Depression and the Link Between Radiographic OA, Knee Pain
Depression and the Link Between Radiographic OA, Knee Pain
Factor analysis for knee pain questions (dichotomous variables) identified only one factor and all items showed a factor loading higher than 0.86, and a global Cronbach's Alpha of 0.66 (Table 2). So we decided to consider these items as a score for knee pain and we classified participants as having no knee pain (score −1), knee pain but no other positive answer (score 0), and score 1, score 2 or score 3, according to the number positive answers. Overall radiographic knee OA (KL ≥ 2) was present in 45.4% of the participants and 42.2% reported "having had knee pain not related with any trauma or injury". Stratifying by radiographic classification, knee pain was reported by 53.2% of those with radiographic KL ≥ 2 and by 33.2% of those with radiographic KL < 2. Also, the proportion of participants with higher pain scores was larger among those with radiographic KL ≥ 2 (Figure 1).
(Enlarge Image)
Figure 1.
Distribution of participants according to knee pain and radiographic findings.
We found that participants with higher pain scores had higher odds of having depressive symptoms. After adjustment for age, BMI and gender and considering participants with no pain (pain score −1) as reference category we found: OR = 1.22 (95% CI 0.53; 2.79) for pain score 0, OR = 1.71 (95% CI 0.77; 3.77) for pain score 1, OR = 3.51 (95% CI 1.92; 6.44) for pain score 2 and OR = 5.64 (95% CI 2.85; 11.16) for pain score 3.
The association between pain score and radiographic knee OA by categories of depressive symptomatology is presented in Table 3. The odds of having radiographic knee OA (KL ≥ 2) was higher in participants with higher scores of pain, both in patients with and without depressive symptoms; however the differences were higher among those with BDI < 14, even after after adjustment for age, BMI and gender.
The sensitivity, specificity and likelihood ratio regarding each knee pain question is described in Table 4. The question with the lowest sensitivity was "During the last 6 months, did knee pain last longer than a week?" (32.9%). The question regarding pain episodes in the last month presented the highest sensitivity value (72.8%). In general, small likelihood ratios were obtained regarding the pre and post-test probability of having radiographic OA, showing the low ability of these questions to identify participants with knee radiographic OA.
Considering the pain score (Table 5), score −1 (participants that reported no pain) showed a high sensitivity to identify patients with knee OA (46.8%), which probably reflects the high number of participants with knee radiographic KL ≥ 2 but without pain; however this score also presented a very low specificity (33.1%).
Among those that reported at least one positive answer on the pain questionnaire (scores from 0 to 3) the ability of knee pain to identify patients with radiographic knee OA increased with increased scores. When we used only the "have you ever had knee pain" question (score 0) or even with score 1 ("have you ever had knee pain" and another positive question) a very low sensitivity was reached (9.3% and 8.3%, respectively) and we obtained a low discrimination ability (positive likelihood ratio of 0.80 and 0.91). Scores 2 and 3 presented a high specificity (>90%) but a low sensitivity (<20%), but nevertheless this was twice as high as the sensitivity obtained with the score 1 and 2. Based on the positive likelihood ratio, the pos-test probability of having a knee radiographic KL ≥ 2 for those who scored 2 was twice that of the pre-test probability and this increased to 4.35 when participants scored 3. Similar results were found by sex. However, additional positive answers in males contributed to higher likelihood ratios than in females (Table 5).
To analyze the role of depressive symptoms in the discrimination ability of knee pain to identify individuals with radiographic OA, we decided not to stratify by sex in order to have enough power. The prevalence of depression (BDI > 14) was 19.9% among participants with radiographic KL ≥ 2 and 12.6% in those with radiographic KL < 2 (p = 0.01). Among participants with BDI ≤ 14 additional positive answers (increased knee pain score) allowed an increase in the positive likelihood ratio: those with score 2 had twice the probability of having radiographic KL ≥ 2 than before the questions and this increased to 7.34 for those who scored 3. In the presence of depressive symptoms BDI > 14 the ability of these questions to identify participants with radiographic knee OA became lower, with a positive likelihood ratio of 1.92 for those who scored 2 and 1.82 for those who scored 3 (Table 6).
Results
Factor analysis for knee pain questions (dichotomous variables) identified only one factor and all items showed a factor loading higher than 0.86, and a global Cronbach's Alpha of 0.66 (Table 2). So we decided to consider these items as a score for knee pain and we classified participants as having no knee pain (score −1), knee pain but no other positive answer (score 0), and score 1, score 2 or score 3, according to the number positive answers. Overall radiographic knee OA (KL ≥ 2) was present in 45.4% of the participants and 42.2% reported "having had knee pain not related with any trauma or injury". Stratifying by radiographic classification, knee pain was reported by 53.2% of those with radiographic KL ≥ 2 and by 33.2% of those with radiographic KL < 2. Also, the proportion of participants with higher pain scores was larger among those with radiographic KL ≥ 2 (Figure 1).
(Enlarge Image)
Figure 1.
Distribution of participants according to knee pain and radiographic findings.
We found that participants with higher pain scores had higher odds of having depressive symptoms. After adjustment for age, BMI and gender and considering participants with no pain (pain score −1) as reference category we found: OR = 1.22 (95% CI 0.53; 2.79) for pain score 0, OR = 1.71 (95% CI 0.77; 3.77) for pain score 1, OR = 3.51 (95% CI 1.92; 6.44) for pain score 2 and OR = 5.64 (95% CI 2.85; 11.16) for pain score 3.
The association between pain score and radiographic knee OA by categories of depressive symptomatology is presented in Table 3. The odds of having radiographic knee OA (KL ≥ 2) was higher in participants with higher scores of pain, both in patients with and without depressive symptoms; however the differences were higher among those with BDI < 14, even after after adjustment for age, BMI and gender.
The sensitivity, specificity and likelihood ratio regarding each knee pain question is described in Table 4. The question with the lowest sensitivity was "During the last 6 months, did knee pain last longer than a week?" (32.9%). The question regarding pain episodes in the last month presented the highest sensitivity value (72.8%). In general, small likelihood ratios were obtained regarding the pre and post-test probability of having radiographic OA, showing the low ability of these questions to identify participants with knee radiographic OA.
Considering the pain score (Table 5), score −1 (participants that reported no pain) showed a high sensitivity to identify patients with knee OA (46.8%), which probably reflects the high number of participants with knee radiographic KL ≥ 2 but without pain; however this score also presented a very low specificity (33.1%).
Among those that reported at least one positive answer on the pain questionnaire (scores from 0 to 3) the ability of knee pain to identify patients with radiographic knee OA increased with increased scores. When we used only the "have you ever had knee pain" question (score 0) or even with score 1 ("have you ever had knee pain" and another positive question) a very low sensitivity was reached (9.3% and 8.3%, respectively) and we obtained a low discrimination ability (positive likelihood ratio of 0.80 and 0.91). Scores 2 and 3 presented a high specificity (>90%) but a low sensitivity (<20%), but nevertheless this was twice as high as the sensitivity obtained with the score 1 and 2. Based on the positive likelihood ratio, the pos-test probability of having a knee radiographic KL ≥ 2 for those who scored 2 was twice that of the pre-test probability and this increased to 4.35 when participants scored 3. Similar results were found by sex. However, additional positive answers in males contributed to higher likelihood ratios than in females (Table 5).
To analyze the role of depressive symptoms in the discrimination ability of knee pain to identify individuals with radiographic OA, we decided not to stratify by sex in order to have enough power. The prevalence of depression (BDI > 14) was 19.9% among participants with radiographic KL ≥ 2 and 12.6% in those with radiographic KL < 2 (p = 0.01). Among participants with BDI ≤ 14 additional positive answers (increased knee pain score) allowed an increase in the positive likelihood ratio: those with score 2 had twice the probability of having radiographic KL ≥ 2 than before the questions and this increased to 7.34 for those who scored 3. In the presence of depressive symptoms BDI > 14 the ability of these questions to identify participants with radiographic knee OA became lower, with a positive likelihood ratio of 1.92 for those who scored 2 and 1.82 for those who scored 3 (Table 6).
Source...