Article Text
Abstract
Objective: To examine the association between body mass index (BMI) and osteoarthritis (OA) leading to total hip (THR) or knee (TKR) joint replacement.
Methods: Case–control study design. All patients still living in Iceland who had had a THR or TKR resulting from OA before the end of 2002 were invited to participate. First-degree relatives of participating patients served as controls. A total of 1473 patients (872 women) and 1103 controls (599 women), all born between 1910 and 1939 and who had answered a questionnaire including questions about height and weight, were analysed. A randomly selected sample, representative of the Icelandic population, was used as a secondary control group.
Results: The OR, adjusted for age, occupation and presence of hand OA, for having a THR was 1.1 (95% CI 0.9 to 1.5) for overweight men and 1.7 (95% CI 1.0 to 2.9) for obese men. The OR for having a TKR was 1.7 (95% CI 1.1 to 2.6) for overweight men and 5.3 (95% CI 2.8 to 10.1) for obese men. The OR for having a THR was 1.0 (95% CI 0.8 to 1.3) for overweight women and 1.0 (95% CI 0.6 to 1.5) for obese women. The OR for having a TKR was 1.6 (95% CI 1.1 to 2.2) for overweight women and 4.0 (95% CI 2.6 to 6.1) for obese women.
Conclusion: This study supports a positive association between high BMI and TKR in both sexes, but for THR the association with BMI seems to be weaker, and possibly negligible for women.
Statistics from Altmetric.com
The relationship between high body mass index (BMI) and knee osteoarthritis (OA) is well established.1 It has been shown that high BMI precedes knee OA2 3 and that high BMI is also a risk factor for progression of established knee OA.4 This relationship appears stronger for women than for men.2 3 5 The relationship between high BMI and hip OA is more controversial. Several studies have indicated that there is no relationship between BMI and hip OA,6–13 others that there is.14–21 A review of the literature until 2000 found that there was moderate evidence for a positive association between high BMI and hip OA,22 and that studies based on clinical symptoms of OA were more likely to find a positive relationship than studies based on radiological diagnosis of OA.22 23
Previous studies varied in design. Some included only one gender and some included a specific subset, such as a certain occupational category.11 14 18 20 24 25 Studies also varied in their definition of OA, some using clinical symptoms and others using radiological OA or total hip replacement (THR) and/or total knee replacement (TKR). Each definition has its pros and cons. The definition of clinical or radiological OA can vary between studies, making comparison difficult, and indications for arthroplasty can vary between orthopaedic surgeons.
The purpose of this study was to explore the relationship between BMI and hip and knee OA leading to arthroplasty in Icelandic men and women. Therefore, using a case–control study design, we compared 1473 patients, who had undergone hip and or knee arthroplasty because of OA, with relatives of the same age who had not had THR or TKR. The study is population based, with subjects recruited from all cases of THR and TKR in Iceland. The primary controls were all relatives of the patients, thus coming from the same environment and genetic pool. In addition, we used an age-matched sample randomly drawn from the Icelandic population to ascertain the validity of our primary controls.
METHODS
The study was granted ethical approval by the ethics committee of Akureyri Central Hospital, where the study was based.
Our cases were patients who had received a THR or TKR because of OA. In a clinical setting, an orthopaedic surgeon had decided from the patient’s symptoms and radiological findings that the severity of the disease warranted prosthetic surgery. We identified, using computerised hospital records, all patients with a diagnosis of OA who had undergone THR and/or TKR during the period from the introduction of THR and TKR surgery in Iceland in 1967 until the end of 2002.26 27 By cross-referencing with the National Census and Death Register, we ascertained if the patient was still alive and their current residential address. We identified 4215 patients who had undergone THR, TKR, or both in the relevant period. Of these, 1289 patients were deceased at the start of the study in 1998. A further 384 patients identified as living at the start of our study died before being invited to participate. Patients who had moved abroad (n = 4) had only the country of residence registered and were therefore not traceable.
Thus, we had 2538 eligible cases, 1843 of whom agreed to participate in the study (72.6%). After reviewing the individual patient records, we excluded those who had surgery for the following reasons, which were incorrectly recorded as OA: inflammatory arthritis including ankylosing spondylitis, Legg–Calvé–Perthes disease, physiolysis of the femoral head, hip dysplasia, fracture. We selected only patients born between 1910 and 1939, who had adequately completed our questionnaire, to become our cases. This left us with 1473 participating index cases, of which 601 were men and 872 were women (fig 1). All data were collected retrospectively after surgery. For 75% of the cases for whom we have data on surgical date available, the median time to follow-up was 5 years (range 0–26).
All patients were asked to supply information about their first-degree relatives, and these were contacted and asked to participate in the study as controls. As for the patients, we included only relatives born between 1910 and 1939 who had completed the questionnaire, leaving 504 men and 599 women as our primary controls. The patients and controls were thus of similar age and from the same genetic pool, but were not matched on an individual basis.
Patients and controls gave written informed consent and answered a standardised questionnaire. The questionnaire contained 79 questions and asked for the patient’s height and weight, general health status, occupation, family history, physical activities, previous injuries and a detailed description of all musculoskeletal symptoms. The hospital records of all cases were reviewed to confirm that the diagnosis was correct. If x-ray films were available, they were also examined.
The presence of hand OA was determined by a trained research nurse or an orthopaedic surgeon at a clinic visit. All participants with classical symptoms of hand joint pain and stiffness and two or more Heberden’s nodes on either hand were defined as having hand OA.28
The Age, Gene, Environment, Susceptibility (AGES) Study of the Icelandic Heart Association has compiled reference data on members of the Icelandic population who have had height and weight measured.25 We used data from the AGES Study collected in 2003 on 2269 people (1301 women) born between 1910 and 1939, randomly selected from the Icelandic Census Register, to ascertain the validity of our primary control group.29 By comparing the self-reported values of our primary control group with the measured values of our secondary control group, we were able to estimate the validity of self-reported BMI. By the end of 2002, the Icelandic population was 288 471 (144 287 men and 144 184 women). A total of 16 988 men and 20 083 women born between 1910 and 1939 were alive at the end of 2002.
Information on the occupations of the Icelandic population was obtained from Statistics Iceland, the national statistical bureau (Hagstofan). The occupations of patients and controls were classified according to the Icelandic translation of the ISCO-88 (International Standard Classification of Occupations 1988) published by The International Labour Organization (ILO).30 Several occupations included only a handful of people. We therefore categorised the occupation of our study subjects according to physical load into sedentary (classes 11–21 and 41), light work (classes 22–34 and 42), moderate work (classes 51, 52, 73 and 81–91) and heavy work (classes 61–72, 74, 92, 93).
We performed separate analyses for men and women and for THR and TKR. Statistical comparisons between groups were carried out using the independent samples t test. We calculated odds ratios (ORs) using logistic regression in a model adjusted for age, occupation and the presence of hand OA. All analyses were carried out at the level of the person. As the group of patients with both THR and TKR was small, we did not include these in the final adjusted analysis. We considered p⩽0.05 to be significant, and all tests were two-tailed (SPSS V15.0).
RESULTS
Characteristics of patients and controls
The mean age of patients—that is, subjects who had undergone THR or TKR—was slightly higher than the mean age of primary controls. BMI was also higher in all patient groups than their controls except for women who had undergone THR (table 1). For women, occupation classes were similarly distributed among cases and primary controls. However, for men this distribution was not even, patients with THR and TKR being over-represented in the heavy work class (p = 0.006).
THR or TKR and BMI in men
Mean BMI was 0.5 kg/m2 higher (95% CI 0.1 to 0.9) in the THR group than in the control group and 1.3 kg/m2 higher (95% CI 0.7 to 1.9) in the TKR group than in the control group. For those who had undergone both THR and TKR, the mean BMI was 1.6 kg/m2 higher (95% CI 0.7 to 2.5) than in the control group. Men with TKR had significantly higher BMI than men with THR, adjusted for age, occupation and the presence of hand OA (p = 0.01).
We also performed sex-stratified multivariable logistic regression, with case–control status as the dependent variable adjusted for age, occupation and hand OA. To facilitate the interpretation of our findings, we divided patients and controls according to the World Health Organization (WHO) classification into underweight (BMI<18.5 kg/m2), normal weight (BMI 18.5–24.9 kg/m2), overweight (BMI 25.0–29.9 kg/m2) and obese (BMI 30.0 kg/m2 or above), and used the “normal” group as the reference category. We found that the OR for having a THR for overweight men was 1.1, but this was not significant when 95% CI (0.9 to 1.5) was taken into account. The OR for obese men was 1.7 (95% CI 1.0 to 2.9) (fig 2). The OR for having a TKR was significant for both overweight men (OR 1.7, 95% CI 1.1 to 2.6) and obese men (OR 5.3, 95% CI 2.8 to 10.1) (fig 3).
THR or TKR and BMI in women
The mean BMI of the THR group did not differ from that of the primary control group (mean difference 0.0 (95% CI −0.5 to 0.5)), whereas that of the TKR group was 1.7 kg/m2 higher (95% CI 1.1 to 2.3) than that of the control group. In those who had both THR and TKR, the mean BMI was 0.8 kg/m2 higher (95% CI −0.1 to 1.7) than that of the controls. Women with TKR had significantly higher BMI than women with THR, adjusted for age, occupation, and the presence of hand OA (p<0.001).
In the same way as for the men, we used a logistic regression model with adjustment for age, occupation and hand OA. We found that the OR for having a THR was 1.0 (95% CI 0.8 to 1.3) for overweight women and 1.0 (95% CI 0.6 to 1.5) for obese women (fig 2). Thus, neither overweight nor obesity in women was associated with THR. The OR for TKR was significant for both overweight (OR 1.6, 95% CI 1.1 to 2.2) and obese (OR 4.0, 95% CI 2.6 to 6.1) women (fig 3).
DISCUSSION
This study was performed to investigate the relationship between BMI and THR and TKR in Icelandic men and women. For the knee, our results corroborate conclusions from previous reports—that is, a positive association between high BMI and subsequent TKR in both sexes.
The relationship between high BMI and subsequent THR is more controversial, some studies claiming that there is a positive relationship, others that there is none. Our findings indicate that both may be right to some extent. It has been previously published that BMI is a stronger risk factor for knee OA for women than for men,2 3 5 which is also supported by our data, but our findings indicate that for hips, BMI is a stronger risk factor for men than women. We have found no previous reports stating this. There may be anatomical, mechanical and/or biochemical reasons for these different effects of high BMI on knee and hip OA leading to joint replacement. It has been proposed that the gender difference in the effect of BMI on the knee is due to the difference in anatomical structure. Joint alignment and BMI interact as risk factors for the knee,31 and such an interaction is perhaps not present for the hip joint.
One challenge with case–control studies is the recruitment of appropriate controls. The controls should optimally be drawn from the same source population as the cases. Further, to avoid bias, it is also important that the data for cases and controls are gathered in the same way and at the same time. The Icelandic Heart Association has been following the development of BMI in Iceland over the past few decades and has shown that BMI tends to increase with age and has also been increasing with each generation—that is, people born in 1930 had a higher BMI at the age of 70 than people born in 1920. We evaluated cases and controls at a similar point in time. We also recruited relatives of the cases as controls, which is advantageous in that we achieve a larger degree of genetic homogeneity.
We believe that the reason for the discrepancy in mean BMI between our primary controls and our secondary controls (taken from the AGES Study) is that data on weight and height for the present study were based on self-report, whereas in the AGES Study these variables were measured. It has been shown that people underestimate their weight, but that the information is nevertheless reliable and valid.32–34 Our primary controls had a mean BMI that was about 95% of that of the AGES population reference group. This is the same degree of underestimation as previously published, which validates our primary control group.33
There are several limitations to our study. Participation was voluntary, and those who did not participate might differ from those who did. Even though the response rate was high (72.6%), non-response bias may still have an influence, although its effect cannot be assessed accurately. High BMI correlates with higher mortality, which may bias our results towards a null finding for all subgroups, except for women with hip OA, where no such bias is expected because of similar BMI to controls. Height and weight was estimated by self-report in the same time period but a variable time after the joint replacement surgery. It is possible that our patients on average may have gained weight after the onset of symptomatic OA, more than they would have gained if no symptomatic OA had occurred, introducing bias into the finding of an association between high BMI and joint replacement surgery for OA. It is likely that the most obese subjects with symptomatic OA were never offered joint replacement surgery (because of being high-risk candidates), biasing the results towards the null (assuming there is a positive association between high BMI and joint replacement surgery). By the use of controls recruited from relatives of our patients, it is possible that the BMI of the controls was more similar to that of the cases than would be expected by purely random control sampling from the general population. This may have biased our estimates of association towards the null, but the strength of such bias, if any, is unpredictable. Importantly, we would expect that any such bias would act similarly on results for men and women. Thus, it does not explain why we found a weaker association between BMI and THR for women in particular. The direction of the sum of all biases and its strength is difficult to ascertain, but our positive finding of an association between TKR and high BMI in both men and women, in line with previous reports, supports the validity of our methodology and findings.
In this study we found that high BMI is a risk for knee OA leading to TKR for both sexes. However, we found that high BMI was a risk factor for THR only in obese men, and not in overweight or obese women.
Acknowledgments
We thank Jan-Åke Nilsson for statistical guidance. We also thank the Icelandic Heart Association for giving us access to their research material.
REFERENCES
Footnotes
Competing interests: None.
Funding: Supported by: Scientific Foundation of Akureyri Central Hospital, The Swedish Research Council (medicine), Lund Medical Faculty and University Hospital, the King Gustaf V 80-year Fund, The Swedish Rheumatism Association and the Kock Foundations.