Effect of Opioid vs Nonopioid Medications on Pain-Related Function in Patients with Chronic Back Pain or Hip or Knee Osteoarthritis
JAMA, 2018 · paper · Clinical trials
A 12-month randomized trial in 240 veterans with chronic back pain or hip or knee osteoarthritis comparing opioid and nonopioid medications. Pain-related function did not differ; pain intensity was slightly better and medication side effects fewer with nonopioids.
8 findings
Substance Table 1 'Women' row prints "36 (13)" in both groups (n = 120 each), giving 72 women; the text and abstract give 32 women in total, and 36/120 is 30%, not 13%.
- p. 5 Table 1, Women row
36 (13)
- p. 5 Results, paragraph 2
32 patients (13.0%) were women (Table 1)
- p. 1 Abstract, Results
women, 32 [13.0%]
- p. 5 Results, paragraph 2
32 patients (13.0%) were women
Judge's note: The Table 1 crop and the page 5 image show 36 (13) in both columns. Both group sizes are n = 120. 2 × 36 = 72, not the 32 printed in the abstract and Results. The printed 13% fits 16 per group (16/120 = 13.3%; 16 + 16 = 32). No other population is described.
Second check: another model agreed
Report a problem with this finding
Substance Table 2, VR-12 mental health baseline row: "47.3 (11.2)" vs "47.8 (13.0)" with difference "−0.3 (−3.4 to 2.8)"; the rounded means allow only −0.4 to −0.6.
- p. 6 Table 2, VR-12 mental health, Baseline row
47.3 (11.2)
- p. 6 Table 2, VR-12 mental health, Baseline row
−0.3 (−3.4 to 2.8)
- p. 6 Table 2, VR-12 mental health baseline
47.3 (11.2) 47.8 (13.0) −0.3 (−3.4 to 2.8)
Only the authors can say what is right here.
Judge's note: Page 6 image. Footnote a says the differences are unadjusted time-specific comparisons, and both columns have n = 119. The difference of the rounded means is −0.5, and rounding cannot reach −0.3. The CI is centred on −0.3, with a half-width (~3.1) that fits the SDs. Every other VR-12 row agrees with its means to within 0.1. One of the means or the difference appears misprinted; only the authors can say which.
Second check: another model agreed
Report a problem with this finding
Substance Table 2 MIDAS 12 mo: "3.7 (11.6) 3.2 (11.6) −0.5 (−2.7 to 3.6)"; opioid minus nonopioid is +0.5 and the CI is centred on +0.45, so the minus sign on the estimate appears wrong.
- p. 7 Table 2 (continued), MIDAS, 12 mo row
−0.5 (−2.7 to 3.6)
- p. 7 Table 2, MIDAS 12 mo
3.7 (11.6) 3.2 (11.6) −0.5 (−2.7 to 3.6)
Judge's note: Page 7 image: the minus sign is printed. All other rows use opioid minus nonopioid, and the differences are unadjusted (footnote a). 3.7 − 3.2 = +0.5. The CI midpoint is +0.45. The half-width (~3.15) fits SD 11.6 with about 105 per group.
Second check: another model agreed
Report a problem with this finding
Substance Table 2 MFI mental fatigue 12 mo: "9.2 (3.9) 9.3 (4.3) 0.1 (−1.3 to 1.0)"; opioid minus nonopioid is −0.1 and the CI is centred on −0.15, so the estimate's sign appears wrong.
- p. 7 Table 2 (continued), MFI mental fatigue, 12 mo row
0.1 (−1.3 to 1.0)
- p. 7 Table 2, MFI mental fatigue 12 mo
9.2 (3.9) 9.3 (4.3) 0.1 (−1.3 to 1.0)
Judge's note: The page 7 image shows 0.1 with no minus sign. From the rounded means the difference can only lie between −0.2 and 0.0, so +0.1 cannot be reached. The CI midpoint is −0.15.
Second check: another model agreed
Report a problem with this finding
Substance Table 2 MFI reduced motivation 12 mo: "8.6 (3.2) 8.8 (3.7) −0.2 (−0.7 to 1.6)"; the estimate matches the means, but the interval is centred on +0.45 and is wider than the SDs and n imply.
- p. 7 Table 2 (continued), MFI reduced motivation, 12 mo row
−0.2 (−0.7 to 1.6)
Only the authors can say what is right here.
Judge's note: Page 7. Methods (p5) say two-sided t tests were used for unadjusted between-group comparisons, and footnote a says the comparisons are unadjusted, so the CI should be symmetric about the estimate. The midpoint is +0.45 against an estimate of −0.2. With 105 and 101 patients (footnote j missing counts), the SE is about 0.48, giving a half-width of about 0.95, so neither printed bound fits. The other rows are symmetric. The correct bounds can only come from the authors.
Second check: another model agreed
Report a problem with this finding
Substance Table 3 all-cause hospitalization '1' row: "15 (13) 16 (13) 1 (−9 to 8)"; 15/120 − 16/120 = −0.8 percentage points, so the estimate should be −1 (the CI is centred on −0.5).
- p. 8 Table 3, All-cause hospitalization, row '1'
1 (−9 to 8)
- p. 8 Table 3, All-cause hospitalization, row 1
15 (13) 16 (13) 1 (−9 to 8)
Judge's note: Page 8 image shows '1' with no minus sign. Footnote d gives n = 120 per group, and the convention is opioid minus nonopioid (ED row: −11). The category differences must sum to 0: 0 + (−0.83) + 0.83 = 0, and the ≥2 row correctly shows +1. The CI fits −0.8.
Second check: another model agreed
Report a problem with this finding
Substance Table 3 falls, opioid group: "63 (53)", "26 (22)", "29 (25)" sum to 118, but footnote f states "Missing data for 1 patient in the opioid group" (n = 119); 29/119 = 24.4%, not 25%.
- p. 8 Table 3, Number of falls in 12 mo after enrollment
29 (25)
- p. 8 Table 3, footnote f
Missing data for 1 patient in the opioid group.
Only the authors can say what is right here.
Judge's note: Page 8 image. The opioid categories sum to 118 against the 119 implied by footnote f. 29/119 rounds to 24%, while 30/119 = 25.2% would match the printed 25%. The nonopioid column (63 + 17 + 39 = 119) is also one short of 120, and the footnote gives no missing data for that group. The counts need checking by the authors.
Second check: another model agreed
Report a problem with this finding
Substance Table 3, 'Unexplained prescription drug positive' row: '11 (10)' vs '9 (8)' with difference '3 (−5 to 10)'. With the denominators given in footnote g, the count, percentage and difference do not agree with one another.
- p. 8 Table 3, Unexplained prescription drug positive row
3 (−5 to 10)
- p. 8 Table 3, urine drug tests
11 (10) 9 (8) 3 (−5 to 10)
Only the authors can say what is right here.
Judge's note: I checked the Table 3 image on p8. Footnote g says 'Missing data for patients: 4 in the opioid group and 6 in the nonopioid group', so the denominators are 116 and 114 out of 120 per group. 11/116 = 9.5%, which rounds to 9, not the printed 10. 9/114 = 7.9%, matching the printed 8. The difference is 1.6, which rounds to 2, but 3 is printed. The CI midpoint (−5+10)/2 = 2.5 is also above 1.6. The adjacent rows reproduce with the same denominators: illicit drug 6/116 − 12/114 = −5.3, printed −5; hazardous alcohol 2/116 − 4/114 = −1.8, printed −2; misuse behaviour, with n = 119 in each group, 11/119 − 8/119 = 2.5, printed 3. An opioid count of 12 would give 10.3%, which fits the printed 10 and the CI bounds, but the difference would still round to 2. The results text gives no other explanation. Only the authors can say which figure is wrong.
Second check: another model agreed
Report a problem with this finding
7 minor (typos and the like; most checked by the judge only)
Substance Abstract and Results print "women, 32 [13.0%]" / "32 patients (13.0%) were women" for 240 randomized patients; 32/240 is 13.3%.
- p. 1 Abstract, Results
women, 32 [13.0%]
- p. 5 Results, paragraph 2
(13.0%) were women
- p. 5 Results, paragraph 2
32 patients (13.0%) were women
Judge's note: Viewed the abstract (p1) and Results paragraph 2 (p5). Both use 240 randomized patients as the denominator, and no other denominator is stated. 32/240 = 13.33%, which rounds to 13.3%, not 13.0%. The Table 1 count mismatch is handled separately in C006.
Second check: another model agreed
Report a problem with this finding
Reference or label Table 1 labels baseline depression "Moderate depression (PHQ-9 score ≥10)" (abbreviation PHQ-9, 9-Item), whereas Methods and Table 2 name the depression measure as the "8-Item Patient Health Questionnaire (PHQ-8)".
- p. 5 Table 1 and its abbreviations
Moderate depression (PHQ-9 score ≥10)
- p. 4 Secondary Health Outcomes
8-Item Patient Health Questionnaire (PHQ-8) depression measure
Only the authors can say what is right here.
Judge's note: Viewed p4 (Secondary Health Outcomes names PHQ-8), p5 (Table 1 row and abbreviations say PHQ-9) and p6 (Table 2 says PHQ-8). The Methods never mention a PHQ-9. The PHQ-9 might have been used as a screener, but the paper does not say so, so this is a label query for the authors.
Second check: another model agreed
Report a problem with this finding
Wording "There were no significant differences in adverse outcomes or potential misuse measures (Table 3)" directly follows the report that the primary adverse outcome (medication-related symptoms) differed significantly (P = .03).
- p. 7 Adverse Outcomes
There were no significant differences in adverse outcomes or potential misuse measures (Table 3).
Judge's note: Page 7 text and Table 3 (p8). The primary adverse outcome was significant. The sentence evidently means the secondary adverse outcomes, but as worded it covers all adverse outcomes, so it is ambiguous.
Second check: another model agreed
Report a problem with this finding
Typo Goldsmith's affiliation reads "University of Minnesota School of Public Health, Minneapolis, Minneapolis, Minnesota (Goldsmith)", with the city printed twice.
- p. 1 Author Affiliations
Minneapolis, Minneapolis, Minnesota
- p. 1 Author Affiliations
Minneapolis, Minneapolis, Minnesota (Goldsmith)
Judge's note: The page 1 image of the affiliations column shows 'Minneapolis, Minneapolis, Minnesota'. It is not an extraction artefact.
Second check: none (judge only)
Report a problem with this finding
Typo Byline gives "Kurt Kroenke, MD; Matthew J. Bair; Siamak Noorbaloochi, PhD": Bair is the only author printed without degrees.
- p. 1 Author byline
Matthew J. Bair; Siamak Noorbaloochi, PhD
Only the authors can say what is right here.
Judge's note: Rests on outside knowledge: that Bair holds degrees which should appear in the byline. The page 1 image confirms that every other author has credentials and Bair has none. Whether that is an omission is for the authors to say.
Second check: none (judge only)
Report a problem with this finding
Typo Pain intensity response "63 patients (53.9%) in the nonopioid group": with the denominator of 117 implied by the neighbouring percentages, 63/117 = 53.85%, which rounds to 53.8%.
- p. 6 Results, response rates
63 patients (53.9%)
- p. 7 Results, response rates
difference, −12.8% (95% CI
- p. 6 Results, response analysis
63 patients
- p. 7 top of page
(53.9%) in the nonopioid group; difference, −12.8%
Judge's note: Pages 6–7. 69/117 = 59.0%, 71/117 = 60.7% and 48/117 = 41.0% fix the denominator at 117. 63/117 = 53.846%, which rounds to 53.8%, not 53.9%. The printed difference −12.8% is correct from the counts (−15/117 = −12.82%), so the only defect is the rounding of one percentage.
Second check: none (judge only)
Report a problem with this finding
Typo Conflict of Interest Disclosures ends 'No other disclosures are reported' with no full stop.
- p. 10 Article Information, Conflict of Interest Disclosures
No other disclosures are reported
Judge's note: The crop of p10 shows the sentence ending with no period before 'Funding/Support:'. This is a cosmetic punctuation omission.
Second check: none (judge only)