Tuesday, May 24, 2005

More on bended knees, and shoe fetishes - reviewers and editors asleep at the switch

BrooklynDodger has been thinking about lower extremity disorders related to occupation. The Dodger found an adulterated [was going to say toxic] abstract of a paper with some interesting data.

Interviews in participants' homes of 29 cases identified from orthopaedic hospital surgical waiting list, and 83 controls identified by general practitioners were conducted. Women aged 50 to 70 years. Univariate analyses identified several significant associations with OAK including past knee injury, arthritis of the feet, heavy smoking, being overweight (BMI 25 or above) and various occupational activities. Virtually all measures of high heeled shoes use were associated with reduced risk of OAK, although none of these findings were statistically significant. In multivariate analysis only BMI 25 or above at age 36-40 remained significantly associated with OAK (OR 36.4, 95% CI 3.07 to 432), although weak evidence suggested certain occupational activities might increase risk. Being overweight before the age of 40 considerably increased the risk of subsequent symptomatic OAK in women. Wearing high heeled shoes did not.

Obviously, the Dodger is frosted about the "weak evidence" part. The authors also give us two more weasel works, "suggested" and "might" for what is really pretty good data given the limits of the study. First, this is a seriously underpowered study. Underpowered studies are usually condemned to uncertain results, but not necessarily "weak" results.

Second, the "weak" applies to the p value [as discussed below], not the risk ratio. The Dodger hates the term weak, but it conventionally applies to the risk ratio alone, with statistical uncertainty addressed with some other adjective.

The occupational activities were lifting, where a reported 24 to 33 years gave a significant risk ratio of 7.31; bending, where 21-33 years gave a signficant risk ratio of 12.36, and kneeling where 33 years plus gave a risk ratio of 4.18, cleaning floors on knees gave a risk ratio of 5.28. In each category the controls had substantially higher proportions in the lowest tertile than victims. There was falloff in risk ratio in the highest tertile of victims, plausibly explainable by disability related termination of exposure. None of these are "weak" associations.

Multivariate analysis gave a risk ratio for 21–34 years of bending of 12.91, with a upper limit of 167 (!!!), marginally significant. For 34 years, the risk ratio of 9.81 was not signficant, but upper limit was 125.

Lacking a plausible mechanism for smoking increasing OA, it is plausible that the increase of smoking among controls is caused by pain. The association of BMI with increased OA is mechanistically plausible, but increased BMI caused by pain may account for some of the association. Persons in industrial jobs are frequently selected for being big at the start, another source of uncontrolled confounding with a bias to the null. The cut point of BMI = 25 is equivalent to 155 pounds of weight for a 5'6" tall person, which is not exactly tubby.

Although Manolo Blahnik and shoe fetishists can take some comfort in these data, the Dodger would like to see the cross tabs for BMI and high heeled shoes, and occupational risk factors and high heeled shoes, to rule out confounding.

J Epidemiol Community Health. 2003 Oct;57(10):823-30.

An investigation of risk factors for symptomatic osteoarthritis of the knee in women using a life course approach.

Dawson J, Juszczak E, Thorogood M, Marks SA, Dodd C, Fitzpatrick R.Oxford Centre for Health Care Research and Development (OCHRAD) School of Health and Social Care, Oxford Brookes University, 44 London Road, Oxford OX3 7PD, UK. jdawson@brookes.ac.ukSTUDY

No comments: