Article Text
Abstract
Objective A systematic and long-term data collection on the treatment of focal cartilage defects (FCDs) of the knee is needed. This can be achieved through the foundation of a National Knee Cartilage Defect Registry. The aim of this study was to establish the nationwide burden of knee cartilage surgery, defined as knee surgery in patients with an FCD. We also aimed to identify any geographical differences in incidence rates, patient demographics or trends within this type of surgery.
Setting A population-based study with retrospective identification of patients undergoing knee cartilage surgery in Norway through a mandatory public health database from 2008 to 2011.
Participants We identified all patients undergoing cartilage surgery, or other knee surgery in patients with an FCD. All eligible surgeries were assessed for inclusion on the basis of certain types of ICD-10 and NOMESKO Classification of Surgical Procedures codes.
Primary and secondary outcome measures The variables were diagnostic and surgical codes, geographic location of the performing hospital, age and sex of the patients. Yearly incidence and incidence rates were calculated. Age-adjusted incidences for risk ratios and ORs between geographical areas were also calculated.
Results A total of 10 830 cases of knee cartilage surgery were identified, with slight but significant decreases from 2008 to 2011 (p<0.0003). The national incidence rate was 56/100 000 inhabitants and varied between regions, counties and hospitals. More than 50% of the procedures were palliative and nearly 400 yearly procedures were reparative or restorative.
Conclusions Knee cartilage surgery is common in Norway, counting 2500 annual cases with an age-adjusted incidence rate of 68.8/100 000 inhabitants. There are significant geographical variations in incidence and trends of surgery and in trends between public and private hospitals. We suggest that a national surveillance system would be beneficial for the future evaluation of the treatment of these patients.
- EPIDEMIOLOGY
This is an Open Access article distributed in accordance with the Creative Commons Attribution Non Commercial (CC BY-NC 4.0) license, which permits others to distribute, remix, adapt, build upon this work non-commercially, and license their derivative works on different terms, provided the original work is properly cited and the use is non-commercial. See: http://creativecommons.org/licenses/by-nc/4.0/
Statistics from Altmetric.com
Strengths and limitations of this study
This cohort study presents the national burden of knee cartilage surgery in Norway.
The geographical differences and differences in trends are reliable as the data collection is mandatory for all hospitals.
ICD codes were used for inclusion and this represents a limitation, as there are no specific codes for ‘non-acute focal cartilage defect’, which leads to unspecific diagnosis. This limitation is partly corrected for by adding NOMESKO Classification of Surgical Procedures surgical codes to the inclusion criteria.
Compliance and validity are limitations for the data quality in most registry studies. The register included in the present study has previously been shown to both overestimate and underestimate clinical conditions; however, studies that are more recent have demonstrated high validity.
Introduction
Knee cartilage injury is a well-known condition after the introduction of knee arthroscopy and MRI. Cartilage injury might consist of a single or several focal lesions or it might constitute generalised degenerative changes within the knee. Focal lesions are classified as traumatic or degenerative and some exist without causing symptoms. They are believed to lead to a chronic osteoarthritic stage with pain and reduced function, which however has been demonstrated only in animal models.1 ,2 Arthroscopic studies have shown that focal cartilage defects (FCDs) within the knee occur in 19–67% of patients with painful knees.3–6 A systematic review found a prevalence of 36% in athletes examined by arthroscopy, MRI or both, whereas 14% were asymptomatic.7 Another study conducted MRI of the tibiofemoral joint in persons aged 50 years or more from the general population (mean age of 62.3 years).8 They found cartilage abnormalities in 69%. We suspect FCDs to be common also in the general population including participants under the age of 50 years.
Several years of research on cartilage surgery have still not led to a clear gold standard treatment of FCDs within the knee. The results from randomised controlled trials (RCTs) are variable,9–16 the patient population is heterogeneous17 and a group of non-operated controls has still not been included in an RCT, making it difficult to decide the role of rehabilitation alone. Also, the quality of clinical studies on cartilage research is low.18 ,19 The most commonly performed procedures on patients with knee cartilage injuries are palliating procedures, such as chondroplasty (CP) and debridement, which have demonstrated symptomatic relief in uncontrolled cohort studies but failed to do so in RCTs.20 ,21 Unfortunately, in this area of orthopaedic surgery, the practice of evidence-based medicine is lacking and the procedures are still used for patients with degenerative changes within their knees.
Results from other orthopaedic registries have led to improved treatment quality and we are currently looking into the potential benefits and challenges of establishing a National Knee Cartilage Defects Registry. Before establishing such a registry, several conditions must be explored. This study intends to present the burden of surgery for the disease.
Two studies from the USA have calculated incidence rates from an insurance database.22 ,23 Montgomery et al showed incidence rates ranging from 1.27 to 1.57/10 000, while McCormick et al presented incidence rates ranging from 63 to 104/10 000. These numbers would represent 635–52 000 yearly procedures when applied to the number of inhabitants in Norway, which is a very wide interval. In 2014, a study on trends of cartilage injuries documented by arthroscopy in Denmark was published.24 It excluded patients with osteoarthritis (OA) and found an incidence of 40/100 000 person-years for the years 1996–2011.
The aim of this study was to establish the nationwide burden of surgery on knees with knee cartilage defects in Norway. This will play an important role in the evaluation of the possible establishment of a National Knee Cartilage Defects Register in Norway. We calculated the national and regional incidences and aimed at detecting any geographical variations. The latter is of major interest for health development research, the medical industry as well as healthcare providers. Our hypothesis was that cartilage surgery is uncommon and performed mainly in hospitals around the larger cities and that only University hospitals perform advanced cartilage surgery.
Patients and methods
Data source
The study is descriptive with population-based data from the years 2008 to 2011 in Norway. It is a retrospective cohort study through the continuous data collection done by the Norwegian Patient Registry (NPR). The NPR is run by the Norwegian Directorate of Health and contains data on the activity in specialist health services. Norway has approximately 5 million inhabitants. The country consists of 4 health regions and 19 administrative counties. The South East region is most populous, followed by the West, Mid and North regions. Norway has a national public healthcare system aiming at equal health services to all inhabitants regardless of their income or private insurances. Also, a growing number of private hospitals and surgical centres offer mainly elective orthopaedic surgery to patients with private insurance, reimbursed by public funding through government contracts or paying out of pocket (previously 10–15% of specific elective surgeries, however, influenced by substantial geographical variation25).
The NPR contains reports on the International Statistical Classification of Diseases and Related Health Problems (ICD) code and the NOMESKO Classification of Surgical Procedures (NCSP) code along with other reported factors. It is obligatory for all public hospitals, and for private hospitals with a contract with the public healthcare system, to report their activity to NPR. The arrangements thereby also involve all major private hospitals. The present patient pool consists of all Norwegian patients.
We aimed at detecting cases undergoing surgery for knee cartilage defects. Distinguishing between traumatic and degenerative lesions is often difficult clinically and the development from an FCD to OA might be seen as a continuum. In addition, the ICD-10 coding system is unspecific and further challenges this distinction. Cases were identified from the NPR through predefined surgical procedure codes (all NCSP codes constituting surgery on the knee and/or calf) and ICD-10 codes (table 1) and retrieved as eligible for inclusion if any combination of surgical and diagnostic codes, according to table 1, was present. ICD-10 codes for concomitant injuries are not included. The list (table 1) was chosen after a consensus meeting between head orthopaedic surgeons of the largest hospital in our region. We also contacted experienced orthopaedic surgeons from other hospitals by mail in order to ensure that all possible codes were included. We included diagnosis M17 after these interchanges as several stated that they use M17 also for FCDs. Patients coded with M17 may have degenerative changes, although some have actual focal lesions. Therefore, we made an upper age limit of 67 years for inclusion and presented descriptive analyses with a distinction between those under and above 50 years of age.
Our data were anonymous and considered as statistical data rather than information on health from individual participants. We received the data set within an SPSS file and recognised all cases that underwent knee cartilage surgery during the 4 years 2008–2011. Cases more likely to constitute OA were excluded; therefore, patients aged 67 years or more, patients undergoing prosthesis surgery and patients with M17 in combination with non-cartilage procedures (only meniscal resection for instance) or high tibial osteotomy were excluded. Cases with M17 and procedures classified as cartilage surgery were included. The final number after exclusion was 10 830 in the 4-year period (figure 1).
Variables and data
The variables were ICD code, NCSP code, age, gender and length of the hospital stay. Additionally, we requested data on the health region, county and institution and received geographical variables only for the years 2008–2009.
Statistics
We defined NCSP codes as cartilage surgery, meniscal surgery or other types of surgery. The different types of cartilage surgery were defined as palliative, repairing or restorative. All cases were divided into subgroups on the basis of these definitions. We chose the term palliative as these procedures are meant to decrease pain for the patients, although its efficacy is not proven for all indications. CP or debridement was defined as palliative surgery, cartilage repair included microfracture (MF) and cell-based repair with either autologous chondrocyte implantation (ACI) or stem cells and restorative techniques included techniques aiming at restoring the articular cartilage without cartilage repair tissue produced on-site as well as mosaicplasty (MP) and allograft transplantation (which is currently not in use in Norway).
The data were analysed with IBM SPSS Statistics (V.22.0). We assessed the distribution of the data with age as the dependent value and concluded with a non-normality distribution. The categorical variables on events of cartilage surgery were assumed to fulfil the criteria of a Poisson distribution. Cases were stratified by age, sex, health region, county and year of surgery. Incidences of cartilage surgery were given per 100 000 inhabitants and were adjusted to age group, region or county by calculation based on population data from Statistics Norway, which is an academically independent organisation administered under the Ministry of Finance in Norway. The data were assembled from their web pages. We compared the incidences for each of the 4 years to each other using rate ratios (RRs) and tested for significance using Wald tests. We used the Cochran-Armitage trend test for comparing trends in the current study with the existing literature.
Demographics were considered by descriptive statistics. Differences in categorical variables were calculated with ORs and tested with Pearson χ2 tests with geographical localisation as the dependent variable. We explored age differences between subgroups with box plots and performed a Kruskal-Wallis test to test the statistical difference. A Bonferroni correction adjusted the new α level to 0.0125 with four independent analyses (CP vs MF, MF vs ACI, CP vs ACI, MF vs MP) before Mann-Whitney U tests were performed. We were not able to address potential confounders such as actual differences in the prevalence of knee cartilage defects, or differences in the willingness to seek medical assistance for painful knees or the willingness to undergo surgery.
Ethics
We received anonymous data from the NPR, which acts under approvals of the Norwegian Directorate of Health. The study was evaluated by the Regional Committees for Medical and Health Research Ethics (REC) (ref: 2010/777) and approval is not necessary as the data are anonymous. We consulted the Norwegian Data Protection Authority and the study is not obliged for notification due to the collection of anonymous data. The data are to be considered as statistical data rather than information on health in individual participants.
Results
A total of 10 830 cases matched our inclusion criteria for cartilage surgery for the years 2008–2011 and a flow chart is presented in figure 1. There were 2897 cases in 2008, 3114 in 2009, 2732 in 2010 and 2087 in 2011. A total of 21 143 procedures (see online supplementary appendix 1) were reported throughout the 4 years, which results in a mean of 1.96 procedures per included case. The most common cartilage surgery was resection of the articular cartilage (NGF3y) followed by fenestration or forage or bone/MF (NGK29). The most common non-cartilage surgery was meniscal surgery followed by synovectomy. The mean age for all years was 45.0 (SD 13.7), whereas the mean age for 2008 was 45.6 (SD 13.7) and for 2011 was 43.1 (SD 14.2), which was significantly lower than for the other years (p value <0.001). The male ratio varied from 55.2% to 58.7%.
Incidences
The incidence rate of having experienced cartilage surgery in Norway throughout 2008–2011 is 56/100 000 inhabitants and age-adjusted incidence rate is 68/100 000 inhabitants between 4 and 66 years of age. Table 2 displays the age-adjusted incidence rates for the different years and age groups. The incidence rate from 2008 was set as the reference when calculating RR between included years. The only significant RR was for 2011, which was 0.69 (95% CI 0.65 to 0.73, p value <0.0003).
The incidences of cartilage surgery in public hospitals in the four different health regions display great diversity as cartilage surgery is twice as common within the Northern region as opposed to the South East region (figure 2). However, when all the procedures performed privately are included, the regional differences change and the Western region becomes the region with the highest incidence (figure 3). The incidence in the Western region (161/100 000 inhabitants) is four times higher than that in the South East region, which has the lowest incidence (37/100 000 inhabitants). The incidences throughout the 19 different counties also display large variations (figure 2). The incidences range from 7.3 to 278.1/100 000 inhabitants.
Trends
The trends for type of surgery varied between both regions and between private and public hospitals (table 3). Whereas private hospitals had nearly 90% debridement, this represented only approximately half of the procedures in public hospitals. Advanced cartilage surgery (repair or restorative techniques) accounted for almost 400 procedures per year. The middle health region had the lowest proportion of advanced cartilage surgery (13.4%) in 2009. In comparison, the northern region performed 42.6% of such procedures in 2009. The corresponding numbers for 2008 were 11.7% and 49.6%. The OR of having advanced cartilage surgery performed in the northern region compared to the other regions was 7.44 (6.11–9.06). Nationwide, the MP/OAT was the most frequent of the repair or restorative procedures for all years, ranging from 57.6% to 62.8%, whereas 4.2%–6.6% were cell transplantation techniques.
A substantial part of all included cases of cartilage surgery was performed in private institutions, whereas they performed 19.8% of the repair or restorative procedures (table 3). The OR of being treated with these methods over palliative procedures in private rather than public institutions was 0.18 (0.08–0.43). A Pearson χ2 confirmed a highly significant association between the regions and between private and public hospitals. Most patients were treated in an outpatient setting and this accounted especially for private institutions. University hospitals performed 44.5% of cases with advanced cartilage surgery, whereas they performed 57.5% of all transplantation techniques, 56.8% of MP procedures and only 13.6% of MF procedures.
Age
The ages between the seven different subgroups were statistically significantly different (p<0.001); whereas the CP group (median 51.0) was significantly older than both the MF (median 39.0) and ACI groups (median 29.0), the MF group was older than the ACI group and not statistically significant different from the MP group (median 42.0). The age distribution of advanced cartilage surgery showed that the majority of procedures are performed on patients aged 20–50 years. Transplantation procedures were seldom performed in the oldest age group (50–67 years of age), whereas the youngest group (<20 years of age) was more commonly treated with MF followed by transplantation. ORs demonstrated that MP/OAT and ACI were more common for patients under 50 years of age, whereas MF and MP/OAT were more common for patients under the age of 20 years.
Discussion
A total of 10 830 cases were included and represent the nationwide load of knee cartilage surgery in Norway throughout 2008–2011. There are 2500 cartilage surgeries yearly and 400 of these are advanced cartilage surgery. The total incidence of all cartilage surgery over these four years is 56/100 000. These numbers are within the range of incidences for knee ligament surgery in Norway, which is considered a common surgery. Granan et al27 found an incidence of ACL surgery of 34/100 000 inhabitants, although there were 85/100 000 in the age group 16–39 years of age in Norway in their baseline study of the Scandinavian Knee Ligament Registries.
Although common, the yearly incidence varies greatly among age groups, health regions, counties and between public and private hospitals. Cartilage surgery is not in use mainly around the largest cities or regional hospitals and University clinics, in contrast to our hypothesis. Private institutions accounted for 43% of all cases, whereas only 40% of the public cases were performed in the South East region. These findings imply that if a cartilage registry is developed, an important consideration is whether to include hospitals from several health regions in addition to private hospitals. Furthermore, the data demonstrate a significant reduced frequency of advanced cartilage surgery for patients treated at private institutions (p<0.001). It is not possible to outline whether this is a case of reduced accessibility, but it is likely that procedures leading to more overnight stays are less available at these institutions.
Similar differences between public and private hospitals are seen in other Scandinavian countries for meniscal surgery,28 and these differences might also be due to financial incentives. Codes for palliative procedures were mainly in use for middle-aged patients in combination with M17. It has been previously demonstrated in studies that debridement is no better than sham surgery20 or rehabilitative training with a Physiotherapist,21 whereas the latter also failed to show the efficacy of surgery in patients with mechanical symptoms. These studies changed the trends in surgery on patients with OA as the rates of arthroscopy declined in the following years, at least in the USA.29 It is possible that a larger part of these procedures is now performed on patients with actual FCDs, although these procedures are also still used in patients with knee OA. On the basis of the recent literature, this type of surgery should be abandoned.
Few studies have explored incidences of cartilage surgery, whereas one study presents national numbers on cartilage injuries diagnosed with arthroscopy.24 Two studies presented remarkably different numbers based on data from the PearlDiver database in the USA. Montgomery et al23 report an incidence rate of 1.27–1.57/10 000 (2004–2009) patients and McCormick et al22 report an incidence rate of 90/10 000 (2004–2011). McCormick seems to calculate incidences on the basis of all individual patients within the database, whereas Montgomery calculates incidences on the basis of all patient records, which may explain the different results. Our incidence rates are within the same range as those reported by Montgomery et al when compared to the number presented in the articles. However, when we recalculated new incidence rates on the basis of the numbers provided by the two articles and applied the same approach as used in this study, we found quite different incidence rates from both articles. Consequently, the incidence rates from this study then appear in the vicinity of McCormick et al (table 4). Both studies focused on cartilage surgery only, and excluded patients with simply the diagnosis of an FCD or patients undergoing osteotomy in the absence of knee OA. These two subgroups accounted for <10% in this study and were excluded when comparing incidence rates for the years 2008–2011 (table 4). The same table displays the numbers from the Danish study, which are in close range with the numbers from this study.
Trends
We found that 56 hospitals performed cartilage surgery, whereas 15 hospitals operated <10 cases throughout 2009. Katz et al30 found that patients operated in low-volume hospitals by low-volume surgeons had worse functional outcomes 2 years after total knee replacement. When performing procedures that have failed to prove efficacy, the volume of the operating surgeons means less. However, this is a field with many patients and presumably low evidence-based adherence. Cartilage surgery is a complex treatment where several options exists, indicating that the availability of several techniques as well as an optimised rehabilitation programme is needed. In order to form a standardised treatment for as many patients as possible, each hospital or surgeon probably needs to see a certain, but not yet defined, number of patients yearly to maintain adequate quality of care. A discussion on whether to make specific cartilage centres must be made.
This study cannot explain the reasons for the geographical differences, but possible factors might be differences between the orthopaedic surgeons’ personal preferences and experience more than differences in the patient populations. A study aiming to describe the practice of MF among Canadian orthopaedic surgeons found widespread variation concerning indication for surgery.31 A patient's willingness to undergo surgery is also an important consideration and is higher in areas with an already high incidence of surgery.32
Knee cartilage surgery consists of several different techniques and although attempts on recommendations have been made, there is no gold standard treatment.9–16 ,33 MF is traditionally chosen for smaller defects, whereas OAT and ACI are chosen for larger defects.34 More specific recommendations do not exist, and we know little of the decision-making for surgical technique other than the size of the lesion and the patient's age. We do not have data on the size or location of the lesions in this study. CP is the most common procedure in our material and is performed for both FCDs and in knees with developing degenerative changes. The study by Montgomery et al23 found that MF and CP are the preferred procedures in 98% of cases with cartilage surgery. These procedures constituted 71.1% of all procedures in our material. The study by Mor et al24 found repair procedures (MF, osteochondral transplantation or chondrocyte transplantation) to be performed in 16.7% of the cases. The trends from the articles of Montgomery et al23 were significantly different from the trends of our material when compared with a χ2 test (p value<0.001). The difference was still significant after excluding the groups which had no cartilage surgery or osteotomies. Also, the trends in procedures from the study by Mor et al were different from the trends of this study with a lower proportion of palliative procedures, also after excluding the cases with no cartilage surgery or osteotomies.
Limitations
The ICD-10 codes available for diagnosing FCDs do not reflect the complexity of the clinical situation of these lesions. The distinction between focal lesions that are traumatic or degenerative is often difficult clinically, and location, size and depth matter greatly. The ICD-10 does not account for these conditions, and a distinction based on these codes is impossible. Although the ICD-10 contains both ‘acute FCD’ (S83.3) and several codes for knee cartilage pathology, there are no codes for the common ‘non-acute FCD’, which might be subacute or chronic. Our predefined codes matched with 92.3% of the reported diagnostic codes from the Norwegian Arthroscopic Association. However, the response rate was only 13.2%. The low response rate has limited effect on our final numbers since we have included most of the possible codes from the ICD system, but these challenges coexist with the fact that some orthopaedic surgeons might not code for FCDs at all if other intra-articular pathology is recognised. This is probably the largest limitation and cannot be defeated by any methodological changes, but by the information and education of orthopaedic surgeons. This is therefore a challenge concerning cartilage pathology and the ICD system and is as such a problem for the entire research field and not only for this study.
Among 11 566 ICD-10 codes, there are 789 coded as S83.3. The frequency of M17 codes increases with age; however, several orthopaedic surgeons have stated that they use M17 also for focal lesions. The inclusion of patients with an M17 diagnosis might lead to an overestimation of surgery for cartilage injury. However, an exclusion of these would definitely lead to an underestimation. This study reports a lower portion of palliative procedures than the Danish study24 (where they excluded all patients with OA), which might imply that most of those included in this study are actual knee cartilage defects and not OA.24 We did not include the ICD-10 code for ‘painful joint’ (M25.5) which might have underestimated the results.
The patient records or surgical protocols are considered the gold standard. However, large administrative databases allow the process of data collection to be efficient, detailed and precise, within its limitations. The Norwegian healthcare system is public and tax funded, which balances out possible geographic or socioeconomic differences. Studies have demonstrated that numbers extracted from electronic databases are being both overestimated and underestimated. Lofthus et al35 found that the Norwegian NPR overestimated hospitalisation for hip fractures by 29%, although the number of those having surgery for hip fractures was underestimated. Readmissions due to the same hip fracture were registered as a new hospitalisation for a new hip fracture by the NPR, which inflated the number. In our material, 297 cases (4.9%) were duplicates and only 73 procedures (0.67%) were classified as reoperations. We believe that procedure codes are reported in more detail as they are the basis of 60% of the government reimbursement in Norway and, as such, are reviewed several times by hospital controllers to ensure correct coding. For this study, we were interested in the burden of cartilage surgery and a combination of diagnostic and procedure codes seemed most appropriate.
The validity for the Norwegian NPR database was later assessed in a national study on hip fractures and the accuracy was found to be 98.2% (CI 96.5% to 99.9%) when diagnostic codes were combined with procedure codes.36 In that same study, the authors suggested possible coding errors from fractures that were treated conservatively or from patients that were admitted to hospital with such a fracture, but died before the operation. This does not apply to this study, as the diagnosis is set during the operation. The study by Mor et al24 assessed the validity against surgical descriptions in the medical records as the gold standard and found the positive and negative predictive values to be 88% and 99%, respectively. As for all studies with inclusion based on surgical procedures, FCDs diagnosed with MRI and treated conservatively are not included. An underestimation or overestimation might exist; however, the main goal of this study was to estimate the nationwide burden of cartilage surgery with the numbers available in NPR.
Future clinical implications
Cartilage surgery concerns a large and severely troubled patient group with no gold standard treatment. No nationwide surveillance currently exists to study the efficacy or effectiveness of treatment for this patient group. Development of a cartilage registry emphasising cartilage treatment being palliative, reparative or regenerative, in addition to non-surgical procedures, will be essential for clinical progression in this field.
Our numbers indicate that CP or debridement is still performed in degenerative knees.
Conclusion
In Norway, there are 2500 annual procedures classified as cartilage surgery, resulting in an age-adjusted incidence rate of 68.8/100.000 inhabitants. There are large variations between the different regions and between public and private hospitals.
This illustrates the need for a larger surveillance database for evaluation of results and calculation of costs in order to secure high quality treatment for all knee cartilage patients.
Acknowledgments
The authors would like to thank Ingar Holme, PhD for his advice concerning relevant statistics.
References
Supplementary materials
Supplementary Data
This web only file has been produced by the BMJ Publishing Group from an electronic file supplied by the author(s) and has not been edited for content.
- Data supplement 1 - Online supplement
Footnotes
Contributors LE and AÅ conceived the study, whereas all the authors planned the method. CNE was responsible for data collection and analyses, whereas all the authors participated in the interpretation. CNE drafted the manuscript, whereas AÅ and LE revised it. All the authors approved the final version.
Funding The study was supported through grants from the Oslo Sports Trauma Research Centre (OSTRC), where the corresponding author is employed. The South-Eastern Norway Regional Health Authority, the Royal Norwegian Ministry of Education and Research, the Norwegian Olympic Committee & the Confederation of Sport and Norsk Tipping finance the centre. The Faculty of Medicine, University of Oslo (UiO) also supported the work as CNE previously was a student at the Medical Student Research Programme at UiO and thereby received financial support from the Research Council of Norway.
Competing interests None declared.
Provenance and peer review Not commissioned; externally peer reviewed.
Data sharing statement No additional data are available.