Abstract
Objective
We evaluated the association of parity to both risk of knee replacement (KR) and knee osteoarthritis (OA).
Design
The NIH-funded Multicenter Osteoarthritis Study (MOST) is a longitudinal observational study of persons age 50 to 79 years with either symptomatic knee OA or at elevated risk of disease. Baseline and 30-month knee radiographic OA (ROA) was defined as Kellgren/Lawrence (K/L) grade≥2 or KR. Women were grouped based by number of births: 0; 1 (reference group); 2; 3; 4; and 5 or more. We examined the relation of parity to the incidence over 30 months of ROA and KR using a Poisson regression model. Generalized estimating equations were used to control for correlation between two knees within a subject. We adjusted for age, BMI, race, education, occupation, baseline estrogen use, clinical site, injury, and for KR analyses WOMAC pain and use of pain medication.
Results
Among 1618 women who reported parity information, mean age was 62.6 years, mean BMI 30.7 kg/m2, mean WOMAC pain subscale score 3.7 at baseline. There were 115 KRs and 134 cases of incident knee ROA over 30 months. The relative risk of incident KR was 2.7 times as high (95% CI: 1.0, 7.3) and relative risk of incident knee ROA was 2.6 times as high (95% CI: 1.2, 5.3) among women with 5–12 children compared with those with one birth.
Conclusion
Parity in women at risk for OA is associated with both incident ROA and KR, particularly for those with more than 4 children.
Keywords: parity, knee, osteoarthritis, joint replacement
Introduction
Osteoarthritis (OA) is the most common cause of disability in the elderly, and has a higher prevalence in women as compared with men. The reasons for this remain unclear, although differences in hormonal milieu have been suspected (1, 2).
The most definitive treatment for knee pain is knee replacement (KR). In general, the decision to replace a knee rests on the severity of pain and limitations of function, and often surgeons and patients together make a joint decision on replacement. In a postal survey of orthopedists, most reported severe daily pain with radiographic joint space narrowing as the most common reason they performed knee replacements (3). Race, age and gender have all been identified as affecting the likelihood of a particular patient receiving a joint replacement (4), and thus likely confounding the relation between pain and knee replacement. Despite these prior findings, the complex interaction of physical disease characteristics, medical practice approaches, and cultural and individual choices that combine to lead to KR continues to require exploration.
Recently, a large prospective study of 1.3 million women from England revealed that both parity and hormone replacement therapy were associated with joint replacement. Specifically, each additional birth increased the relative risk of KR by 8% (5). However, the investigator did not examine the association of parity with radiographic OA (ROA), a major risk factor for KR.
In order to confirm an increased risk for KR with parity in the U.S., we evaluated the relation of number of births to incidence of KR among women participants in the Multicenter Osteoarthritis (MOST) Study. In addition, we assessed whether parity was associated with an increased incidence of ROA over 30 months.
Methods
The Multicenter Osteoarthritis Study (MOST) is an NIH-funded study of risk factors for individuals with or at risk of knee OA due to factors including obesity, knee pain, aching or stiffness on most of the previous 30 days, a history of knee injury or a history of knee surgery. 3026 subjects, age 50–79 years at enrollment were recruited from the areas surrounding Birmingham, Alabama and Iowa City, Iowa. A detailed description of the study population has been published previously (6).
Knee osteoarthritis was assessed at baseline and at 30 months using weight-bearing fixed flexion in positioning frame posteroanterior x-rays. Two experienced readers blinded to parity status interpreted these paired baseline and 30-months radiographs as part of a central reading process, with excellent reliability (weighted kappa for inter-rater reliability for Kellgren/Lawrence grade equal to 0.90) (7, 8). Disagreements were adjudicated by discussion and consensus between the two readers. In this study we defined tibiofemoral knee OA as Kellgren/Lawrence grade greater than or equal to 2.
KR was assessed by self-report at baseline and at 30-month visits. The self-report was confirmed by radiograph or by medical record documentation. Parity was assessed by self-report at the 30-month clinic visit with the question “how many children did you give birth to?”
Covariates used in this analysis were collected by questionnaire at baseline in the MOST study and were chosen because they might affect either ROA or risk of KR, along with estrogen use which the prior study by Liu et al had examined. These covariates included age, body mass index (BMI), race, educational attainment, occupation (skilled, unskilled/semi-skilled labor, sales, farming, technician, and worked in the home were classified as “labor”; office/clerical and office/professional were classified as “not labor”; all other occupations were classified as “other”), baseline Western Ontario McMaster (WOMAC) pain subscale scores, clinic site, use of estrogen for hormone replacement therapy at baseline (based on information collected in the medication inventory form: the participants “brought in or identified ALL prescription and over-the-counter medications that they took during the last 30 days”. Subjects with any use of estrogen during the prior 30 days were considered as users), pain medication use (again based on information collected in the medication inventory form: subjects with any use of analgesics, narcotic analgesics, COX II inhibitors, or NSAIDS during the prior 30 days were considered as users), and history of knee injury or surgery to the knee. BMI was collected during the first clinical assessment.
We examined the relation of number of births to prevalence of ROA and KR at baseline, as well as to incidence of ROA and KR over a 30-month follow up period. For the cross-sectional study, we included the knees of all women with baseline radiographs who also answered the questions about pregnancy. For the incidence study, we included the knees of all women who had readable radiographs at both baseline and 30 months.
The potential knees included in the joint replacement analysis are those which had no KR at the baseline examination and for which we have information at 30 months on whether a KR was performed in the interval. We defined baseline ROA as including baseline KR, and incident ROA as including incident KR. We defined incident radiographic OA as K/L≥2 at follow-up among knees that had K/L=0 or 1 at baseline.
Statistical Analysis
We divided parity into 6 categories: zero children; one child (report of one child served as the referent group in all analyses); two children; three children; four children; and 5 or more children. Using an analysis of variance for continuous variables and a chi-square test for categorical variables, we compared the characteristics of the participants according to parity categories. We examined the relation of parity to prevalence of baseline ROA and to incident ROA over 30 months and incident KR over 30 months with a log linear regression model using the Poisson assumption and robust variance estimates for KR and ROA, respectively. Generalized estimating equations (GEE) were used to account for the correlation between two knees within a person. In the regression model, we adjusted for age, BMI, race, educational level (in three levels: high school graduate or below, some college, college graduate or above), occupation (manual labor versus non-labor versus “other”), baseline WOMAC pain subscale scores, clinic site, estrogen use, history of knee injury or surgery. Age and BMI and WOMAC were entered into the regression model as continuous variables.
This analysis was in compliance with the Helsinki Declaration. The MOST Study was performed at clinical centers at University of Alabama, Birmingham, and University of Iowa with the Coordinating Center at University of California, San Francisco, the Analysis Center at Boston University, and with the first and last authors at University of California, Davis; the appropriate institutional review boards granted approval for the analysis and for the MOST study. All analyses were done using SAS version V9.2 (SAS Institute Inc., Cary, NC, USA) and analyses performed at the 5% significance level.
Results
Of 1820 women enrolled at baseline, 202 were excluded from this analysis because of missing information on parity (n=200, mostly because they were lost to follow-up so could not provide parity information at the second visit) or had rheumatoid arthritis (N=2). Total loss to follow up from baseline to 30 months in MOST was 198 participants, and all of these subjects were excluded from this analysis. The total number of subjects included in this analysis was 1618 women. The women included in this analysis had a mean age of 62.6 (SD 7.9), mean BMI of 30.7 (SD 6.3) and 83.6% were Caucasian. Participants overall had a mean of 2.5 children (SD=1.7). As shown in Table 1, women with 5–12 children had lower education levels, were more likely to be African American, and were more likely to be in labor occupations than women with fewer children. No significant differences were noted between parity categories for BMI or WOMAC pain levels. (See Tables 1 and 2). At baseline there were 45 prevalent KRs; over the 30 months there were 115 incident KRs reported, 113 of which were confirmed by radiograph or by medical documentation. 1244 knees had prevalent ROA at baseline. 129 knees developed incident ROA during 30-month follow-up.
Table 1.
Baseline characteristics among 1618 women by parity groups.
All (N=1618) | # of children
|
|||||||
---|---|---|---|---|---|---|---|---|
0 (N=210) | 1 (N=181) | 2 (N=514) | 3 (N=372) | 4 (N=191) | 5–12 (N=150) | p-value for comparison among # of children groups | ||
Age, mean (SD) | 62.6 (7.9) | 61.1 (8.2) | 60.9 (7.9) | 60.9 (7.4) | 63.4 (7.7) | 65.0 (7.1) | 67.5 (6.8) | <.0001 |
| ||||||||
BMI, mean (SD) | 30.7 (6.3) | 31.2 (6.5) | 30.9 (6.8) | 30.3 (6.2) | 30.7 (6.6) | 31.0 (5.7) | 31.2 (6.0) | 0.3798 |
| ||||||||
Race, N (%) | ||||||||
African American | 245 (15.1) | 22 (10.5) | 42 (23.2) | 83 (16.1) | 44 (11.8) | 31 (16.2) | 23 (15.3) | 0.0038 |
White | 1352 (83.6) | 186 (88.6) | 133 (73.5) | 426 (82.9) | 325 (87.4) | 156 (81.7) | 126 (84.0) | |
other | 21 (1.3) | 2 (0.9) | 6 (3.3) | 5 (1.0) | 3 (0.8) | 4 (2.1) | 1 (0.7) | |
| ||||||||
Educational Level, N(%) | ||||||||
High School or less | 475 (29.3) | 23 (22.7) | 41 (22.7) | 139 (27.0) | 128 (34.4) | 80 (41.9) | 64 (42.7) | |
Some College | 472 (29.2) | 69 (38.1) | 69 (38.1) | 154 (30.0) | 115 (30.9) | 42 (22.0) | 46 (30.7) | <.0001 |
College Graduate or Higher | 671 (41.5) | 71 (39.2) | 71 (39.2) | 221 (43.0) | 129 (34.7) | 69 (36.1) | 40 (26.7) | |
| ||||||||
Occupation, N (%) | ||||||||
Not labor | 774 (47.8) | 134 (63.8) | 100 (55.3) | 265 (51.6) | 149 (40.1) | 81 (42.4) | 45 (30.0) | |
Labor | 465 (28.7) | 29 (13.8) | 40 (22.1) | 120 (23.3) | 142 (38.2) | 68 (35.6) | 66 (44.0) | <.0001 |
Other | 379 (23.4) | 47 (22.4) | 41 (22.7) | 129 (25.1) | 81 (21.8) | 42 (22) | 39 (26.0) | |
| ||||||||
Pain medication use, N (%) | 1126 (69.6) | 132 (62.9) | 128 (70.7) | 355 (69.1) | 265 (71.2) | 140 (73.3) | 106 (70.7) | 0.2628 |
Table 2.
Baseline knee-based radiographic characteristics, history of knee injury or surgery and WOMAC knee pain.
All (N=1618) | # of children
|
|||||||
---|---|---|---|---|---|---|---|---|
0 (N=210) | 1 (N=181) | 2 (N=514) | 3 (N=372) | 4 (N=191) | 5–12 (N=150) | p-value for comparison among # of children groups | ||
K/L grade, (%) | ||||||||
0 | 43.5 | 44.3 | 49.4 | 46.6 | 41.4 | 41.5 | 32.9 | |
1 | 17.4 | 19.4 | 18.4 | 19.3 | 14.4 | 15.6 | 16.1 | 0.4456 |
2 | 15.5 | 16.0 | 12.6 | 13.8 | 18.5 | 17 | 15.1 | |
3–4 | 23.6 | 20.3 | 19.6 | 20.2 | 25.8 | 25.9 | 35.9 | |
| ||||||||
Knees with history of injury or surgery, (%) | 25.7 | 28.6 | 27.4 | 23.9 | 26.5 | 25.4 | 23.7 | 0.5095 |
| ||||||||
WOMAC knee pain subscale (0–20) | 3.7 (3.8) | 3.4 (3.6) | 3.7 (4.0) | 3.4 (3.7) | 4.0 (3.9) | 3.8 (4.0) | 3.7 (3.9) | 0.0731 |
*For the K/L grade listings, 53 knees had K/L grade missing, 45 had KR at baseline, 8 excluded from knee x-ray reading due to osteonecrosis or poor film quality
When we dichotomized simple baseline K/L grade by ≥2 vs. <2 and performed the Chi-square test, the p-value was significant for difference by parity (p<0.0001). We found a small but statistically significant increased prevalence of baseline ROA in all women with 3 or more births compared with those who had one birth (Table 3). We also found a statistically significant increased relative risk of incident ROA over 30 months in women with 5–12 births as compared with women with 1 birth, but no association for other levels of parity (Table 4). Very similar relative risk levels and significance levels were observed when baseline or incident ROA were evaluated without including KR (results not shown).
Table 3.
Parity and baseline prevalent radiographic knee OA, including KR.
Crude | Age, BMI, race-Adj. | Multi-Adj.* | |||||
---|---|---|---|---|---|---|---|
| |||||||
Number of Births | Total # Knees | # (%)With ROA | RR | RR | p-value | RR | p-value |
0 | 416 | 153 (36.78) | 1.1 (0.9, 1.4) | 1.2 (0.9, 1.5) | 0.2133 | 1.2 (0.9, 1.5) | 0.1952 |
1(ref) | 362 | 119 (32.87) | 1.0 | 1.0 | 1.0 | ||
2 | 1028 | 360 (35.02) | 1.1 (0.9, 1.3) | 1.2 (0.9, 1.4) | 0.1527 | 1.2 (1.0, 1.4) | 0.1177 |
3 | 742 | 335 (45.15) | 1.4 (1.1, 1.7) | 1.3 (1.1, 1.6) | 0.0052 | 1.3 (1.1, 1.6) | 0.0063 |
4 | 380 | 168 (44.21) | 1.3 (1.1, 1.7) | 1.2 (1.0, 1.5) | 0.0725 | 1.2 (1.0, 1.5) | 0.0770 |
5–12 | 300 | 154 (51.33) | 1.6 (1.2, 2.0) | 1.3 (1.0, 1.6) | 0.0405 | 1.3 (1.0, 1.6) | 0.0420 |
adjusted for age, BMI, race, education, occupation, injury/surgery, estrogen use, clinical site. KR was considered to be prevalent ROA.
Table 4.
Parity and incident radiographic knee OA (ROA) over 30 months, including new KR.
Crude | Age, BMI, race-Adj. | Multi-Adj.* | |||||
---|---|---|---|---|---|---|---|
| |||||||
Number of Births | Total # Knees | # (%)With ROA | RR | RR | p-value | RR | p-value |
0 | 263 | 18 (6.84) | 1.3 (0.6, 2.7) | 1.2 (0.6, 2.6) | 0.5628 | 1.3 (0.6, 2.7) | 0.5358 |
1(ref) | 241 | 13 (5.39) | 1.0 | 1.0 | 1.0 | ||
2 | 657 | 40 (6.09) | 1.1 (0.6, 2.2) | 1.2 (0.6, 2.4) | 0.5128 | 1.3 (0.7, 2.5) | 0.4087 |
3 | 407 | 29 (7.13) | 1.3 (0.7, 2.6) | 1.4 (0.7, 2.7) | 0.3523 | 1.5 (0.8, 3.0) | 0.2291 |
4 | 212 | 17 (8.02) | 1.5 (0.7, 3.2) | 1.5 (0.7, 3.1) | 0.2945 | 1.7 (0.8, 3.6) | 0.1459 |
5–12 | 143 | 17 (11.89) | 2.2 (1.1, 4.6) | 2.1 (1.0, 4.5) | 0.0489 | 2.6 (1.2, 5.3) | 0.0120 |
| |||||||
P for linear trend | 0.0258 |
adjusted for age, BMI, race, education, occupation, injury/surgery, estrogen use, clinical site. New KR was considered to be incident ROA.
We also found that women with 3 or 5–12 children were significantly more likely to have incident KR at 30 months, and it appeared that there was a linear trend in the data, where those with more children were more likely to have KR (see table 5). There was no apparent association between estrogen use and risk of KR (RR=1.3 (0.8, 2.1; p=0.3151).
Table 5.
Parity and 30-month incident KR.
Crude | Age, BMI, race-Adj. | Multi-Adj.* | |||||
---|---|---|---|---|---|---|---|
| |||||||
Number of Births | Total # Knees | # (%)With incident KR | RR | RR | p-value | RR | p-value |
0 | 417 | 6 (1.44) | 0.7 (0.2, 2.2) | 0.7 (0.2, 2.3) | 0.5540 | 0.9 (0.3, 2.9) | 0.8580 |
1(ref) | 358 | 7 (1.96) | 1.0 | 1.0 | 1.0 | ||
2 | 1013 | 33 (3.26) | 1.6 (0.7, 4.0) | 1.8 (0.7, 4.5) | 0.2398 | 2.0 (0.8, 5.1) | 0.1624 |
3 | 732 | 38 (5.19) | 2.6 (1.1, 6.4) | 2.4 (1.0, 6.1) | 0.0594 | 2.7 (1.1, 6.8) | 0.0375 |
4 | 373 | 14 (3.75) | 1.9 (0.7, 5.2) | 1.7 (0.6, 4.8) | 0.3199 | 1.8 (0.6, 5.3) | 0.2590 |
5–12 | 298 | 17 (5.70) | 2.8 (1.1, 7.5) | 2.2 (0.8, 6.2) | 0.1299 | 2.7 (1.0, 7.3) | 0.0494 |
| |||||||
P for linear trend | 0.0070 |
adjusted for age, BMI, race, education, occupation, WOMAC knee pain, pain medication, estrogen use, clinical site
Discussion
We confirmed an association between parity and incident KR, and found an association between parity and incident and prevalent ROA. The association between parity and risk of KR may be attributable to osteoarthritis status, changes in weight and body composition, social, neurological, hormonal or other factors.
Liu et al. found an 8% per birth increased relative risk for incident knee replacement over a mean of 6.1 years. Liu utilized the Million Women Study, a prospective study of 1.3 million middle-aged women recruited in 1996–2001 through the United Kingdom National Health Service. Liu assigned the nulliparous group as the reference group and then divided the cohort into levels of parity of 1, 2, 3, and “4 or more” births, finding an increased risk in each of those groups compared with nulliparous women. They also examined a variety of other factors related to hormonal milieu, including oral contraceptive use, hormone replacement therapy, age at menarche and age at menopause, and found a significant association only with current hormone therapy use and age at menarche. However, due to limited covariate information, they were not able to address whether parity is associated with increased risk of osteoarthritis itself, the primary diagnosis engendering KR, and they were not able to adjust for potentially important covariates, such as occupation and history of injury to the joint (5). In the MOST, a longitudinal study of risk factors for knee OA, we collected additional data including careful radiologic evaluation of all knees, history of joint injuries, and occupation.
The specifics of the MOST population indicate that the associations we have identified between parity and OA may or may not hold true in a more general population. It is of some interest that Liu et al identified an association with KR in a general population, but it may be that the associations in the two different populations reflect different mechanisms by which pregnancy increases relative risk.
Jorgensen et al. recently reported that an increasing numbers of live births reported by both men and women was associated with increased risk of “first OA hospitalization” in a Danish cohort using International Classification of Diseases (ICD) codes for their definition of OA (9). Physician clinical diagnosis assessed by ICD codes carries the potential for misclassification bias. The direct diagnosis of radiographic knee OA and KR in the current study avoids this type of bias. In addition, the effect was observed in both men and women in the Danish cohort, which may suggest that childrearing factors (carrying children, etc.) are an important contributor to the increased OA risk; unfortunately we do not have information in MOST on number of children for men.
Wei et al. performed a cross-sectional study of 489 women and found that increased number of live or still births was associated with decreased total knee cartilage volume on MRI but not with joint space narrowing or osteophytosis on radiograph in adjusted analyses (10). They also found “no association between parity and change in either cartilage volume or cartilage defects over 2.7 years”. The authors stated “this might be expected given that pregnancy was an exposure in the relatively distant past”. Although our measures of OA, using radiographs, are obviously different from cartilage volume measurements on MRI, it is interesting that they found no association with cartilage change over time, where we found significant evidence of incident OA disease over a similar timespan. It is possible that this is due to smaller numbers of subjects in Wei’s study, or due to a different population in Tasmania, Australia which may have different genetic risks and different cultural approaches to occupation or childrearing. It is also worth noting that the MOST women were recruited specifically because they either had knee OA or were at high risk of the disease, where the Tasmanian subjects were recruited from a general health study cohort.
A relationship between parity and KR might be due to increased risk of osteoarthritis. The fact that we found a significant association with incident and prevalent ROA supports the idea that at least part of the increased risk of KR may be related to the observed increase in radiographic disease. The similarity of effect size magnitudes in the associations between parity and ROA and between parity and KR suggest that ROA is an important and perhaps dominant factor, perhaps more important than social factors related to community and family size, exposure to community members who have had joint replacements, differences in willingness to consider surgery, and other such factors. It is unclear why the association for the group with 4 children did not reach significance while the 3 and 5–12 children groups were significant, but it may be due to variations inherent in the relatively small outcome numbers in this analysis
The relationship between parity and incident ROA is interesting, and suggests that there may be biological events during pregnancy or childrearing which do not manifest as incident disease until many decades after the pregnancies. These may include damage to cartilage due to extra weight carried or increased “clumsiness” during pregnancy due to the temporary alteration of weight distribution, which may predispose to later disease without overtly causing it at the time of pregnancy. Alternatively, rather than direct damage to joints that occurs during pregnancy, it is possible that permanent changes occur in the body during pregnancy which predispose to adverse joint health in the decades following reproductive life. These changes may include additive weight retention after pregnancy (11) or hormonal changes or other factors, perhaps acting in a “multi-hit” model. Although we adjusted for BMI in the current study, it was not possible to adjust for BMI prior to entering the MOST study. It is also worth noting that despite some controversy, there is evidence that excess weight gain may occur after first pregnancy but not after further pregnancies (12–16); our finding of increasing risk of ROA with increasing numbers of births suggests that weight gain after pregnancy is not sufficient to explain the patterns of ROA we see. It is also possible that caring for children causes ongoing physical insults to the joints during childrearing years, which again may only manifest years later. In this sense, pregnancies might be considered to be a period of “injury” (due to transient increased weight, hormonal changes, etc) similar, for example, to ligamental or meniscal injuries which predispose to the disease of OA later in life. Other potential sources of residual confounding, such as differential occupational exposures by parity group, are also possible.
A number of earlier studies found no association between knee OA and parity or pregnancy (17–19). However, these studies differed in significant ways from ours. Dawson et al. examined the number of full term pregnancies, but had relatively small numbers of subjects and assigned parity levels differently, dividing into 0–2, 3, and >3 (18). Samanta et al. did not have knee OA as a specific outcome but instead used “large joint OA” as an outcome and “ever-pregnant” or “live birth” as the risk factor, finding no association (19). Anderson and Felson found a non-significant 5% per birth increased risk of radiographic osteoarthritis of the knee in the NHANES I dataset in an age-adjusted analysis, possibly due to the fact that non-weight-bearing AP knee films were used in that study (17). It is not entirely clear why each of these earlier studies failed to find an association where our study did, but it is likely because of differences in variable definition or collection, or in the populations studied.
It is important to consider that the question asked of the women in MOST was “how many children did you give birth to” rather than, for example, “how many times were you pregnant?” In Liu’s study, she found an increased risk of joint replacement with hormone replacement therapy, a finding which we did not see in our dataset; this could be due to differences in the subject populations in terms of age or other factors, differences between “hormone replacement therapy” (as defined in Liu) and “estrogen use” (as defined in MOST), or simply a smaller sample size in the MOST Study. However, there is no information about age at menarche and menopause, breastfeeding history, number of pregnancies, and other hormonal factors in the MOST dataset, and exploring these factors might potentially elucidate our findings.
Some readers may be interested in how our results would have differed if we had used nulliparous women as the referent group. We have redone the analysis using no children as referent point and we provide the results as online appendix Tables A1–A3 for your inspection. The same general patterns hold in this alternate analysis, but because women with zero children systematically have slightly higher risk of ROA and of KR than do women with one child, some of the results become statistically insignificant. Given known and likely unknown differences between women with zero births and all other women, we have presented the analysis using women with one birth as referent as our core results and we believe this serves to highlight the patterns of relative risk.
Given that some women were missing information on parity, we considered whether missing information on parity may potentially bias the association of parity and incident ROA. Since women who did not come to the 30 month visit had no information on parity, we took an indirect approach by using all women who do have parity data available at the 30-month visit and testing if parity is associated with loss to follow-up at the 60-month visit. After adjusting for age, BMI, race and clinic site, compared with nulliparous women, the relative risk of loss to follow-up for those with 1, 2, 3, 4, and 5–12 children were 1.1, 1.1, 1.2, 1.0, and 1.2, respectively (all p values > 0.50) (see Table A4 in the online appendix). These findings indicate that parity is not significantly associated with risk of loss to follow-up. Thus, we believe that missing information may have only a limited impact on the effect estimates, although these findings provide only indirect evidence on this potential bias.
Perhaps the most important limitation of our study is that the MOST cohort may not be representative of the general population of women who experience pregnancy due to the inclusion criteria employed for the study, and therefore our findings may not be as generalizable as some larger cohorts. Although it is one of the largest comprehensive prospective cohorts organized around questions of osteoarthritis, there are still only a limited number of the outcomes, which may have prevented reaching significance for some of the questions. Some parity groups had small numbers of outcomes. 30 months is a relatively short time for observing for incident OA outcomes.
There are also important strengths in our study, which include a large longitudinal dataset of over 1800 women in which risk factors, radiographs and information on joint replacement has been obtained in a standardized, comprehensive manner. The radiographs have been read by central readers in this cohort and have excellent reliability. Loss to follow-up has been extremely low in this cohort, so the incident results are robust and subject to relatively little of this type of bias.
In summary, increasing numbers of births in women with or at high risk for knee OA is associated with increased risk of prevalent ROA and incident KR and ROA. The observation of new cases of ROA late in life associated with pregnancy earlier in life suggests that the effect of pregnancy or children continues to express itself decades after the pregnancy and may reflect a model of joint injury engendering OA in later years.
Supplementary Material
Acknowledgments
We would like to acknowledge the MOST Study participants and staff as well as the funding sources. NIH grant numbers for the MOST project are: Felson –UO1 AG18820; Torner – UO1 AG18832; Lewis – UO1 AG18947; Nevitt – UO1 AG19069. Dr. Lane’s work is supported by NIH grants K24-AR048841; P50-AR060752; and R01-AR052000. Dr. Wise’s work is supported by the UC Davis Building Interdisciplinary Research Careers in Women’s Health Program – NIH K12HD051958; P50-AR063043.
Footnotes
Author Contributions:
Concept and Design: BLW, JN, YZ, NEL
Acquisition of Data: MN, DTF
Analysis and Interpretation of Data: BLW, JN, YZ, LB, NS, JK, MN, NEL
Preparation and Revision of Manuscript: BLW, JN, YZ, LB, NS, JK,MN, NEL
Final Approval of Submitted Version: BLW, JN, YZ, DTF, LB, NS, JK, MN, NEL
Competing Interests:
Dr’s Wise and Zhang have received research funding from Pfizer, Inc. for work unrelated to the present manuscript.
Publisher's Disclaimer: This is a PDF file of an unedited manuscript that has been accepted for publication. As a service to our customers we are providing this early version of the manuscript. The manuscript will undergo copyediting, typesetting, and review of the resulting proof before it is published in its final citable form. Please note that during the production process errors may be discovered which could affect the content, and all legal disclaimers that apply to the journal pertain.
Contributor Information
Barton L Wise, University of California, Davis School of Medicine.
Jingbo Niu, Email: niujp@bu.edu, Boston University.
Yuqing Zhang, Email: yuqing@bu.edu, Boston University.
David Felson, Email: dfelson@bu.edu, Boston University.
Laurence A. Bradley, Email: Braddog@uab.edu, University of Alabama, Birmingham.
Neil Segal, Email: neil-segal@uiowa.edu, University of Iowa School of Medicine.
Julie Keysor, Email: jkeysor@bu.edu, Boston University.
Michael Nevitt, Email: MNevitt@psg.ucsf.edu, University of California, San Francisco.
Nancy E Lane, Email: nancy.lane@ucdmc.ucdavis.edu, University of California, Davis School of Medicine.
References
- 1.Nevitt MC, Felson DT. Sex hormones and the risk of osteoarthritis in women: epidemiological evidence. Annals of the rheumatic diseases. 1996;55(9):673–6. doi: 10.1136/ard.55.9.673. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 2.Silman AJ, Newman J. Obstetric and gynaecological factors in susceptibility to peripheral joint osteoarthritis. Annals of the rheumatic diseases. 1996;55(9):671–3. doi: 10.1136/ard.55.9.671. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 3.Mancuso CA, Ranawat CS, Esdaile JM, Johanson NA, Charlson ME. Indications for total hip and total knee arthroplasties. Results of orthopaedic surveys. The Journal of arthroplasty. 1996;11(1):34–46. doi: 10.1016/s0883-5403(96)80159-8. [DOI] [PubMed] [Google Scholar]
- 4.Katz BP, Freund DA, Heck DA, Dittus RS, Paul JE, Wright J, et al. Demographic variation in the rate of knee replacement: a multi-year analysis. Health services research. 1996;31(2):125–40. [PMC free article] [PubMed] [Google Scholar]
- 5.Liu B, Balkwill A, Cooper C, Roddam A, Brown A, Beral V. Reproductive history, hormonal factors and the incidence of hip and knee replacement for osteoarthritis in middle-aged women. Annals of the rheumatic diseases. 2009 doi: 10.1136/ard.2008.095653. [DOI] [PubMed] [Google Scholar]
- 6.Felson DT, Niu J, Guermazi A, Roemer F, Aliabadi P, Clancy M, et al. Correlation of the development of knee pain with enlarging bone marrow lesions on magnetic resonance imaging. Arthritis Rheum. 2007;56(9):2986–92. doi: 10.1002/art.22851. [DOI] [PubMed] [Google Scholar]
- 7.Felson DT, Nevitt MC, Yang M, Clancy M, Niu J, Torner JC, et al. A new approach yields high rates of radiographic progression in knee osteoarthritis. J Rheumatol. 2008;35(10):2047–54. [PMC free article] [PubMed] [Google Scholar]
- 8.Neogi T, Felson D, Niu J, Nevitt M, Lewis CE, Aliabadi P, et al. Association between radiographic features of knee osteoarthritis and pain: results from two cohort studies. BMJ. 2009;339:b2844. doi: 10.1136/bmj.b2844. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 9.Jorgensen KT, Pedersen BV, Nielsen NM, Hansen AV, Jacobsen S, Frisch M. Socio-demographic factors, reproductive history and risk of osteoarthritis in a cohort of 4.6 million Danish women and men. Osteoarthritis Cartilage. 2011;19(10):1176–82. doi: 10.1016/j.joca.2011.07.009. [DOI] [PubMed] [Google Scholar]
- 10.Wei S, Venn A, Ding C, Martel-Pelletier J, Pelletier JP, Abram F, et al. The associations between parity, other reproductive factors and cartilage in women aged 50–80 years. Osteoarthritis Cartilage. 2011;19(11):1307–13. doi: 10.1016/j.joca.2011.07.020. [DOI] [PubMed] [Google Scholar]
- 11.Williamson DF, Madans J, Pamuk E, Flegal KM, Kendrick JS, Serdula MK. A prospective study of childbearing and 10-year weight gain in US white women 25 to 45 years of age. Int J Obes Relat Metab Disord. 1994;18(8):561–9. [PubMed] [Google Scholar]
- 12.Boardley DJ, Sargent RG, Coker AL, Hussey JR, Sharpe PA. The relationship between diet, activity, and other factors, and postpartum weight change by race. Obstet Gynecol. 1995;86(5):834–8. doi: 10.1016/0029-7844(95)00283-W. [DOI] [PubMed] [Google Scholar]
- 13.Gunderson EP, Abrams B. Epidemiology of gestational weight gain and body weight changes after pregnancy. Epidemiol Rev. 2000;22(2):261–74. doi: 10.1093/oxfordjournals.epirev.a018038. [DOI] [PubMed] [Google Scholar]
- 14.Gunderson EP, Murtaugh MA, Lewis CE, Quesenberry CP, West DS, Sidney S. Excess gains in weight and waist circumference associated with childbearing: The Coronary Artery Risk Development in Young Adults Study (CARDIA) Int J Obes Relat Metab Disord. 2004;28(4):525–35. doi: 10.1038/sj.ijo.0802551. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 15.Gunderson EP, Quesenberry CP, Jr, Lewis CE, Tsai AL, Sternfeld B, Smith West D, et al. Development of overweight associated with childbearing depends on smoking habit: The Coronary Artery Risk Development in Young Adults (CARDIA) Study. Obes Res. 2004;12(12):2041–53. doi: 10.1038/oby.2004.255. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 16.Smith DE, Lewis CE, Caveny JL, Perkins LL, Burke GL, Bild DE. Longitudinal changes in adiposity associated with pregnancy. The CARDIA Study. Coronary Artery Risk Development in Young Adults Study. JAMA. 1994;271(22):1747–51. [PubMed] [Google Scholar]
- 17.Anderson JJ, Felson DT. Factors associated with osteoarthritis of the knee in the first national Health and Nutrition Examination Survey (HANES I). Evidence for an association with overweight, race, and physical demands of work. Am J Epidemiol. 1988;128(1):179–89. doi: 10.1093/oxfordjournals.aje.a114939. [DOI] [PubMed] [Google Scholar]
- 18.Dawson J, Juszczak E, Thorogood M, Marks SA, Dodd C, Fitzpatrick R. An investigation of risk factors for symptomatic osteoarthritis of the knee in women using a life course approach. J Epidemiol Community Health. 2003;57(10):823–30. doi: 10.1136/jech.57.10.823. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 19.Samanta A, Jones A, Regan M, Wilson S, Doherty M. Is osteoarthritis in women affected by hormonal changes or smoking? Br J Rheumatol. 1993;32(5):366–70. doi: 10.1093/rheumatology/32.5.366. [DOI] [PubMed] [Google Scholar]
Associated Data
This section collects any data citations, data availability statements, or supplementary materials included in this article.