Hall and colleagues1 have presented an interesting study linking a number of physiological variables to outcomes in persons who have undergone hip arthroplasty. Presumably, the motivation for the study was to identify strategies to reduce the perioperative stress response, which would in turn aid recovery.
Essentially this study found that there was a relationship between the inflammatory response, as measured by interleukin 6 (IL-6) and the number of postoperative days before patients could walk 10 and 25 m independently. Outcomes 1 and 6 months after surgery, as measured by the WOMAC (a self-report pain and activity questionnaire with excellent psychometric properties), were not influenced by any of the physiological variables measured in the study. The endocrine response did not appear to influence either the short-term or the long-term outcome.
A question comes to mind in interpreting these findings. Is there a biological reason for suspecting that any of the postoperative physiological variables could have enduring effects such that they might influence walking and activity?
There is supporting evidence that surgical procedures necessitating extensive tissue damage produce a greater inflammatory response.2 For example, in the 11 studies comparing open with laparoscopic cholecystectomy, summarized recently by Kehlet,3 the lower inflammatory response in the laparoscopic group was associated with shorter length of hospital stay.
Length of stay in hospital, however, is not necessarily the best indicator of recovery. This measure is influenced by the health-care system and the administrative culture in which the patient and physician function, as well as by the patients expectation of hospital stay and the availability of community and family postoperative support. In other words, administrative end-points such as the length of stay are influenced by many factors outside the investigators control. Thus, an investigator may be unwise to rely solely on these outcomes, however attractive they may be from an administrative point of view. With length of stay as the only outcome, there is the risk of a study finding no effect when there is one, or of wrongly assuming the patient had recovered just because he or she had been discharged from the hospital.
Clearly, if the interest is in measuring recovery after surgery, it would be best to pick a measure of this desirable outcome. This was the dilemma that Hall and colleagues encountered in their study, because, as they point out, there are no accepted and validated measures of postoperative recovery.
As a proxy measure of recovery, they chose the number of days it took patients to walk 10 and 25 m.4 This measure has face validity (makes sense) for recovery, but it may not satisfy another desirable property of an outcome measure, namely being independent of outside influences. Many factors could delay the day on which a patient walked a target distance. Perhaps a test was scheduled, a family crisis intervened, or the physiotherapist was absent. Notwith standing factors outside the investigators control, time-to-event data have statistical features that make them more complex to analyse and interpret. Statistically speaking, the data are discrete and ordinal (i.e. 1, 2 or 3). The problem is that much valuable information on the patients ability to walk is not incorporated in this crude measure. Loss of information introduces misclassification such that like and unalike can be grouped together and like and like can be separated.
For example, consider that a person who walks 6 m four times a day for the first 3 days and then the targeted 10 m on day 4 (24 m x 3 days) would be classified the same as someone walking not at all for 3 days and then walking the full 10 m once on day 4.
When misclassification is random, i.e. it is unrelated to the exposure (to IL-6 in this study), then only noise is introduced, making it more difficult to detect an association. In this study, there may be concern that non-random misclassification (systematic error) would have crept in if persons with a lower inflammatory response had been able to walk a slightly longer distance or more frequently during the day, but did not necessarily achieve the target distance any earlier. If this is true, the result is a biased estimate of the effect of inflammatory response on recovery of walking and the bias is towards the null hypothesis, underestimating the effect of IL-6 on recovery. As there were many non-associations in this paper, particularly with respect to endocrine responses, questions arise as to whether these were true negatives or false negatives introduced by non-random misclassification.
So what is an optimal measurement strategy for interventions that have important physiological and patient effects? Clearly, the outcome must be biologically related to the intervention and should not be influenced by extraneous factors. Moreover, information should not be lost because of imprecision of the measure. Ideally, the precision of the measurement should be close to the biological precision of the underlying construct. Walking ability is not measured on a yes/no scale. It is a complex construct influenced by strength, range of motion, balance and the presence of pain, and can be assessed in terms of stride length, distance covered, time and degree of assistance. In the study of Hall and colleagues, the exposure variables (inflammatory and endocrine variables) were measured with great accuracy; however, the measure of outcome was not subjected to the same precision. As a general rule, outcome should be measured with as much accuracy as is biologically possible.
What are the measures that are biologically related to recovery after total hip replacement in the short term and in the long term? Postoperative mobility is an appropriate construct reflecting not only hip-related function but also general condition. Walking is one measure of mobility, but the concepts of distance, speed, endurance and pain need to be incorporated. For example, measures such as gait speed,5 2- or 6-min walks,6 7 the timed Up-and-Go test8 and total distance walked per day are more precise measures of walking, and are not difficult to incorporate into routine clinical practice and research.
While the WOMAC is considered an excellent measure of health status for people with osteoarthritis of the hip or knee,9 the areas covered in the WOMAC may not be directly influenced by the physiological variables measured in the hours and days after surgery. Rather, the link may be through an intermediate outcome such as mobility. It is more likely that a study would show an association between two variables that have a direct biological link than between two variables that are more loosely connected.
Mobility is just as relevant a measure 1 and 6 months after hip replacement as it is postoperatively. The WOMAC captures mobility but also taps other constructs, such as emotional well-being and social participation, which are connected to postoperative variables more remotely and are also influenced by outside elements. In the long term, measures of community mobility are particularly relevant and are associated with overall health status and quality of life. A test such as the 6-min walk,7 a measure of functional exercise capacity, is attractive as it is measured objectively with a standardized protocol (distance covered in 6 min); its influencing variables (effort and practice) are known; it is measured on a continuous scale, which makes statistical analyses simpler yet more powerful; and it has a natural zero point when an individual is unable to walk, thus eliminating the problem of what to do with people with limited walking ability. The test also has age and sex norms.
In the design of studies of surgical outcomes, it could be proposed that the measurement strategy should consider the link between biology and outcome, the influence of other variables on the chosen outcome, and the statistical properties of the measures. Clearly, the timing of the measurement must also reflect the time course of recovery, which would differ by type of surgery. For example, for hip arthroplasty, 6 months is a relevant time point, whereas for abdominal surgery 6 weeks may be more appropriate.
In choosing a measurement strategy, it is helpful to consider how the various parameters that could be studied influence each other and what are the likely causal pathways. We propose the simple model depicted in Figure 1. It is well known that surgery is a stressor and that both the surgery and the associated stress produce immediate biological changes that can be measured in the perioperative period. These changes are physiological and systemic. In our model, we have depicted these strong relationships with solid arrows. It is biologically plausible that these short-term physiological effects have an impact on short-term function. Hall and colleagues studied this association and provided some evidence of a relationship, but the evidence was weakened by a less than optimal choice of outcome. Further study of this association would be important. Hall and colleagues did not find any relationship between the short- and long-term outcomes. At the moment, we cannot tell whether the lack of association was because of the selected short-term outcomes (the long-term outcome measure was ideal) or because there is no biological reason for an association in the context of hip arthroplasty. The open arrows indicate that an association has not yet been shown. There was also no association between short-term biological changes and the long-term outcome. We suggest that there is no true association, except through short-term effects, and we have depicted our hypothesis with a dashed arrow. Therefore, in our model we have placed short-term outcomes as an intermediary between biological changes and long-term outcomes. Clearly, more research on these pathways is needed, particularly if interventions can be devised to improve outcome in both the short and the long term.
|
F. Carli
Department of Anesthesia
McGill University Health Centre
Montreal
Canada
N. Mayo
Department of Clinical Epidemiology
McGill University
Montreal
Canada
References
1 Hall GM, Peerbhoy D, Shenkin A, Parker CJR, Salmon P. The relationship of the functional recovery after hip arthroplasty to the neuroendocrine and inflammatory responses. Br J Anaesth 2001; 87: 53742
2 Cruickshank AM, Fraser WD, Burns HJG, Van Damme J, Shenkin A. Response of serum interleukin-6 in patients undergoing elective surgery of varying severity. Clin Sci 1990; 79: 1615[ISI][Medline]
3 Kehlet H. Surgical stress response: does endoscopic surgery confer an advantage? World J Surg 1999; 23: 8017
4 Peerbhoy D, Keane P, Maciver K, Shenkin A, Hall GM, Salmon P. The systematic assessment of short-term functional recovery after major joint arthroplasty. J Quality Clin Pract 1999; 19: 16571
5 Holden MK, Gill KM, Magliozzi MR. Gait assessment for neurologically impaired patients. Standards for outcome assessment. Phys Ther 1986; 66: 15309[ISI][Medline]
6 Butland RJ, Pang J, Gross ER, Woodcock AA, Geddes DM. Two-, six-, and 12-minute walking tests in respiratory disease. Br Med J Clin Res Ed 1982; 284: 16078[Medline]
7 Guyatt GH, Thompson PJ, Berman LB, et al. How should we measure function in patients with chronic heart and lung disease? J Chronic Dis 1985; 38: 51724[ISI][Medline]
8 Podsiadlo D, Richardson S. The timed Up & Go: a test of basic functional mobility for frail elderly persons. J Am Geriatr Soc 1991; 39: 1428[ISI][Medline]
9 Bellamy N, Buchanan WW, Goldsmith CH, Campbell J, Stitt LW. Validation study of WOMAC: a health status instrument for measuring clinically important patient relevant outcomes to antirheumatic drug therapy in patients with osteoarthritis of the hip or knee. J Rheumatol 1988; 15: 183340[ISI][Medline]