Cancer risk after total hip replacements. A prospective study of 41,402 patients in the Norwegian Arthroplasty Register linked to the Cancer Registry of Norway

Background: Concerns have been raised that implants used in total hip replacements (THR) could lead to increased cancer risk. Several different materials, metals and �xation techniques are used in joint prostheses and different types of articulation can cause an increased invasion of particles or ions into the human body. Methods: Patients with THR registered in the Norwegian Arthroplasty Register during 1987-2009 were linked to the Cancer registry of Norway. Patients with THR due to osteoarthritis, under the age of 75 at time of surgery, were included. Standardized incidence ratios (SIR) were applied to compare cancer risk for THR patients to the general population. Types of THR were divided into cemented (both components), uncemented (both components), and hybrid (cemented femoral and uncemented acetabular components). To account for selection mechanisms, time dependent covariates were applied in Cox-regression, adjusting for cancer risk the �rst 10 years after surgery. The analyses were adjusted for age, gender and if the patient had additional THR-surgery in the same or the opposite hip. The study follows the STROBE guidelines. Results: Comparing patients with THR to the general population in Norway we found no differences in cancer risk. The overall SIR for the THR-patients after 10 years follow-up was 1.02 (95% CI: 0.97-1.07). For cemented THR, the SIR after 10 years follow-up was 0.99 (95% CI: 0.94-1.05), for uncemented, 1.16 (95% CI: 1.02-1.30), and for hybrid 1.12 (95% CI: 0.91-1.33). Adjusted Cox analyses showed that patients with uncemented THRs had an elevated risk for cancer (hazard ratio: HR=1.24, 95% CI: 1.05-1.46, p=0.009) when compared to patients with cemented THRs after 10 years follow-up. Strati�ed by gender the increased risk was only present for men. The risk for patients with hybrid THRs was not signi�cantly increased (HR=1.07, 95% CI: 0.85-1


Background
In total hip replacement (THR) surgery, implants consisting of metals, polymers, ceramics and are inserted, some of which are xated by means of bone cement.The numbers of metals and other materials, and the variety of sizes and bearing surfaces used in these implants over the years, have been substantial.Concerns have been raised whether the insertion of implants might lead to subsequent malignancies [1][2][3][4].Tumours could hypothetically develop at the implant site, due to local reactions, or elsewhere in the body, caused by systemic in uences.In animal studies, different materials have been used to model cancer development, but questions have been raised whether biomaterial-related tumours in animals have relevance to humans [5].Most studies have found no increased cancer risk after THRs compared to the general population [6][7][8][9][10][11][12][13][14][15][16].In a meta-analysis, Visuri and colleagues observed a decreased cancer risk for patients with arthroplasties [17].Another meta-analysis did not con rm an overall increased cancer risk after THR and Total Knee Arthroplasty (TKA), but described an elevated risk for prostate cancer and melanomas [18].A group from Sweden has reported an increased cancer risk among patients who had received a TKA due to osteoarthritis and rheumatoid arthritis, and also reported a latency effect for cancer after insertion of joint replacements [19].
In the present study, we have linked data on THRs from the Norwegian Arthroplasty Register (NAR) to the Cancer Registry of Norway (CRN).The primary aim of the study was to determine if there were differences in the long term (after 10 years) cancer risk according to types of prosthesis xation; cemented (both femoral and acetabular components cemented), uncemented (both femoral and acetabular components uncemented), and hybrid (cemented femoral and uncemented acetabular component).

Methods
In this prospective cohort study, follow-up time was measured from insertion of the initial prosthesis and until cancer, emigration, death, or December 31st 2009 (end of study), whichever came rst.
The Norwegian Arthroplasty Register (NAR) started registration of total hip replacements (THR) in September 1987 [20].More than 95% of patients receiving a THR are reported to the NAR [21,22].Patients registered with primary THR, from 1987 to 2009 with known prosthesis xations, osteoarthritis, under the age of 75 at time of surgery, were followed from their rst/initial operation.Since this study looks at the late (after 10 years) risk for cancer, patients older than 75 years were excluded.The selected patients were linked to the Cancer Registry of Norway (CRN) using the 11-digit personal identi cation number unique for all Norwegian citizens.The CRN was established in 1953 and registration of new cancer cases is compulsory.The registry has information on type of malignancy, date of diagnosis and initial treatment and demographics on 99% of all cancer patients in Norway [23,24].THR patients with cancer prior to the THR were excluded from the analysis.Hence, 41,402 patients were included.In the les from the CRN, ICD-7 code 189 (basal cell carcinoma) was not included.
Type of xation was coded as fully cemented, fully uncemented, and hybrids (cemented femoral and uncemented acetabular components).THR patients with reversed hybrids (uncemented femoral and cemented acetabular components) were excluded due to few observations and short follow-up.The outcome variable of this study was the incidence of cancer occurring 10 years after insertion of THR.The Cancer Registry of Norway is a mandatory national health registry, regulated by law.All hospitals, laboratories, and general practitioners are obligated to report new cancer cases to the registry within two months.The Norwegian Arthroplasty Register (NAR), started in 1987, is a voluntary register licensed by the Data Inspectorate of Norway (16/01622-3).Patients give a written informed consent to be included in the registry.The operating surgeon reports the operation to the registry on a one-page standard form.The present study was approved by the Regional Committee for Medical and Health Research Ethics, Western Norway (170.06) and The Norwegian Data Protection Authority (06/01218-2).The study is reported according to the STROBE guidelines.
Standardized incidence ratio (SIR) was calculated to quantify the difference in cancer risk between THR patients and the population with corresponding age, gender and calendar year.The SIR will equal the hazard rate for patients divided with the hazard for the corresponding population [25,26].SIRs after 10 years follow-up were the main focus, but SIRs for the complete follow-up and SIRs before 10 years followup were also examined.SIR before 10 years follow-up may indicate possible selections of healthier or sicker patients for the different types of prosthesis.A SIR below 1 can indicate healthier patients than the population, while a SIR above 1 indicate sicker patients.
A Cox proportional hazards regression model, for the risk for cancer after 10 years follow-up, including a time-dependent indicator to adjust for the hazard for cancer in the rst decade was set up to adjust for potential selection mechanisms.Hence, in this model, the hazard ratios after 10 years follow-up between the different types of prosthesis xation were adjusted for the baseline hazards (cancer risk) the rst 10 years follow-up.This difference-in-difference model minimizes the effect of unknown covariates.The regression models were adjusted for gender, age at operation, and time-dependent covariates for time to a contralateral THR and/or a revision operation.For testing of the proportionality assumption, a test of the Schoenfeld residuals after tting the Cox-model was performed.To see if the results were consistent within subgroups, the analysis was strati ed based on the age-categories and gender.Follow-up time was measured from insertion of the initial prosthesis and until cancer, emigration, death, or December 31st 2009 (end of study), whichever came rst.Median follow-up was calculated using the inverse Kaplan-Meier method [27].IBM-SPSS version 22 (IBM-SPSS, Chicago Ill), Stata (version 13-IC), and Fortran [26] were used for the statistical analyses.P-values less than 0.05 (5%) were considered statistically signi cant.

Results
There were 13,954 (34 %) males and 27,448 (66 %) females, with a mean age of 65 (SD=7) and 66 (SD=7) respectively.Of the THR patients, 6,167 were diagnosed with at least one cancer after THR, 1,789 of these occurring more than 10 years after the rst primary hip implant (Table 1).There was huge variation in the types of cancer and the most common single cancer type after 10 years follow-up was prostate cancer (257 males) and breast cancer (183 females) (Table 2).Total person years in the study were 453,950 and median follow-up was 11.9 years (Table 1).Uncemented THRs were predominantly given to younger and healthier individuals.For the included patients, the mean age for uncemented THR was 58.1 (sd=8.3)years, for cemented 67.0 (sd=5.9)years, while for hybrid prostheses mean age was 63.3 (sd=7.2) years.
For THR-patients the overall standardized incidence ratio after 10 years follow-up was not statistically signi cantly different from the general population in Norway.SIR=1.02,95% CI: 0.97-1.07.
Testing of the proportionality assumption in the regression model showed that in a model with no timedependent covariates, the dummy variable for uncemented (versus cemented) implants interacted with time (p=0.008).However, for the fully adjusted model, with all the mentioned time-dependent covariates, none of the covariates had an interaction with time (overall p=0.96).
Strati ed by age categories, there were no signi cant differences in risk for cancer between the different types of xation (Table 3).Males with uncemented THRs had an increased risk for cancer compared to males with cemented THRs (HR=1.41,95% CI: 1.11-1.80,p=0.004), while this was not found for females (HR=1.09,95% CI: 0.87-1.36,p=0.47).Strati ed by gender and age categories we found no statistically signi cant differences comparing uncemented and hybrid prostheses to cemented prostheses (Table 4).

Discussion
There has been focus on the risk for cancer after insertion of joint replacements [2][3][4].The majority of studies are from the national arthroplasty registries in Finland [9,10,12,15,16], Scotland [6], England and Wales [14], and Sweden [19].All but one of these studies conclude that there is no (or a negligible) increased risk for cancer after insertion of joint replacements.On the other hand, Wagner and colleagues have reported an overall increased cancer risk for total knee arthroplasty (TKA) patients compared to the general population.In addition, they reported ndings of speci c cancer types, which they argue can be a result of TKA exposure [19].Our study supports previous ndings showing no overall increased risk for cancer after THR, neither before nor after 10 years follow-up.For uncemented THRs we found an association with a small increased risk for cancer for males.
There are limitations in this study.As shown by Lie and colleagues [28], patients with a THR have reduced overall mortality compared to the general population, while THR patients under 60 years have increased mortality and patients over 80 years of age have considerably reduced mortality compared to the population in general.Furthermore, uncemented prostheses have predominantly been given to younger and healthier patients, while cemented prostheses have been given to elderly and frailer patients [29].
Consequently, adjusting for the risk for cancer, the rst 10 years after primary THR (difference-in-difference model) would adjust for (unknown) risk factors contributing to the baseline risk for cancer for the different categories of patients.Still, there is a potential for a complex selection mechanism for receiving a THR and also for receiving the different types of THR, which we are not able to adjust for.Hypothetically, receiving a THR can increase the attention to own health.Subsequently, this can lead to more visits for medical care (e.g.general practitioner), which may increase the number of tests and also the probability of being tested for cancer.This would particularly be the case for prostate cancer in men.
Recently, Cartilage Oligomeric Matrix Protein (COMP), which plays an important role in the organization of the extracellular matrix of cartilage, has been identi ed as a potent driver of the progression of prostate cancer, acting in an anti-apoptotic fashion by interfering with the Ca2+ homeostasis of cancer cells [30].In a retrospective case control series in prostate cancer patients with and without osteoarthritis, this condition was identi ed as an independent risk factor for metastatic disease.However, when joint arthroplasty was included in the model, osteoarthritis was no longer an independent risk factor [31].It is unlikely that this association can explain the small increase in cancer risk in men with uncemented compared to cemented THR in our study.
The common analytical approach to study cancer risk for THR and TKA patients is to compare the observed cancer risk for arthroplasty patients with cancer rates in the general population.When SIRs are used to compare the cancer risk for the patients studied with the cancer risk in the population, it is assumed that prosthesis patients are comparable to the general population.Previous studies nd no increase in risk for cancer after an arthroplasty compared to the general population [6][7][8][9][10][11][12][13][14][15][16][17][18].Overall, this agrees with our nding, using the same analyses techniques.
From studies of secondary cancer related to anti-neoplastic treatment, it is known that the latency from the rst to the subsequent malignant tumour is 10 years or more [32,33].In the regression models in this study, we took into consideration that the development of cancer related to arthroplasty can take at least one decade, and that cancer diagnoses during the rst years after a THR operation are most likely related to factors other than the arthroplasty itself.To compare the different types of xations, we therefore used baseline cancer risk at the rst 10 years follow-up as a reference.In these analyses we thus compared the difference in cancer risk between different types of xations, and other factors, adjusting for the crucial selection for receiving the THR.
We found an increased cancer risk for patients with two uncemented prostheses components, compared to patients where both prostheses components were cemented.Patient with hybrid prostheses had not a statistically increased risk for cancer compared to patients with two cemented components.
In analyses of cancer after THR, death is a competing risk.For the present analyses we did not take competing risks into account.The reason is that since there are differences in selection mechanisms between the different prosthesis, which will be apparent in analyses with death as endpoint, death may also be a collider in causal terminology when we study the risk for receiving cancer.Accordingly, using models for competing risks, a false and elevated risk between the types of THR and cancer was present (analyses not shown).The relative differences in the SIRs in our analyses correspond to the ndings from the Cox model with time-dependent adjustment, which we consider strengthen our ndings.
There has been a concern about cancer risks associated with metal on metal articulations for THRs [9,10,14,34], but other and newer types of articulations should also be studied [35].Articulation has not been included in the present study since the majority of THR prostheses in the Norwegian Arthroplasty Register have a metal head and polyethylene cup, and other articulations have lower numbers or shorter follow-up [22].Metal-on-metal has rarely been used in Norway in the study period.Only approximately 200 cemented and less than 200 uncemented implants of this type, most of which with small heads (<32 mm) were used in the time period studied [36].Metal on metal resurfacing implants were excluded from the study, because this type of THR is a marginal problem in this study, and omitting these implants would not alter our ndings.

Conclusions
In the present study, we found no increased risk for cancer in THR patients compared to the general population.However, we found a small increased risk for cancer after insertion of THR where both components were uncemented, compared to prostheses where both were cemented.In gender-strati ed analysis, this increased risk was only found for men, but not found in age-strati ed analysis for men.The difference was small and prone to unmeasured confounding.The risk for cancer after joint replacements and possible mechanisms related to cancer for patients with musculoskeletal diseases and/or joint replacements should be studied further.Surveillance of new products, materials and prostheses, with respect to rare and adverse outcomes like cancer, is important, also in the future.

Declarations
analyses in collaboration with orthopaedic surgeons OF and LIH, and oncologists SDF and CT.ED was responsible for writing of the draft manuscript.All the authors participated in critical review and preparation of the nal manuscript.
34. Rosengren BE: Metal-on-metal hip implants and the risk of cancer.Bmj 2012, 345:e4605.35.Levine BR, Hsu AR, Skipor AK, Hallab NJ, Paprosky WG, Galante JO, Jacobs JJ: Ten-year outcome of serum metal ion levels after primary total hip arthroplasty: a concise follow-up of a previous report.

Supplementary Files
This is a list of supplementary les associated with this preprint.Click to download. STROBEChecklistv4MSWord.doc
Number of hips, cases of cancer, males and follow-up time.

Table 2 :
Types of cancer following a total hip replacement, with more than 100 observed cases,

Table 3 :
SIR and Cox model with time dependent covariates for the excess risk after 10 years

Table 4 :
Cox model with time dependent covariates for subgroups of gender and age groups.