INHE 3003 Week 7 Statistics Interpretation Example

Detectable Is Not the Same as Large: A Statistics Interpretation of Four Results From the Tasmanian Turmeric Trial, Read as Reported and Not Recalculated

Student Name

Walden University

INHE 3003: Research Literacy for Integrative Health Professionals

Instructor Name

Month Day, Year

Detectable Is Not the Same as Large: A Statistics Interpretation of Four Results From the Tasmanian Turmeric Trial, Read as Reported and Not Recalculated

The Two Questions

Every result in a trial answers two separate questions. Was a difference detectable, meaning the data are hard to explain by chance alone? And how large was it, meaning would a person notice it? A p-value below 0.05 or a confidence interval that excludes zero speaks to the first question only. The size of the estimate and the range of its interval speak to the second. Treating a small p-value as proof of a large effect is among the most common misreadings of statistics (Greenland et al., 2016). This paper reads four results from the Tasmanian turmeric trial as they were reported (Wang et al., 2020), without recalculating anything, and keeps those two questions apart.

Result 1: Knee Pain on the Visual Scale

The extract group's knee pain fell by 9.1 mm more than the placebo group's on a 100-mm scale, with a 95% confidence interval from 17.8 mm to 0.4 mm and a p-value of 0.039. Detectable: yes, narrowly, since the interval stops just short of zero. Large: probably not. A study of patients with knee osteoarthritis found that the smallest improvement patients themselves considered meaningful was about 20 mm on this scale (Tubach et al., 2005). That threshold describes change within one person, not a difference between group averages, so the comparison is rough. But even the most favorable end of the interval, 17.8 mm, falls short of it, and the other end, 0.4 mm, is too small for anyone to feel.

Result 2: Effusion-Synovitis on MRI

The extract group's effusion-synovitis volume differed from the placebo group's by 3.2 mL, with an interval from -0.3 to 6.8 mL. Detectable: no, since the interval includes zero. But that is not evidence of no effect. The interval allows anything from a slight decrease to an increase of nearly 7 mL, so the trial cannot say what the extract does to swelling. The correct reading is that this trial could not detect a difference, not that the extract failed.

Result 3: WOMAC Pain

On the WOMAC pain subscale, a secondary outcome, the extract group improved by 47.2 mm more than placebo, with an interval from 81.2 to 13.2 mm and a p-value of 0.006. The number looks much larger than the visual-scale result, but it is measured on a different scale. In its visual-scale version, which a result reported in millimeters suggests, the WOMAC pain subscale sums five items each scored up to 100 mm, giving a range of 0 to 500 mm. Judged against that range, 47 mm is a change of under a tenth of the scale, similar in proportion to the primary pain result. The smaller p-value says the difference is more clearly detectable, not that it is larger in any sense a patient would feel.

Result 4: Adverse Events

Adverse events occurred in 39% of the extract group and 53% of the placebo group, with a p-value of 0.16. Detectable: no. With only 34 to 36 people per group, a difference of this size could easily arise by chance, and the trial was not designed to detect differences in harms. The reading is that the trial found no detectable difference in adverse events, not that the extract is safer than placebo.

What the Four Results Say Together

One primary outcome was detectable but small; the other was not detectable in either direction. A secondary outcome was detectable and similar in relative size to the first. Harms showed no detectable difference. A sentence that matches all four might read: in this 12-week trial, the extract produced a small, detectable reduction in knee pain compared with placebo, with no detectable effect on joint swelling.

References

Greenland, S., Senn, S. J., Rothman, K. J., Carlin, J. B., Poole, C., Goodman, S. N., & Altman, D. G. (2016). Statistical tests, P values, confidence intervals, and power: A guide to misinterpretations. European Journal of Epidemiology, 31(4), 337-350. https://doi.org/10.1007/s10654-016-0149-3

Tubach, F., Ravaud, P., Baron, G., Falissard, B., Logeart, I., Bellamy, N., Bombardier, C., Felson, D., Hochberg, M., van der Heijde, D., & Dougados, M. (2005). Evaluation of clinically relevant changes in patient reported outcomes in knee and hip osteoarthritis: The minimal clinically important improvement. Annals of the Rheumatic Diseases, 64(1), 29-33. https://doi.org/10.1136/ard.2004.022905

Wang, Z., Jones, G., Winzenberg, T., Cai, G., Laslett, L. L., Aitken, D., Hopper, I., Singh, A., Jones, R., Fripp, J., Ding, C., & Antony, B. (2020). Effectiveness of Curcuma longa extract for the treatment of symptoms and effusion-synovitis of knee osteoarthritis: A randomized trial. Annals of Internal Medicine, 173(11), 861-869. https://doi.org/10.7326/M20-0990