups were performed and factor structures compared. In both groups, the same psychological profiles could be constructed.
Qualities of the five psychosocial profiles as determined by factor analysis
“psychologically negative” High interference, reduced life control, disturbed mood, negative self-efficacy, catastrophizing, high state and trait anxiety, inadequacy, low self-esteem, poor physical and mental health, reduced vitality
Low pain intensity, low interference, reasonable physical activities level, positive expectations, high physical and social functioning
Presence of a supporting, rewarding, and distracting partner
High self-esteem, dominance, high social adequacy
The relations between the compound factors (i.e., the psychological profiles), baseline VAS scores, and substantial VAS reduction at 3 and 12 months were determined with logistic regression analysis. Outliers were identified with Cook's distance determination, and subsequently discarded . Outcome in VAS reduction was dichotomized (at ≥50%) to obtain more stable model outcomes. For all tests and analyses, the Statistical Package for the Social Sciences (SPSS) was used. P-values of ≤0.05 were defined as statistically significant. P-values between 0.05 and 0.10 are described as trends in text and tables.
For short-term outcome analyses, data on all 161 patients could be obtained. Between 3 and 12 months, 29 patients (15 back pain group; 14 leg pain group) were lost to follow-up, due to premature termination of control visits by the patients or because of incomplete data. During these 9 months, 21.1% of patients received no additional treatment, 40.4% received additional RF treatment, 8.1% received additional steroid injection treatment, and 30.4% received both types of additional treatment. The number of additional treatments varied between 0 and 6. No further assessment of the consequences of the types and numbers of additional treatments on outcome was performed, as the small group sizes would render adequate analysis impossible. Also, during this period no significant analgesic medication changes were made, and patients did not enter rehabilitation programs.
Back Pain Group
In this group, logistic regression analyses between the five psychological profiles constructed by the factor analysis and VAS revealed a significant positive prognostic effect of ADAPT on the outcome at 3 months, whereby success was defined as a reduction ≥50% of both VAS-back and VAS-leg reduction. Long-term outcome, in which both VAS-back and VAS-leg were reduced ≥50%, showed a negative predictive influence of the factor PSYNEG and positive influences of ADAPT and the presence of substantial back and leg pain reduction at 3 months. Separate VAS-back reduction ≥50% at 12 months was predicted by the same factors, except for the 3-month VAS-leg value. Separate VAS-leg reduction ≥50% was positively predicted by ADAPT and the presence of a substantial VAS-leg reduction at 3 months. With this outcome the factor RIGID showed a negative influence (Table 6).
Results of logistic regression analysis of the predictive value of psychosocial profiles in patients with primarily back pain, in patients with leg pain, and in all patients together, controlled for success at 3 months, on the outcome at 12 months
↵¶ Number of outliers determined by Cook's distance.
Only statistically significant values and trends are mentioned.
Leg Pain Group
The analysis at 3 months was found unstable with five clear outliers. Although statistically significant positive effects were found for ADAPT, and a negative influence for PSYNEG, these findings should be interpreted with care. Long-term outcome, in which both VAS-leg and VAS-back were reduced ≥50%, showed a negative prediction by the factor PSYNEG and a positive predictive influence of the presence of a substantial VAS-leg reduction at 3 months. Separate VAS-leg reduction ≥50% at 12 months was predicted positively by ADAPT and the presence of a substantial VAS-leg reduction at 3 months, and negatively by the factor PSYNEG. Separate VAS-back reduction ≥50% at 12 months was negatively predicted by the factor PSYNEG and positively by the presence of a substantial VAS-back reduction at 3 months (Table 6).
Back Pain and Leg Pain Group Together
Long-term outcome in the complete group of patients, in which both VAS-leg and VAS-back were reduced ≥50%, was predicted negatively by PSYNEG and positively by ADAPT and the presence of already substantial back and leg pain reduction at 3 months (Table 6).
The trials from which we derived our CLBP patients' data reported statistically significant VAS reductions in all groups, without statistically or clinically significant differences between patients initially treated by RF vs injection treatment. This enabled us to merge the data of both patient populations in order to perform a more stable and powerful factor analysis. The psychosocial components for the factor analysis were based on validated questionnaires, comprising coping (MPI-DLV), cognitive (PCL), anxiety (STAI-DV), and personality and behavioral aspects of pain (DPQ). Quality-of-life aspects were registered with the internationally approved SF-36 inventory. Only the physical activities scale, based on the system used by the Dutch government Central Bureau of Statistics, has not been validated. These components together present a balanced overview of the patient psychosocial status. We used factor analysis as a means to reduce data, by which we would be able to devise a limited number of relevant psychosocial profiles . As the factor analysis was applied on two, seemingly separate, populations of patients, we executed a prior factor analysis of each group and compared the factors. As no major differences were found, we consider the factor analysis of the whole group as valid. We are aware of the fact that, in factor analyses, the sample size should ideally be about 10 times the number of measures to avoid instability of the factor structure. However, we did manage to come up with a stable factor structure, despite the fact that our sample size was 160 vs a number of measures of 36.
We managed to construct five recognizable and clinically useful profiles, as summarized in Table 5. These profiles are quite consistent with the three to four MPI-derived clusters that were described before: “adaptive coper,” “dysfunctional,” “interpersonally distressed,” and “average”[30,38,39]. Our findings also correlate reasonably well with the three distinct CLBP patient groups Strong identified: “patients who were in control,” “patients who were depressed and disabled,” and “patients who were active copers with high denial”. Both the psychologically negative profile (PSYNEG) and the adaptive manager profile (ADAPT) we identified appear to significantly predict long-term outcome after minimally invasive RF or injection treatments for CLBP. For the whole group (back pain and leg pain patients), PSYNEG has an overall negative predictive value. For the back pain group, this was only found true if VAS-back reduction ≥50% was involved (as the outcome variable). It disappeared when only VAS-leg reduction ≥50% was considered. These findings are generally in agreement with most reports on psychological predictors of outcome after specific interventional or noninterventional treatment for CLBP. For the whole group, ADAPT positively predicts outcome. This was also found for the back pain group separately. However, in the leg pain group it disappeared when VAS-back reduction ≥50% was involved. A positive predictive value of typical ADAPT qualities like physical functioning and coping has been reported previously [19,22]. The only previous report on predictors of efficacy of RF treatment , only identified catastrophizing as a prognostic factor of outcome after RF facet denervation and dorsal root ganglion treatment for chronic cervicobrachialgia. Catastrophizing is one of the qualities of the PSYNEG profile we constructed. Furthermore, the presence of an already substantial reduction (≥50%) in back and leg pain at 3 months showed a strong positive prediction of the presence of a substantial pain reduction in back and leg pain at 12 months. For both groups, separately and combined, applies, that success at 3 months is a strong predictor for success at 12 months of the same type of pain (leg and/or back pain.). Additional differences between both groups were found for PSYNEG and ADAPT. In the back pain group, presuming predominant back pain, both factors have their predictive influence. For the leg pain patients, presuming predominant leg pain, this is only valid when reduction in that type of pain is taken into account. When in these patients back pain is included, PSYNEG prevails. Thus, it seems that back pain and leg pain patients do differ in effect of predictive factors.
As no clinically significant additional VAS reduction occurred between 3 and 12 months, additional RF and/or steroid injection therapies for remaining back and/or leg pain did not clearly contribute to further VAS reduction. Whether the additional treatments consolidated the initial VAS reduction could not be distilled from our data, as no adequate control group was available during this study phase.
In the original trials, we had to face the challenge of including a control “placebo” treatment, that could not be distinguished from the initial effects of RF treatment. A local anesthetic injection treatment control group appeared to be the only logical option. This created the subsequent dilemma whether this control should be considered as placebo or active. We performed this analysis on the assumption that the control group was active; in fact as active as the RF group. In the literature there is substantial evidence to suggest that local anesthetic injection treatment can have effects far exceeding the pharmacological duration of effect [41–44].
On the other hand, one may regard the VAS reductions after both RF and injection treatment as not directly treatment related, but due to placebo effect, perhaps combined with the natural course of the disease and regression to the mean [45,46], especially in the light of the controversies still surrounding RF treatment. We are not in a position to differentiate between these two views. As clinicians, however, we do strongly feel that in chronic pain patients, even with scientifically proven active treatment modalities, outcome is codetermined by a placebo effect. It may be that pain, more than other conditions, stimulates placebo responsiveness .
The question also arises what the outcome would have been if a nonintervention control group would have been included. Without going into the complexities of designing such a trial, in which blinding obviously would be an impossibility, we hypothesize that the outcomes may not be as good as the ones we have seen in the original trials.
In whatever way one views the treatment effects observed in the original trials, as a resultant of either active treatment or placebo, our study clearly shows that psychological profiles can predict the positive or negative response of patients to these interventions. This is an important finding, as we may have touched upon a means of identifying placebo responders. This, however, will require further research far beyond the scope of this study.
Our findings become even more relevant if one considers recent data revealing that injection treatments account for a significant proportion of since the early 1990s almost exponentially rising CLBP-associated health care costs .
In conclusion, long-term substantial pain reduction (≥50% VAS reduction) in patients with CLBP and radiating pain to the leg, who were treated with minimally invasive blockades, is strongly and statistically significantly influenced by the psychosocial profiles as determined prior to the treatment. Psychologically vulnerable patients with a psychosocial profile characterized by high interference, reduced life control, disturbed mood, negative self-efficacy, catastrophizing, high anxiety levels, inadequacy, low self-esteem, poor physical and mental health, and reduced vitality (PSYNEG-type), tend not to respond to these treatments. On the other hand, patients characterized by reduced pain and interference levels, reasonable physical activities levels, positive expectations, and reasonable physical and social functioning (ADAPT-type), perform more favorably on these interventions. These findings are irrespective of whether one views the pain reduction following these procedures as placebo or due to active treatment. Therefore, from both a clinical and financial perspective, psychosocial evaluation and selection of patients with chronic low back or leg pain prior to applying minimally invasive techniques seems appropriate.
This study was supported by a grant (OG 95-027) from the Dutch Health Insurance Council and by a contribution from the Pain Expertise Center Nijmegen, The Netherlands. We are indebted to the patients that participated in the studies (TECL trial) and to the referring physicians. We gratefully acknowledge all contributors to the study: ET Kamphuis MD PhD, AJ de Gier MD, EJ Buijs MD, ASM Wassink MD, anesthesiologists; JMSP Quirijnen PhD, Michiel P. Mastenbroek, medical psychologists; DG Roseval MD, FDM Hommes MD, AJ van Nie MD, EJ Bijsmans MD, residents in anesthesiology; CGMM Beckers, coordination; the qualified nurse technicians assisting with the procedures; all secretaries from the pain clinics administrations; P Anema, information technologist; BMG Annink, BM den Bandt, trial bureau; and M. Oldenborg, data analyses. Professors AJP Schrijvers MD PhD and CJ Kalkman MD PhD are acknowledged for their general advice and support.
. A systematic review of sociodemographic, physical, and psychological predictors of multidisciplinary rehabilitation-or, back school treatment outcome in patients with chronic low back pain. Spine 2005;30:813–25.
Roelof M. A. W.Van Wijk, Jos W. M.Geurts, RichelLousberg, Herman J.Wynne, EdwinHammink, Johannes T. A.Knape, Gerbrand J.GroenPain Med(2008)9 (2):
212-221DOI: http://dx.doi.org/10.1111/j.1526-4637.2007.00367.xFirst published online: 1 March 2008 (10 pages)