Most of the tables reported in this paper are for all patients, irrespective of the knee replacement system that was implanted, except where the point of the analysis is to show how comparisons between different procedures may be presented. Apart from these exceptions, comparisons between Attune and Sigma patients are noted, and a data supplement contains the separate tables for them.
Dimensions and levels
Overall quality of life
Figure 1 shows the number of respondents (including both Attune and Sigma patients, N = 1587) in each level of each EQ-5D-3L dimension at different observation points, for patients who completed all 3 questionnaires. Numbers and percentages are shown in Additional file 1: Table S1. The distributions of pre-surgery levels give a picture of the amount of unmet need for patients undergoing surgery. Pain & Discomfort was the most frequently observed problem: 1.6% of patients reported no problems (Level 1), 27.2% severe problems (Level 3) and the remainder some problems (Level 2). Usual Activities was the second most frequent: 13.8% no problems, 7.8% severe problems and the remainder some problems. The mobility dimension also had less than half of patients reporting no problems (16.4%), 83.3% had some problems, but only 0.2% reported severe problems. This is likely to be due to a known limitation of the EQ-5D-3L Mobility Level 3, which labels this “Confined to bed” . This ceiling effect issue has been corrected in the EQ-5D-5L , where the most severe problem with mobility is “Unable to walk about” and there is an option for “Severe problems in walking about”. More than half of patients reported no problems for Anxiety & Depression (61.4%) and Self-care (79.3%), and relatively small numbers reported extreme problems (3.1% and 0.4%); in each case the remainder had some problems.
The distributions of post-surgery levels show marked improvements in every dimension. However, the point at which this was observed differed between dimensions. For Mobility, Self-Care and Anxiety & Depression, most of the improvement was seen in the first 10 months (pre-surgery to 10 months or less), with small further improvements in the following year (10 months or less to 11–22 months). For Usual Activities and Pain & Discomfort, there was a large improvement in the first 10 months and a further large improvement in the following year. In both cases, the second-year improvements were an increase in the number of patients reporting no problems, rather than a reduction in the number of patients reporting severe problems.
For Pain & Discomfort, the most frequently observed problem pre-surgery, 29.6% reported no problems during the first 10 months after surgery, rising to 56.0% in 11–22 months. Only 2% had severe problems in 11–22 months. Most of the improvement in Pain & Discomfort occurred in the first 10 months. There was a similar pattern for the second most frequently reported problem, Usual Activities, with no problems rising to 54.6% during the first 10 months and to 70.1% in 11–22 months, and severe problems falling to 1.9% and subsequently to 0.8%. For Mobility, 71% had no problems during the first 10 months, rising to 78.8% in 11–22 months.
Although there were large improvements in the dimensions most associated with the underlying condition, post-surgery there remained (even in 11–22 months) a significant number of patients experiencing problems with Mobility, Usual Activities and, in particular, Pain & Discomfort. For the other two dimensions, there were also large improvements, although from a lower level of pre-surgery problems. 5.6% of patients reported problems with Self Care post-surgery, and 16.0% with Anxiety & Depression; these may represent an underlying level of problems consistent with the general population.
Comparison of attune and sigma patients
Additional file 1: Tables S2 and S3 show the overall pattern of levels within dimensions and changes over time for Attune and Sigma patients, which were very similar to each other, with the exception of Mobility. Both pre- and post-surgery, Sigma patients had fewer mobility problems than Attune patients, reflecting a slightly different pre-surgery case mix. Pre-surgery, more Attune (87.5%) than Sigma (77.4%) patients reported Level 2 problems and three Attune patients reported Level 3 problems, compared with none for Sigma. At 11–22 months, the percentages with Level 2 problems were 23.8% and 17.4% respectively; there were no patients with Level 3 problems in either the Attune or Sigma group.
Table 1 compares between Attune (N = 937) and Sigma (N = 650) patients the increases in the percentage of patients reporting no problems within each dimension in the first- and second- years post-surgery compared with pre-surgery. It shows that in every dimension, the improvement was greater in the second-year post-surgery for Attune than Sigma, including Mobility. Except for Usual Activities, the same is the case in the first-year post-surgery.
Changes in levels within dimensions
Table 2 shows the pattern of changes between the different time points (N = 1587). There are three possible start and end levels, so there are 9 possible pathways of change within each dimension. These can be divided into improvement pathways (2–1, 3–1 and 3–2), worsening pathways (1–2,1–3 and 2–3) and stable pathways (1–1, 2–2 and 3–3). In the table, improvement pathways are highlighted in bold and worsening pathways in italics. The first pair of data columns show the number and percentage in each change pathway, measured from the pre-surgery baseline to the first post-surgery assessment (< 1 year). The second pair show the changes from the pre-surgery baseline to the second post-surgery assessment (> 1 year). The third pair show the changes from the first post-surgery assessment to the second. The third pair are therefore not directly comparable with the other pairs, because they start from a different baseline.
This table adds more detail to the finding that most improvements occurred in the first-year post-surgery, although in some dimensions there are also substantial improvements in 11–22 months. In every dimension there were few patients in worsening pathways by 11–22 months: the largest percentage was in Anxiety & Depression, where 73 out of 1587 patients were in these pathways (4.6%); the others were in decreasing order Self Care (2.33%), Usual Activities (2.27%), Pain & Discomfort (1.13%) and Mobility (0.6%). Moreover, few patients who had no problems pre-surgery developed problems after surgery. In every dimension the majority of patients who had at least some problems (that is, Level 2 or Level 3) 11–22 months post-surgery had also had at least some problems in that dimension pre-surgery: in decreasing order Pain & Discomfort (98.9%), Mobility (97.0%), Usual Activities (95.1%), Anxiety & Depression (76.7%) and Self Care (65.5%).
Additional file 1: Tables S4, S5 and S6 compare improvement and worsening pathways between Attune and Sigma. In the first 10 months post-surgery there was a mixed pattern of the relative numbers of patients in improvement pathways and worsening pathways; however it should be remembered that there were few patients in worsening pathways in either group, so the differences between them are small. For Self-Care, there was a slightly larger percentage (1.9%) of Attune patients in improvement pathways and a slightly smaller percentage (− 0.86%) in worsening pathways; for Mobility and Anxiety & Depression there were also more Attune patients in improvement pathways (3.2% and 2.8%) but also more in worsening pathways (0.17% and 0.18%); for Pain & Discomfort there were fewer Attune patients in both improvement (− 1.4%) and worsening (− 0.54%) pathways; and for Usual Activities there were fewer Attune patients in improvement pathways (− 1.9%) and more in worsening pathways (1.8%). However, after 11–22 months, there were in every dimension a slightly greater percentage of Attune patients in improvement pathways and a slightly smaller percentage in worsening pathways; for example, 3.4% more improving and 0.76% fewer worsening in Mobility.
Analysing changes in profiles
Looking at how levels within dimensions change in the distribution of EQ-5D-3L health states is valuable, but it does not give a picture of how individual patients’ health states as a whole change. The Paretian Classification of Health Change (PCHC)  summarises these by classifying a patient’s health state into four categories:
Improve – an improvement in at least one dimension and no deterioration in any other dimension. For example, pre-surgery 21321; post-surgery 21221.
Worsen – a deterioration in at least one dimension and no improvement in any other dimension. For example, pre-surgery 21321; post-surgery 31321.
No change – no change in any dimension. For example, pre-surgery 21321; post-surgery 21321.
Mixed – an improvement in at least one dimension and a deterioration in at least one dimension. For example, pre-surgery 21321; post-surgery 22221.
Table 3 shows the distribution of change from pre-surgery to the first 10 months (N = 1789) and 11–22 months (N = 1627) post-surgery. 76% of patients improved over the first 10 months, rising to 85% in 11–22 months. This improvement of the patient group as a whole resulted from a reduction in all of the other three categories: the number who worsened fell from 7.5% in the first 10 months to 3.5% in 11–22 months; those whose quality of life did not change fell from 9.2 to 5.9%; and those who had a mixed change fell from 7.8 to 5.5%. Additional file 1: Table S7 compares the PCHC between Attune and Sigma patients. At both first- and second- years post-surgery a greater percentage of Attune patients improved and fewer had a mixed change, and by 11–22 months fewer also worsened and more had no change.
The characteristics of those who either worsened or experienced no change according to the PCHC were further investigated by examining their profiles and for those who worsened the specific dimensions in which they worsened.
For patients who experienced no change, there was no evidence that this was related to any particular profile. The most common profiles for such patients, both Attune and Sigma, were those that were most common in patients as a whole pre-surgery (see analysis of profiles below).
For patients whose health state worsened (N = 134 to < 11 months, N = 57 to 11–22 months), Table 4 shows the number of dimensions in which they became worse. In both the first- and second- years post-surgery, of the patients who worsened, the large majority worsened in one dimension only (75.4% and 70.2% respectively). Although the number of patients who became worse in more than one dimension fell to almost half (33–17) in 11–22 months, they formed a slightly larger percentage of the total (24.6–29.8%). Additional file 1: Table S8 compares Attune and Sigma patients. In the first 10 months there were slightly more Attune patients with larger numbers of worse dimensions (25.2–23.5%), but in 11–22 months there were far fewer (20–40.7%).
Table 5 shows for these patients who became worse overall the dimensions in which they worsened. In both the first- and second- years post-surgery, the smallest numbers were in those dimensions in which there were most problems pre-surgery; Mobility and Pain & Discomfort. Although there were differences between the other dimensions in the number of patients in the first 10 months, in 11–22 months they were the same. Additional file 1: Table S9 compares Attune and Sigma patients. Attune patients followed the same pattern as patients overall, but there were some differences for Sigma patients. The relatively small numbers in each patient group make detailed interpretation difficult, but for Sigma patients Pain & Discomfort worsening in the first 10 months post-surgery was as prevalent as in other dimensions apart from Mobility. The reduction in numbers worsening in both Mobility and Pain & Discomfort was much less apparent for Sigma patients, such that the percentage who worsened in Mobility, which was much smaller in 11–22 months compared with the first for Attune patients (12–6.7%), rose (11.8–18.5%).
Health profile grid
The Health Profile Grid  compares each patient’s profile at baseline with their profile at follow-up, in this case at 11–22 months (N = 1627). The profiles are arranged in rank order from best to worst defined by a value, in this case the US Time trade off (TTO)-based Value Set. Figure 2 shows these pairings. The black diagonal line shows the points at which baseline- and follow-up profiles are the same, in other words the patient experienced no improvement or deterioration in any dimension of the EQ-5D. Points above the line represent improvements, and points below a deterioration. (Note that gaps between apparent clusters of patients and points apparently in straight lines are artefacts of the method and have no real interpretations in terms of differences between patients.) The points are also identified by their change category according to the PCHC, which in effect identifies those who had a Mixed Change.
As expected, a large majority of patients (1439/1627 = 88.4%) are above the line. There are more patients who experienced no change (96/1627 = 5.9%) than a deterioration (92/1627 = 5.7%). Additional file 1: Figure S1 compares Attune and Sigma patients. A larger percentage of Attune patients improved (89.3% Attune, 87.3% Sigma) and had no change (6.2% Attune, 5.5% Sigma), and therefore a smaller percentage deteriorated (4.5% Attune, 7.3% Sigma).
Table 6 shows the most common EQ-5D-3L profiles observed pre-surgery and at follow up for patients who completed questionnaires at both < 11 months and > 1 year follow-up as well as pre-surgery (N = 1587). As is usual with EQ-5D-3L data, a small proportion of all of the possible profiles describe a large proportion of the observations. The table shows the most frequent profiles that describe 90% of the observations.
Pre-surgery, 4 profiles accounted for 50% of all observations and 17 for 90% of them, all of which include some or severe Pain & Discomfort. The most common, accounting for over a quarter of all patients, is 21221: some problems with Mobility, Usual Activities and Pain & Discomfort and no problems with Self Care and Anxiety & Depression. The other profiles differ by having a more severe level in one or more dimensions, though four of them have a less severe level in one or both of Mobility and Usual Activities.
Post-surgery, 12 profiles account for 90% of observations at both less than and greater than 11 months follow up; these are the same profiles at each time point with one exception amongst less common observations – a more severe condition, 22221, replaced by less severe, 21211 – and other more severe conditions are relatively less frequent. The profiles are even more concentrated than pre-surgery. The most common profile 11,111 – no problems in any dimension – covers 22.1% and 45.1% of observations at less than and greater than 11 months; at least 50% of observations are covered by 3 and 2 profiles respectively.
Additional file 1: Tables S10 and S11 show the most frequently observed pre-surgery profiles for Attune and Sigma patients. These were mostly the same, except for some of the less common profiles, 21122 and 21223 (in Attune but not Sigma) and 11222 (Sigma but not Attune). However, more severe profiles were relatively more frequent in Attune than Sigma; the top 50% for Sigma included the less severe profile 11221 instead of 22211 and the less severe profiles 11221 and 11121 were relatively more frequent in Sigma (16% of all profiles) than Attune (8%). This finding is consistent with that from “Results” section, that there were slightly greater mobility problems amongst Attune patients pre surgery.
Post-surgery, the most common profiles were again very similar. Both had 11111 and 11121 covering 50% of patients, although a slightly greater percentage of Attune patients were in the ‘full health’ profile 11111. In contrast to the pre-surgery distribution, where there were relatively more severe profiles amongst Attune patients, post-surgery more severe profiles were slightly more frequently observed amongst Sigma patients.
Combining EQ-5D descriptive system and EQ VAS data
As well as data in the form of the EQ-5D descriptive system, the EQ-5D questionnaire provides data in the form of the EQ VAS. Using these data together can provide further insights into respondents’ HRQOL. Two examples are given here: the relationship between profiles and the EQ VAS, and the relationship between change categories and the EQ VAS.
The relationship between profiles and the EQ VAS
Information on the distribution of levels and dimensions over a patient population demonstrates their relative importance in terms of prevalence within that population. An equally valuable question is how important they are to patients. Examining the relationship between the profiles and the EQ VAS gives an indicator of this.
Regression analysis using Ordinary Least Squares was used to estimate the impact of each level and dimension of the EQ-5D on the EQ VAS. Table 7 summarises the results, pre- and > 11 months post-surgery (N = 1624). There were so few observations for Level 3 (extreme problems) on Mobility and Self Care that it was not possible to identify an impact separate to that of Level 2 (some problems); for Pain & Discomfort pre-surgery it was only possible to identify Level 3 effects. All other levels and dimensions had the expected pattern, that is their coefficients were negative and their Level 3 coefficients were greater than Level 2 coefficients.
Although the relative sizes of coefficients were very similar pre- and post-surgery, as described below, the size of all of the coefficients that were significantly different from zero were higher post-surgery. The goodness of fit, as measured by R2, was also higher.
The dimension that had the most impact overall on measured quality of life was Anxiety & Depression. It had the largest Level 2 coefficient both pre- and post-surgery, and Level 3 was the largest pre-surgery and the second largest post-surgery. Mobility, Self-Care and Usual Activities had similar coefficients at Level 2 both pre- and post-surgery, but post-surgery the Usual Activities Level 3 coefficient was much higher than pre-surgery and was the largest of the post-surgery coefficients. The Pain & Discomfort coefficients were consistently the smallest of the significant coefficients, at both Level 2 and Level 3 and both pre- and post-surgery.
Additional file 1: Tables S12, S13 and S14 compare Attune and Sigma patients. A dummy variable for knee replacement system included in the all-patients analysis had a coefficient in each year’s equation that was not significantly different from zero and did not change goodness of fit as measured by adjusted R2. However, analysing Attune and Sigma patients data separately, the relative sizes of the coefficients, both for dimension levels and for pre- and post-surgery, were in some cases different. Two examples are that Mobility at Level 2 had a slightly smaller coefficient for Attune patients post-surgery than pre-surgery, but for Sigma patients it was very much larger; and pre-surgery Self-Care had the smallest coefficient for Attune patients, for Sigma patients it was Mobility.
The relationship between change categories and the EQ VAS
The Paretian change categories provide a simple indicator of how successful are interventions such as a TKA. It is valuable to explore factors that might affect that success. A proper assessment of this would only be possible by examining a wide range of factors affecting patients before, during and after surgery, but using the pre-surgery EQ VAS gives an insight into one factor, the severity of the patient’s condition as measured by their reported HRQOL.
The distributions of pre-surgery EQ VAS values for the four change categories are very similar in terms of summary statistics (pre- and > 11 months post-surgery, N = 1624). Means, with standard deviations in parentheses, are Improved 73.3 (17.3), Worsened 72.7 (17.4), No change 74.0 (17.5) and Mixed Change 72.1 (16.3). This can be further explored by visual examination of the distributions, as in Fig. 3. Each dot represents one patient. It shows that those who improved recorded pre-surgery EQ VAS values over the full range of possible values. For the other categories, there was a similar spread over the range of possible values, although with a lower minimum value, because for the most severe conditions, there is less scope for worsening.
Rights and permissions
Open Access This article is licensed under a Creative Commons Attribution 4.0 International License, which permits use, sharing, adaptation, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons licence, and indicate if changes were made. The images or other third party material in this article are included in the article’s Creative Commons licence, unless indicated otherwise in a credit line to the material. If material is not included in the article’s Creative Commons licence and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder. To view a copy of this licence, visit http://creativecommons.org/licenses/by/4.0/.