Mortality Rates After Metal-on-Metal Hip Resurfacing
Mortality Rates After Metal-on-Metal Hip Resurfacing
We conducted a nationwide retrospective cohort study. Data were obtained from the hospital episode statistics database, which holds information on patients admitted to English hospitals in the United Kingdom's health service. Each record in the database relates to one finished consultant episode, describing the time an individual spends under the care of one NHS consultant. Procedures performed in private hospitals are excluded. The information held includes age, sex, area of usual residence, diagnosis or reason for admission to hospital, and procedure undertaken. Further information is available online (www.hesonline.nhs.uk). Hospital episode statistics data were then linked to mortality records from the Office for National Statistics, which provided information about the date and cause of death.
We extracted anonymised records for all patients over 18 years of age who underwent primary hip replacement between April 1999 and March 2012. Patients were included if they had primary total hip replacement (cemented or uncemented) or primary hip resurfacing. We excluded patients who had revision surgery, total hip replacement of unspecified fixation, hybrid prosthetic hip replacement, and total prosthetic replacement of the femur head. The following exclusions were made to remove potential case mix issues: diagnostic codes indicating fracture or cancer of the hip bones; other injuries due to trauma, such as transport accidents and falls; non-elective admissions; and a diagnosis other than primary hip osteoarthritis.
The outcome of interest was date of death (all cause mortality). Patients were followed for up to 10 years from the date of operation. Secondary outcomes included the most common underlying causes of death using codes from ICD-10 (international classification of diseases, 10th revision). Causes of death included malignant neoplasms (C00-C97); ischaemic heart diseases (I20-I25); cerebrovascular diseases (I60-I69); diseases of arteries, arterioles, and capillaries (I70-I79); pneumonia (J12-J18); and bronchitis, emphysema, and other chronic obstructive pulmonary disease (J40-J44). The exposure of interest was prosthesis type (cemented and uncemented total hip replacement, and metal-on-metal hip resurfacing). Episodes involving these procedures were identified using a combination of OPCS4 codes in the procedure fields (codes from the Office of Population Censuses and Surveys that contain information about a patient’s operations) and ICD-10 codes from the diagnostic fields (that contain information about a patient's illness or condition; (Web Appendix).
Confounding variables at the patient level included age, sex, year of operation, and degree of comorbidity classified for each patient by using the Charlson comorbidity index. We used data from across all diagnostic fields to create a weighted score and an ordinal variable (none (0), mild (1), moderate (2), severe (≥3)). The annual volume of hip replacement operations and operations for metal-on-metal hip resurfacings in an NHS hospital trust was derived for each financial year of hospital episode statistics, and categorised into five equal groups based on annual volume over the follow-up period. An NHS hospital trust, known as an acute trust, provides secondary health services within the English NHS. We treated the Royal Orthopaedic Hospital NHS Trust in Birmingham as a separate category owing to the high volume of metal-on-metal hip resurfacings performed. Ecological variables were linked to the lower level, super output area where the patient lived. Super output areas are small areas of England that have a consistent population size with a minimum population of 1,000 people and mean of 1,500 people. Ecological variables included rurality (categorised as urban population ≥10,000; town and fringe; and village or isolated) and the Index of Multiple Deprivation 2004 as a measure of social deprivation.
In randomised controlled trials, each person has an equal probability of being in a treatment or control group. Observational study designs are limited by an inherent imbalance of both known and unknown confounders, which makes some patients more likely to undergo metal-on-metal hip resurfacing than total hip replacement. Because the type of surgery given was not randomly allocated in our study, we accounted for confounding by indication by using propensity score matching methods. These methods for the assessment of causality in epidemiological studies has been previously described. The propensity score represents the probability that a patient received the intervention (that is, metal-on-metal hip resurfacing). We fitted two separate logistic equations where the outcomes were metal-on-metal hip resurfacing versus cemented or uncemented total hip replacement. Age, sex, Charlson comorbidity, rurality, Index of Multiple Deprivation, volume of total hip replacement, volume of metal-on-metal hip resurfacing, and year of operation were introduced as potential confounders of all cause mortality in the long term.
With propensity scores using a 0.02 standard deviations calliper, we matched each patient undergoing metal-on-metal hip resurfacing to three comparable controls undergoing total hip replacement. This is the standard method for minimising confounding by indication, which not only provides participants with balanced baseline characteristics in both surgical groups, but also eliminates patients undergoing metal-on-metal hip resurfacing with no comparable controls.
We included patients undergoing metal-on-metal hip resurfacing and controls in a Cox regression survival model to describe the association between prosthesis type and time to death from any cause. The model is stratified on matched sets to allow for the correlation between matched pairs of patients undergoing metal-on-metal hip resurfacing and controls. We tested for evidence of interactions of prosthesis type with age, sex, and comorbidity. Because clustering exists within the data (patients nested within hospital trusts), we fitted a multilevel survival model by extending the Cox regression model to include a frailty term with a Gaussian distribution. This inclusion allowed adjustment for evidence of unexplained variation across hospital trusts. The proportional hazards assumption was assessed using Shoenfelds residuals. We used Kaplan-Meier plots to estimate the probability of survival up to 10 years after surgery in patients undergoing metal-on-metal hip resurfacing and controls. To assess the potential effect of unmeasured confounders, we conducted a Rosenbaum bounds sensitivity analysis. This analysis estimates the magnitude of hidden residual bias that would have to be present to explain the associations actually observed. Stata version 12.1 was used for all statistical analyses.
Methods
Study Design, Setting, and Source of Data
We conducted a nationwide retrospective cohort study. Data were obtained from the hospital episode statistics database, which holds information on patients admitted to English hospitals in the United Kingdom's health service. Each record in the database relates to one finished consultant episode, describing the time an individual spends under the care of one NHS consultant. Procedures performed in private hospitals are excluded. The information held includes age, sex, area of usual residence, diagnosis or reason for admission to hospital, and procedure undertaken. Further information is available online (www.hesonline.nhs.uk). Hospital episode statistics data were then linked to mortality records from the Office for National Statistics, which provided information about the date and cause of death.
Participants
We extracted anonymised records for all patients over 18 years of age who underwent primary hip replacement between April 1999 and March 2012. Patients were included if they had primary total hip replacement (cemented or uncemented) or primary hip resurfacing. We excluded patients who had revision surgery, total hip replacement of unspecified fixation, hybrid prosthetic hip replacement, and total prosthetic replacement of the femur head. The following exclusions were made to remove potential case mix issues: diagnostic codes indicating fracture or cancer of the hip bones; other injuries due to trauma, such as transport accidents and falls; non-elective admissions; and a diagnosis other than primary hip osteoarthritis.
Primary Outcome and Exposure
The outcome of interest was date of death (all cause mortality). Patients were followed for up to 10 years from the date of operation. Secondary outcomes included the most common underlying causes of death using codes from ICD-10 (international classification of diseases, 10th revision). Causes of death included malignant neoplasms (C00-C97); ischaemic heart diseases (I20-I25); cerebrovascular diseases (I60-I69); diseases of arteries, arterioles, and capillaries (I70-I79); pneumonia (J12-J18); and bronchitis, emphysema, and other chronic obstructive pulmonary disease (J40-J44). The exposure of interest was prosthesis type (cemented and uncemented total hip replacement, and metal-on-metal hip resurfacing). Episodes involving these procedures were identified using a combination of OPCS4 codes in the procedure fields (codes from the Office of Population Censuses and Surveys that contain information about a patient’s operations) and ICD-10 codes from the diagnostic fields (that contain information about a patient's illness or condition; (Web Appendix).
Potential Confounders
Confounding variables at the patient level included age, sex, year of operation, and degree of comorbidity classified for each patient by using the Charlson comorbidity index. We used data from across all diagnostic fields to create a weighted score and an ordinal variable (none (0), mild (1), moderate (2), severe (≥3)). The annual volume of hip replacement operations and operations for metal-on-metal hip resurfacings in an NHS hospital trust was derived for each financial year of hospital episode statistics, and categorised into five equal groups based on annual volume over the follow-up period. An NHS hospital trust, known as an acute trust, provides secondary health services within the English NHS. We treated the Royal Orthopaedic Hospital NHS Trust in Birmingham as a separate category owing to the high volume of metal-on-metal hip resurfacings performed. Ecological variables were linked to the lower level, super output area where the patient lived. Super output areas are small areas of England that have a consistent population size with a minimum population of 1,000 people and mean of 1,500 people. Ecological variables included rurality (categorised as urban population ≥10,000; town and fringe; and village or isolated) and the Index of Multiple Deprivation 2004 as a measure of social deprivation.
Statistical Methods
In randomised controlled trials, each person has an equal probability of being in a treatment or control group. Observational study designs are limited by an inherent imbalance of both known and unknown confounders, which makes some patients more likely to undergo metal-on-metal hip resurfacing than total hip replacement. Because the type of surgery given was not randomly allocated in our study, we accounted for confounding by indication by using propensity score matching methods. These methods for the assessment of causality in epidemiological studies has been previously described. The propensity score represents the probability that a patient received the intervention (that is, metal-on-metal hip resurfacing). We fitted two separate logistic equations where the outcomes were metal-on-metal hip resurfacing versus cemented or uncemented total hip replacement. Age, sex, Charlson comorbidity, rurality, Index of Multiple Deprivation, volume of total hip replacement, volume of metal-on-metal hip resurfacing, and year of operation were introduced as potential confounders of all cause mortality in the long term.
With propensity scores using a 0.02 standard deviations calliper, we matched each patient undergoing metal-on-metal hip resurfacing to three comparable controls undergoing total hip replacement. This is the standard method for minimising confounding by indication, which not only provides participants with balanced baseline characteristics in both surgical groups, but also eliminates patients undergoing metal-on-metal hip resurfacing with no comparable controls.
We included patients undergoing metal-on-metal hip resurfacing and controls in a Cox regression survival model to describe the association between prosthesis type and time to death from any cause. The model is stratified on matched sets to allow for the correlation between matched pairs of patients undergoing metal-on-metal hip resurfacing and controls. We tested for evidence of interactions of prosthesis type with age, sex, and comorbidity. Because clustering exists within the data (patients nested within hospital trusts), we fitted a multilevel survival model by extending the Cox regression model to include a frailty term with a Gaussian distribution. This inclusion allowed adjustment for evidence of unexplained variation across hospital trusts. The proportional hazards assumption was assessed using Shoenfelds residuals. We used Kaplan-Meier plots to estimate the probability of survival up to 10 years after surgery in patients undergoing metal-on-metal hip resurfacing and controls. To assess the potential effect of unmeasured confounders, we conducted a Rosenbaum bounds sensitivity analysis. This analysis estimates the magnitude of hidden residual bias that would have to be present to explain the associations actually observed. Stata version 12.1 was used for all statistical analyses.
Source...