Superior outcomes following cervical fusion vs. multimodal rehabilitation in a subgroup of randomized Whiplash-Associated-Disorders (WAD) patients indicating somatic pain origin-Comparison of outcome assessments made by four examiners from different disciplines

2.5.1 Examination by the neurologist (ÅS)

The neurological examination was performed according to general standards at neurological university departments and included the following: The functions of cranial nerves I–XII were examined with the exception of visual acuity, hearing threshold, gag reflex and swallowing. The cutaneous and tendon reflexes were tested with the exception of anal and genital reflexes. The cerebellar functions were examined with regard to extremity co-ordination, balance, articulation and eye movement. The power and tone of extremity-related muscle groups were tested and the possible presence of abnormal motor activity (dyskinesia) was investigated. Cutaneous sensibility to light touch and pinprick as well as deep sensibility to vibration were tested in the extremities. Sensibility disturbances and/or regional pain were documented graphically. The neck was examined with regard to active and passive mobility as well as tenderness at palpation.

The symptoms and findings were mainly referable to the cervical spine and adjoining regions such as shoulders and arms, extremity motor functions and sensibility (regional/numbness/paraesthesia). The global scale categories of change were conditionally based on these findings accordingly: much worse (deterioration by >2 grades), somewhat worse (deterioration by 1–2 grades), unchanged, somewhat better (improvement by 1–2 grades), and much better (improvement by >2 grades).

2.5.2 Examination by the orthopedic surgeon (IG)

The orthopedic examination included ocular inspection of the carriage of the head, position of the shoulders, especially the position of the scapulae, hyper- or hypotrophy of the muscles of the neck, shoulder, arm and hand. Spontaneous as well as forced movements of the head, shoulders and upper extremities were observed. Palpatory examination included the neck muscles, especially their insertions, and pain recordings during passive, active and forced movements. Finally, neurological testing included motor, sensory and reflex examination. The number of indicators of changed status referring to the dimensions: cranial nerves, cervical spine, cerebellar, motor and sensory functions determined the global scale of change.

2.5.3 Examination by the physiatrist (PS)

Evaluation in physical medicine and rehabilitation is normally based on the dimensions of the International Classification of Function [18], [19], [20]. In our study design the examination by the physiatrist should mainly rely on the dimension impairment without taking function, participation or the environmental and personal factors into account. The assessments comprised an ordinary clinical examination in physical medicine and rehabilitation, with the neurological part omitted due to the study design.

The clinical examination included measurement of range of neck movement with an inclinometer, flexion, extension, bilateral rotation, lateral flexion, provocation of pain through passive stretching of the neck muscles during the above-mentioned movements, the bilateral Spurling’s test, palpation of tenderness of 16 defined tender points in fibromyalgia and assessment of the ability to relax the neck and shoulder muscles. The perceived bodily pain distribution drawing and numerical 9-point pain assessment during activity and rest were also assessed. Perceived tiredness/fatigue and lack of concentration were also assessed on 9-point numerical scales. Subjective health complaints (SHC) during the last month were assessed on the SHC Questionnaire [21]. The examinations covered the following seven dimensions: movements of the neck, bodily pain distribution and intensity, tenderness, fatigue, activity of daily living and subjective health complaints. The number of indicators of changes in the dimensions defined the global categories of change.

2.5.4 Examination by the psychologist (NIL)

In the psychological evaluation the clinical interview was followed by a number of assessments within the following areas; (1) Pain experience, (2) Coping style and pain behavior, such as kinesiophobia and catastrophizing, (3) Psychological symptoms, such as depressive symptoms (based on Beck Depression Inventory II), post-traumatic stress symptoms as well as personality based profiles, (4) Cognitive/neuropsychological tests (i.e. attention, memory, executive and general cognitive/intellectual functions).

The primary endpoint was patient reported pain experience assessed along three dimensions (a) the degree of pain experienced in general during the last week, (b) to what degree did pain influence and restrict daily activities during the last week and (c) to what degree did pain influence and restrict your work function during the last week. The assessments were made on 6-point verbal descriptive scales. The criteria for the categories of the global scale of change were defined conditionally on the change in experienced pain. This means that the categories better and much better required a decrease in perceived pain at follow-up.

2.6.1 Basic approaches

Part of the aim was to evaluate the efficacy of cervical fusion surgery in a well-defined group of WAD patients, which motivates the per-protocol (PP) analysis. This means that only those patients who complied with the allocated treatments were evaluated as surgery (S), rehabilitation (R), and comparison (C) patients, respectively. A group D of ten non-compliers also participated in the follow-up examinations and was included in the final analyses.

The intention-to-treat (ITT) analysis was also performed in order to report the effectiveness of the two treatments, which is the overall effect of the planned interventions, irrespective of the patients’ compliance with the allocated treatments.

Ordered categorical scale assessments generate ordinal data, which means that a category only represents an ordered level and not a numerical value in a mathematical sense. These non-metric properties are well recognized, and several authors have stressed the fact that calculating sums and differences is not appropriate for such data [22], [23], [24], [25], [26], [27], [28], [29]. Statistical methods that take account of these non-metric properties of ordered categorical data were used. The frequency distributions of the experts’ assessments on the global transition scale categories, much worse, somewhat worse, unchanged, somewhat better and much better in the treatment groups were described in tables and bar charts, indicating the quartiles (25%, 75%) and the median (50%) [24], [30], [31].

2.6.2 Inter-group comparisons; treatment effects

The proportions of patients assessed as somewhat or much better by each of the experts at the follow-up examinations after fusion surgery or rehabilitation and the 95 percentage confidence intervals (95% CI) were calculated and compared [30], [32].

The frequency distributions of assessments on the five categories of change for the different PP and ITT groups are shown together with the median categories. Possible differences in the distributions of assessments on the expert-specific outcome scales between the groups were analyzed by the Kruskal-Wallis test [24], [30].

2.6.3 Inter-expert comparisons

The impact of the experts’ different perspectives of assessing the patient’s global outcome was evaluated by Svensson’s approach for paired ordinal data [13], [15], [16], [17], [33], [34], [35], [36], [37]. The assessments made by one of the experts were compared pairwise with each of the other expert’s assessments and described in a square contingency frequency table. The pairs of complete inter-expert agreement in scale assessments appear in the main diagonal of the contingency table. The percentage inter-expert agreement (PA) was calculated.

The type of observed inter-expert disagreement was evaluated. The two sets of marginal frequencies in the contingency table show the distribution of each expert’s assessments. Different marginal distributions indicate the presence of a systematic inter-expert disagreement regarding the patient’s outcome because of the expert-specific operational definitions of the outcome scale categories. The difference between the probability that one of the experts assessed the patients’ outcomes to lower categories than did the other expert and the probability of the opposite defines the measure of relative position (RP). Correspondingly the difference between the probabilities of how the experts concentrate their assessments on the scale categories defines the measure of relative concentration (RC). Possible RP and RC values range from –1 to 1. The value of zero indicates unbiased assessments.

Additional individual variability in pairs of assessments that cannot be explained by systematic inter-expert differences is measured by the relative rank variance (RV) [13], [15], [16], [17], [33], [34], [35], [36], [37]. The RP, RC and RV measures and the 95% confidence intervals (CI) of the measures were calculated using a free software program [38].

3.1.1 Fusion surgery versus multimodal rehabilitation

In the per-protocol evaluation only those 18 and 17 patients, who consented and adhered to the allocated treatments were evaluated as belonging to group S and group R, respectively. The distributions of changes in outcome after the S and R treatments are shown in Fig. 1A and B, respectively.

Fig. 1:

The relative frequency distributions of the four experts’ follow-up assessments of the global outcome of change after treatment. The median (50%) and quartiles (25%, 75%) are indicated. (A) The surgery group (n=18). (B) The rehabilitation group (n=17).

The relative frequency distributions of the four experts’ outcome assessments show that a majority of the patients who underwent fusion surgery (group S) were assessed as somewhat or much better, ranging from 67% to 78% dependent on the clinical discipline, Fig. 1A.

In the neurological examination 78% (95% CI, 55–91%) of the 18 patients were assessed as improved. This means that in a representative group of patients with the same inclusion criteria [9] one can expect that between 55% and 91% of the patients will be neurologically assessed as somewhat or much better after fusion surgery. The orthopedic surgeon and the psychologist judged 72% (95% CI, 49–88%), and the physiatrist 67% (95% CI, 44–84%) of the patients to be somewhat or much better after cervical fusion surgery.

Correspondingly, the proportions of patients who were assessed as somewhat or much better after the multimodal rehabilitation treatment ranged from 29% to 53%, Fig. 1B. The orthopedic surgeon judged 53% (95% CI, 31–71%), the psychologist 35% (95% CI, 17–59%), the physiatrist and the neurologist assessed 29% (95% CI, 13–53%) of the patients as somewhat or much better after multimodal rehabilitation.

The most pronounced difference between the treatment effects was seen by the neurologist, who found 48 percentage units (78% vs. 29%) more patients somewhat or much better after fusion surgery compared with multimodal rehabilitation. The 95% confidence interval ranging from 15 to 69 percentage units confirms the significant difference in treatment effect from the perspective of the neurologist. Correspondingly, 37 percentage units more patients (95% CI, 4–60) were judged somewhat or much better by the psychologist and by the physiatrist. According to the assessments by the orthopedic surgeon, the difference in the proportion of patients with improved outcomes in favor of fusion surgery was 19 percentage units (95% CI, −12 to 46).

3.1.2 Inter-group per-protocol analyses

The per-protocol analyses of all patients evaluated in the follow-up examinations involved four groups: the two treatment groups S and R, the comparison group C, and the group D of ten non-compliers that also participated in the follow-up examinations.

The frequency distributions of independent assessments of global change in patients made by the four experts ranged from much worse to much better (Table 1). The majority of the 18 patients in the surgery (S) group were assessed as much better by the psychologist and the physiatrist, and somewhat better by the neurologist and the orthopedic surgeon. The median outcomes of the 17 patients in the rehabilitation (R) group were unchanged except for the assessments made by the orthopedic surgeon, where the median outcome was somewhat better. The median outcomes of the patients in groups C and D were unchanged.

The statistical analyses confirm the significant differences in outcomes between the groups in favor of better outcomes in patient treated by fusion surgery (group S), the p-values ranging from 0.003 to 0.04.

3.1.3 The inter-group intention-to-treat analyses

The intention-to-treat analysis includes the two treatment groups (S and R) and the comparison group (C). The frequency distributions of the outcome assessments made by each of the experts differed between the three treatment groups in favor of the surgery group (S), with the median level of change being somewhat better (Table 2).

The statistically significant differences in outcomes between the three groups were most pronounced in the neurological evaluation (p=0.00003) and were explained by the superiority of outcomes in group S. The inter-group comparisons of treatment effects made by the physiatrist were less evident for general conclusions, p=0.14.

3.2.1 Inter-expert outcome assessments

The frequency distributions of global changes in patients of the treatment groups S and R that were assessed by the four experts ranged from somewhat worse to much better (Table 1). As also demonstrated in Fig. 1, the majority of the 18 patients in the surgery (S) group were assessed as somewhat or much better, the median being somewhat better according to the neurologist, the orthopedic surgeon, and the physiatrist, and much better according to the psychologist. The median outcomes of the 17 patients in group R were unchanged except for the assessments made by the orthopedic surgeon, the median outcome being somewhat better.

As evident by Table 1 the frequency distributions differ between the experts. To what extent do the experts agree in assessing a patient’s global score of change in whiplash related complaints? For example, the physiatrist judged seven patients in the S group as much better on the follow-up examination than before surgery. Are these patients found among the ten patients with the same level of outcomes according to the psychologist?

3.2.2 Inter-expert disagreements in outcome after fusion surgery

Figure 2 shows the paired distribution of assessments of global changes in the surgery group of patients made by the psychologist and by the physiatrist. These experts agreed in assessing the seven patients as much better after surgery. The remaining three patients, who were much better according to the psychologist, were regarded as somewhat better or unchanged by the physiatrist. Ten of the patients were assessed the same outcome scores by these two experts, the percentage agreement, PA, is 56%. Could the observed disagreement be explained by the different operational definitions for the two experts´ global scale categories?

Fig. 2:

Frequency distribution of pairs of independent assessments of change in WAD related complaints in patients after surgery (Group S) made by the psychologist and the physiatrist. The agreement diagonal is marked.

The two sets of frequency distributions of assessments made by the two experts appear as marginal totals and indicate that the psychologist systematically used higher rather than lower scores of global change than did the physiatrist. This systematic difference in position of the outcome scale assessments is also evident by the corresponding relative frequency distributions of the bar charts in Fig. 1A. Both the median and the third quartile outcome categories of the assessments by the psychologist are much better, compared with the median category somewhat better and the third quartile much better in the assessments by the physiatrist. This systematic difference in outcome scale assessments between these two experts is statistically confirmed by the significant measure of systematic disagreement in relative position, RP: −0.19 [95% CI (RP), −0.34 to −0.04]. The negative RP value indicates that the group of patients who underwent surgery more likely, (0.04–0.34 percentage units) will get higher scored outcome levels by the psychologist than by the physiatrist rather than the opposite.

The relative frequency distributions of the four experts’ outcome assessments of the surgery group of patients, (Fig. 1A), indicate a systematic difference in position regarding the proportion patients assessed as much better after surgery. The significant RP-values of the pairwise comparisons of outcome assessments made by the psychologist and by the three other experts, respectively, (Table 3A), confirm the larger probability that a group of S-patients will get a higher rather than a lower outcome level from the psychologist than from the other experts.

As evident by Fig. 1A and Table 1 the neurologist assessed a majority (67%) of the patients in the surgery group as being somewhat better, which is the median as well as the quartile categories. The paired distribution of assessments by the neurologist and the orthopedic surgeon in the S group of patients, Fig. 3, shows that they agree in outcomes of seven patients, PA, 39%. The neurologist concentrated the assessments of 12 patients to somewhat better. Four of these patients were assessed as much better, and two as somewhat worse by the orthopedic surgeon. The disagreement is partly explained by a systematic disagreement in how they concentrate their assessments; the measure of concentration, RC is −0.31 [95% CI (RC), −0.55 to −0.07], see Table 3A.

Fig. 3:

Frequency distribution of pairs of independent assessments of change in WAD related complaints in patients after surgery (Group S) made by the neurologist and the orthopedic surgeon. The agreement diagonal is marked.

Furthermore, the three disagreeing pairs of assessments: (much better, somewhat worse), and (somewhat better, somewhat worse) explain the high RV-value, 0.24, of disagreement that is not covered by the systematic disagreement.

3.2.3 Inter-expert disagreement on outcome after multimodal rehabilitation

The frequency distributions of the experts’ assessments of change in the patients who underwent the multimodal rehabilitation, Table 1 and Fig. 1B, show that three of the experts regarded most patients as unchanged, contrary to the orthopedic surgeon, who assessed two patients as unchanged on the follow-up occasion. The percentage agreements of the pairwise comparisons between the orthopedic surgeon and the other experts range between 18% and 41%, (Table 3B).

The paired distribution of assessments made by the psychologist and the orthopedic surgeon, Fig. 4, shows that these experts agreed in the outcome of three of the patients, the percentage agreement being 18%, (Table 3B). One of the ten patients who were assessed as unchanged by the psychologist was judged the same outcome level by the orthopedic surgeon. The outcomes of the other nine patients ranged from somewhat worse to much better according to the orthopedic surgeon. Eight of the patients were assessed to lower and six to higher outcome levels by the orthopedic surgeon than by the psychologist. This variability in assessments of patients, as evident by the high RV-value, 0.47, is one of the main reasons for the low level of agreement in assessments. Another reason is the systematic disagreement in how the two experts concentrate the assessments on the outcome levels. The measure of RC, −0.42 [95% CI (RC), −0.75 to −0.09] means that the psychologist will more likely concentrate the assessments to the outcome level unchanged than will the orthopedic surgeon, who used the categories somewhat worse and somewhat better in assessing 12 patient, (Fig. 4). The same systematic inter-expert disagreements in concentration hold for the other experts when compared with the orthopedic surgeon, see Table 3B.

Fig. 4:

Frequency distribution of pairs of independent assessments of change in WAD related complaints in patients after rehabilitation (Group R) made by the psychologist and the orthopedic surgeon. The agreement diagonal is marked.

As evident by the Table 3B, the individual variability in the pared comparison of the assessments of the patients’ outcomes, the RV, ranged from 0.03 to 0.47.

Figure 5 shows the paired comparison between the neurologist and the physiatrist, who agreed in assessing outcomes in eight patients, 47%. The disagreeing pairs are close to the agreement diagonal, which means a negligible individual variability, RV, 0.03. The frequency distributions of assessments of the group of patients are similar which means homogeneous paired assessments.

Fig. 5:

Frequency distribution of pairs of independent assessments of change in WAD related complaints in patients after rehabilitation (Group R) made by the neurologist and the physiatrist. The agreement diagonal is marked.

3.2.4 Systematic inter-expert differences in outcome assessments

In summary, the frequency distributions of outcome levels in the group of patients after fusion surgery differed between the experts. The percentage agreements, PA, in outcome assessments range from 33% to 56% (Table 3A), and the disagreements are explained by systematic disagreements between the professionals’ outcome scale categories. The psychologist systematically assessed more patients as much better and the neurologist systematically concentrated the outcomes to somewhat better than did the other experts. The additional individual variations in paired assessments, the RV-values were mainly negligible; the RV ranged from 0.02 to 0.07 except for the paired assessments of the neurologist and the orthopedic surgeon, the RV being 0.24 (Table 3A).

The frequency distributions of outcome levels in the group of patients after rehabilitation also differed between the experts. The percentage inter-expert agreements, PA, in outcome assessments made by the orthopedic surgeon and each of the other experts ranged from 18% to 41%, and the main reason for disagreements was the systematic disagreement in how the experts concentrated their outcome assessments as evident by the significant RC-values (Table 3B). As shown in Table 1 and Fig. 4 the orthopedic surgeon regarded two patients as unchanged in contrast to the other experts, who assessed a majority of the patients as unchanged on the follow-up examinations. Another reason for disagreement was the individual variations in assessing outcomes, the RV ranged from 0.03 to 0.47 (Table 3B).