Age is perhaps one of the most important confounder that none of us can escape. The image is CC by Sara.
Age is an important confounder in studying most health related outcomes [1, s 5], and perhaps the most commonly adjusted variable. In this and next post I will go into (1) what we know about the age effect in relation to total hip replacements (THR) re-operations and mortality, (2) what I found in my study on age and health-related quality of life (HRQoL) using splines, and (3) how I implemented and evaluated different splines using R for this study.
Re-operations
Increased re-operation rates for younger patients due to aseptic loosening is today widely recognized [2]–[5], [6, s 79], [7], [8]. On the other hand, the risk of fractures and dislocations is higher among elderly. Cook et al. showed that patients age 80 and above have an odds ratio of 4.4 (95% CI, 2.9–6.4) for fractures compared to patients below 80 [9]. Malkani et al. reported that patients 85 and above were associated with an almost 50% higher relative risk of early dislocation compared to patients between 65 and 69 [10].
Mortality
As expected, age is also related to mortality within 90 days. SooHoo et al. reported an odds ratio of 2.6 (95% CI, 2.2 to 3.0) for patients over the age of 75 compared to patients between 65 and 75 [11]. Hunt et al. studied over 400,000 patients and similarly reported an increase in the elderly, although they noted that this increase was primarily among men. The Kaplan-Meier 90-day mortality estimates for patients age 80 and above were 1.9% for men (95% CI, 1,7 to 2•11) vs 1.1% for women (95% CI, 1,0 to 1.2); the corresponding rates for patients between 65 and 69 were 0.3% for men (95% CI, 03 to 0.4) vs 0.2% for women (95% CI, 0.1 to 0.2) [12].
Health-related quality of life
Surprisingly, prior to our study [13] there was no consensus regarding the impact of age on HRQoL after THR [14]. Some studies report decreasing improvement in HRQoL after THR with age [15]–[19], while others indicate little effect if any [20]–[24]. None of these studies used splines, thus potentially missing out on more complex relations.
Our study was based on a nation-wide cohort of almost 30,000 patients operated due to primary osteoarthritis between 2008 and 2010. For estimating HRQoL, we used the generic health outcome questionnaire EQ-5D of the EuroQol group that consists of 2 parts: the EQ-5D index and the EQ VAS estimates.
We found that age had little effect until patient’s late 60:s, after which it had a negative impact on both HRQoL outcomes. Below image illustrates the effect:
The relationship between the EQ-5D index and the EQ VAS one year postoperatively and the patient’s age at surgery. Preoperative EQ-5D index and EQ VAS were set to the most frequently occurring values (index = 0.87; VAS = 50) and are indicated by the horizontal dashed lines. The change before and after surgery is the height above this line, i.e. anything above is an improvement. The 2 lines differ only in height; the solid line with blue confidence interval indicates the optimal combination of covariates (male sex, first hip, and Charnley class A) while the dotted line with pink confidence interval indicates the least favorable combination (female sex, previous contralateral hip surgery, and Charnley class C). The pain VAS was set to the median value, 65 mm.
The effect can also be illustrated using a contrast-table:
| Table 3: Age estimates with 70 years as reference. | |||
| Age (years) | Estimate | 95% CI | P-value |
|---|---|---|---|
| EQ-5D index | |||
| 40 vs. 70 years | 0.011 | 0.003 – 0.018 | 0.004 |
| 50 vs. 70 years | 0.009 | 0.004 to 0.013 | < 0.001 |
| 60 vs. 70 years | 0.007 | 0.005 to 0.009 | < 0.001 |
| 70 vs. 70 years | Ref. | – | – |
| 80 vs. 70 years | -0.019 | -0.022 to -0.017 | < 0.001 |
| 90 vs. 70 years | -0.042 | -0.048 to -0.037 | < 0.001 |
| EQ VAS | |||
| 40 vs. 70 years | 1.2 | -0.1 to 2.4 | 0.078 |
| 50 vs. 70 years | 1.2 | 0.4 to 2.0 | 0.002 |
| 60 vs. 70 years | 1.3 | 1.0 to 1.6 | < 0.001 |
| 70 vs. 70 years | Ref. | – | – |
| 80 vs. 70 years | -4.2 | -4.7 to -3.8 | < 0.001 |
| 90 vs. 70 years | -9.3 | -10.3 to -8.3 | < 0.001 |
Conclusions
It is hardly surprising that the room for improvement is less in elderly, although even the majority of older patients improve their HRQoL. It is also important to note that other outcomes not measured by the EQ-5D tool may be of importance to the patient.
From a purely statistical viewpoint, the bend occurs at the perhaps worst possible place – the median age for THR patients. We can therefore expect that the residual confounding effect may be substantial when modeling age using a regular linear term, i.e. a straight line.
Since many use cut points to adjust for confounding in this setting, I want to remind that these introduce a cut-point bias. Also, if we use a fine-grained-split for age, e.g. cutting age in 5-year intervals, we may remove the confounding but will also spend more degrees of freedom, risking a less stable model. Furthermore, if the effect that we are studying happens to be confined to one or two age-groups, we may encounter multicollinearity.
Implementation in R
In the next post (coming soon) I will show hot to implement splines in R and compare different alternatives.
References
[1] F. E. Harrell, Regression modeling strategies: with applications to linear models, logistic regression, and survival analysis. New York: Springer, 2001.
[2] H. Malchau, P. Herberts, och L. Ahnfelt, ”Prognosis of total hip replacement in Sweden. Follow-up of 92,675 operations performed 1978-1990”, Acta Orthop. Scand., vol 64, num 5, ss 497–506, okt 1993.
[3] P. Münger, C. Röder, U. Ackermann-Liebrich, och A. Busato, ”Patient-related risk factors leading to aseptic stem loosening in total hip arthroplasty: a case-control study of 5,035 patients”, Acta Orthop., vol 77, num 4, ss 567–574, aug 2006.
[4] S. Overgaard, ”Primær hoftealloplastik: Overlevelseskurver Effekt af alder og køn, alle operationstyper”, i Årsrapport 2011, Dansk Hoftealloplastik Register – Årsrapport 2011 Dansk Hoftealloplastik Register, 2011, ss 52–53.
[5] C. Wainwright, J.-C. Theis, N. Garneti, och M. Melloh, ”Age at hip or knee joint replacement surgery predicts likelihood of revision surgery”, J Bone Jt. Surg Br, vol 93-B, num 10, ss 1411–1415, okt 2011.
[6] G. Garellick, C. Rogmark, J. Kärrholm, och O. Rolfsson, Svenska Höftprotesregistret Årsrapport 2012, vol 2012. 2013.
[7] S. Graves, J. Lynch, L. Griffith, S. Miller, M. Lorimer, Y.-L. Liu, K. Cashman, T. Stanford, G. O’Donahue, D. Davidsson, R. de Steiger, A. Tomkins, R. Vial, och L. Miller, ”Outcome by Patient Characteristics”, i Australian Orthopaedic Association National Joint Replacement Registry., AOA, 2013, ss 52–62.
[8] J. J. Prokopetz, E. Losina, R. L. Bliss, J. Wright, J. A. Baron, och J. N. Katz, ”Risk factors for revision of primary total hip arthroplasty: a systematic review”, BMC Musculoskelet. Disord., vol 13, num 1, s 251, dec 2012.
[9] R. E. Cook, P. J. Jenkins, P. J. Walmsley, J. T. Patton, och C. M. Robinson, ”Risk factors for Periprosthetic Fractures of the Hip: A Survivorship Analysis”, Clin. Orthop., vol 466, num 7, ss 1652–1656, jul 2008.
[10] A. L. Malkani, K. L. Ong, E. Lau, S. M. Kurtz, B. J. Justice, och M. T. Manley, ”Early- and late-term dislocation risk after primary hip arthroplasty in the Medicare population”, J. Arthroplasty, vol 25, num 6 Suppl, ss 21–25, sep 2010.
[11] N. F. SooHoo, E. Farng, J. R. Lieberman, L. Chambers, och D. S. Zingmond, ”Factors That Predict Short-term Complication Rates After Total Hip Arthroplasty”, Clin. Orthop. Relat. Res., vol 468, num 9, ss 2363–2371, apr 2010.
[12] L. P. Hunt, Y. Ben-Shlomo, E. M. Clark, P. Dieppe, A. Judge, A. J. MacGregor, J. H. Tobias, K. Vernon, och A. W. Blom, ”90-day mortality after 409 096 total hip replacements for osteoarthritis, from the National Joint Registry for England and Wales: a retrospective analysis”, The Lancet, vol 382, num 9898, ss 1097–1104, sep 2013.
[13] M. Gordon, M. Greene, Paolo Frumento, Ola Rolfson, Göran Garellick, och André Stark, ”Age and health related quality of life after total hip replacement: Decreasing gains in patients above 70 years of age”, Acta Orthop., vol 85, num 3, ss 244–249, jun 2014.
[14] O. Ethgen, O. Bruyère, F. Richy, C. Dardennes, och J.-Y. Reginster, ”Health-Related Quality of Life in Total Hip and Total Knee ArthroplastyA Qualitative and Systematic Review of the Literature”, J. Bone Jt. Surg., vol 86, num 5, ss 963–974, maj 2004.
[15] F. X. McGuigan, W. J. Hozack, L. Moriarty, K. Eng, och R. H. Rothman, ”Predicting quality-of-life outcomes following total joint arthroplasty: Limitations of the SF-36 health status questionnaire”, J. Arthroplasty, vol 10, num 6, ss 742–747, dec 1995.
[16] P. D. Birdsall, J. H. Hayes, R. Cleary, I. M. Pinder, C. G. Moran, och J. L. Sher, ”Health outcome after total knee replacement in the very elderly”, J. Bone Joint Surg. Br., vol 81-B, num 4, ss 660–662, jan 1999.
[17] P. L. Santaguida, G. A. Hawker, P. L. Hudak, R. Glazier, N. N. Mahomed, H. J. Kreder, P. C. Coyte, och J. G. Wright, ”Patient characteristics affecting the prognosis of total hip and knee joint arthroplasty: a systematic review”, Can. J. Surg. J. Can. Chir., vol 51, num 6, ss 428–436, dec 2008.
[18] O. Rolfson, J. Kärrholm, L. E. Dahlberg, och G. Garellick, ”Patient-reported outcomes in the Swedish Hip Arthroplasty Register: Results of a nationwide prospective observational study”, J. Bone Joint Surg. Br., vol 93-B, num 7, ss 867–875, jan 2011.
[19] J. C. Keurentjes, M. Fiocco, och R. G. Nelissen, ”Willingness to undergo surgery again validated clinically important differences in health-related quality of life after total hip replacement or total knee replacement surgery”, J. Clin. Epidemiol., vol 67, num 1, ss 114–120, jan 2014.
[20] P. Rissanen, S. Aro, H. Sintonen, P. Slätis, och P. Paavolainen, ”Quality of life and functional ability in hip and knee replacements: A prospective study”, Qual. Life Res., vol 5, num 1, ss 56–64, feb 1996.
[21] R. Benroth och S. Gawande, ”Patient-reported health status in total joint replacement”, J. Arthroplasty, vol 14, num 5, ss 576–580, aug 1999.
[22] C. A. Jones, D. C. Voaklander, D. W. Johnston, och M. E. Suarez-Almazor, ”The effect of age on pain, function, and quality of life after total hip and knee arthroplasty”, Arch. Intern. Med., vol 161, num 3, ss 454–460, feb 2001.
[23] J. M. Quintana, A. Escobar, U. Aguirre, I. Lafuente, och J. C. Arenaza, ”Predictors of health-related quality-of-life change after total hip arthroplasty”, Clin. Orthop., vol 467, num 11, ss 2886–2894, nov 2009.
[24] B. M. Lawless, M. Greene, J. Slover, Y.-M. Kwon, och H. Malchau, ”Does Age or Bilateral Disease Influence the Value of Hip Arthroplasty?”, Clin. Orthop. Relat. Res., vol 470, num 4, ss 1073–1078, apr 2012.