The longitudinal study of patients with CIN has shown that the EQ-5D-5L was responsive to change in health after surgery, but the effect size was small. And the enhance of UI and EQ VAS after treatment were at least 0.039 and 5.35 which can be considered an improvement in health from patients’ perspective. However, the MCID estimated in this study can only represent truly meaningful change of HRQoL score at group level instead of at individual level.
As results shown, the anxiety/depression dimension was the one with the most improvement in scores after surgery, and the only one with statistically significant of score change. This is similar to the results of longitudinal HRQoL assessment of CIN patients by Xie et al. 1 month after treatment, the average improvement in mental component summary scores (MCS) measured by SF-36 qustionnaire was higher than that of physical component summary scores (PCS) (△MCS:7.05 vs. △PCS:1.47) [30]. A possible explanation is that, in general, CIN does not produce symptoms or signs that affect patients' ability to perform, whereas the CIN diagnosis has a negative psychological impact [2, 3]. Howbeit, the psychological support of doctors, examples of good prognosis of patients after treatment, and increased patient awareness of disease may ameliorate psychological.
In all patients, the positive changes in UI and EQ VAS were also found in other studies. A prospective study in Chinese CIN patients conducted by Zhao et al. expressed that the EQ-5D scores of 1 month after treatment was significantly better than the baseline [31]. Therefore, we considered that changes in health of CIN patients caused by surgical treatment can be qualitatively judged by the score change of the EQ-5D-5L. Interestingly, the UI and EQ VAS of patients whose response to GRCQ were "improvement" increasing by 0.039 and 9.27, and were statistically significant. Different result was presented in patients of “About the same”, UI and EQ VAS raised 0.010 and 4.37, but not significant. Bilbao et al. found that the mean change of the EQ-5D-5L score was positive in “improved” hip or knee osteoarthritis patients who underwent surgery [32]. It can be seen that the judgment result of health change from the patient's point of view through GRCQ was consistent with score change of the EQ-5D-5L, a multi-dimensional and multi-attribute questionnaire, even if the GRCQ has only one question. Thence GRCQ is a simple and credible choice for determining whether or not the health change when multiple items questionnaire cannot be used.
In this study, two of the most commonly used indicators of responsiveness, ES and SRM, were used to estimate the degree of change in patient health [33–35]. The result denoted that the effect size of the EQ-5D-5L in the total sample was only between small and moderate. Similar results have appeared in previous studies. Chen et al. assessed the responsiveness of the EQ-5D-5L in 65 Taiwanese patients with stroke underwent rehabilitation, and the effect size ranged from 0.40 to 0.63 for UI and 0.30 to 0.34 for EQ VAS, which suggested small to moderate responsiveness [36]. Furthermore, the effect size of UI was only 0.20 in patients with cataract surgery [37]. Another study in the obese patients showed that UI and EQ VAS had only small responsiveness to bariatric surgery [38]. These findings suggest that the EQ-5D-5L is responsive to various disease conditions, which can explain that changes in health were clinically relevant, rather than random errors, but small responsiveness was noteworthiness. The reason may be that the study population with chronic diseases. The loss of health in patients with chronic disease is a slow and long-term process compared to that patients with acute disease and recovery rapidly, their perception of changes in health may not be as strong as in acute illness, so the small responsiveness they have.
Some researchers believe that the responsiveness may depend on the direction of changes in health states and the health states at baseline [39], and this theory was supported by the results of this study. When we evaluated the responsiveness in different change directions of health states, we found that the moderate responsiveness of UI and EQ VAS in patients with improved health states, while low or no responsiveness was found in patients with no change. In addition, the baseline scores of UI and EQ VAS in “improvement” surgical patients were lower than those with “About the same”, while score change was higher than latter. Statistically, the responsiveness of patients with improved health states must be better than that of “About the same” patients.
Responsiveness of the EQ-5D-5L in patients with improved health states was also studied in other populations, but results were inconsistent. In acute asthmatics patients who underwent 1 month of treatment and self-reported improvement in health states, the UI had moderate to large responsiveness with effct size of 0.63 to 0.95 [40]. Golicki et al. revealed that the EQ-5D-5L was consistently responsive in stroke patients with improved health 4 months after treatment, which UI showing moderate ES (0.51–0.71) and moderate to large SRM (0.69–0.86), and EQ VAS range from 0.51 to 0.65 for ES and 0.59 to 0.69 for SRM [41]. Another study in patients with osteoarthritis 6 months after surgery showed that patients with improved health states had 1.48 of ES and SRM in UI, 0.82 of ES and 0.90 of SRM in EQ VAS [32]. Through the above, we found that although the responsiveness of “improvement” patients was good, at least moderate, the effect size of each study was quite different. The source of the difference may be different characteristics of subjects or different time intervals for two measurements [36]. Because longer time intervals allows sufficient time to respond to the physical condition, this is reflected in larger score change, resulting in larger effect size to reflect the degree of change in health upon full recovery, and vice versa [22].
MCID, as a clinically significant score change threshold of questionnaire, has become a key issue in questionnaire application. There is still no consensus on best method for estimating MCID [42], but the distribution-based and anchor-based methods are commonly used [43, 44]. As known, the anchor-based method can provide professional explanation of clinical significance for MCID, while the distribution-based method estimates MCID based on the sample variation and the precision index of the instrument, which makes the estimation greatly affected by the measurement characteristics of instrument itself and the sample size, so that the former is preferred, and the latter is used as a supported method [39, 45]. Previous studies taken the mean change of scores as MCID of the anchor-based method [46, 47], but it does not take into account the possible impact of the HRQoL score over time on MCID in patients who reported no change of health during follow-up [24]. However, the absolute value of score change of the person with “A little better” minus the score change of the person with “About the same” was used as the MCID in this study, eliminating the potential impact of the time on the MCID estimation.
Besides the distribution-based and anchor-based methods, the instrument-defined method also used to triangulate the MCID. The instrument-defined method is only relevant to preference-based measurements, such as the EQ-5D-5L, and the MCID estimation is completed on the basis of the simulated transition of health states [27]. The greatest advantage of this approach is that a single-level transition of each baseline health states can act as a reference point or standard for minimally important change, resulting in MCID based on multiple internal anchors [27]. Luo et al. used the instrument-defined method to estimate the MCID for the EQ-5D-3L, and the result was parallel to the published estimate, so the instrument-defined method was regarded as an effective method for MCID estimation [27]. As shown in our results, the MCID obtained by the instrument-defined method differs from the anchor-based method by 0.023, and the difference between the distribution-based method and the anchor-based method was 0.014 to 0.018. So we deem that the instrument-defined method can be used for the MCID estimation of the EQ-5D-5L in CIN patients.
The result of the study on the relationship between MCID and MDC demonstrated that the MCID estimated for UI and EQ VAS by the three methods can at group level explain that score change was owing to change in health rather than measurement error. However, MCID of UI and EQ VAS both can not account for the health change of individual at 95% confidence level, possibly due to the inclusion of patients with different histopathological. In this study, the proportion of patients with carcinoma in situ was 22.00%, although it belongs to CIN [1], compared with other pathological grades, it was at higher risk of progression to invasive cancer [48], and patients had lower psychological expectation of changes in health, so the different criteria for patients to judge health change may cause this result. Another possible explanation may be that, although patients included were all first diagnosed, the HRQoL scores at baseline of some patients with disease duration of more than 0 may be improved compared to those at the beginning of the diagnosis, resulting in the baseline score of the entire sample being raise. Therefore, the possibility of underestimating MCID in this study leads to its being less than MDC95%(ind). The result of the study could be further validated in patients with the same pathological grade and the same disease duration.
This study has some advantages. Firstly, using combination of qualitative and quantitative approaches to assess responsiveness increases the credibility of the results. Secondly, in addition to the distribution-based and anchor-based methods, using of the instrument-defined method for MCID estimation would helpful for obtaining reasonable result. Next, we analyzed whether the MCID estimated by each method can reflect the true health change in patients with individual and group levels to determine the reliability of MCID and avoid the wrong application and interpretation of MCID. Although judging whether the MCID is different from the measurement error is considered a logical next step after MCID estimation [29], only a few studies have been performed [49, 50]. Finally, there were no measurement biase from investigator because of two time points survey for each patient were performed by the same investigator.
There are also several limitations. The disease-specific questionnaire was commonly used anchor besides the GRCQ in previous studies [23, 51]. This study did not simultaneously perform the measurement of the disease-specific questionnire, FACIT-CD, of CIN, since there is no chinese version so far [52]. Although GRCQ has only one question, it is the accepted anchor for MCID estimation at this stage [16]. Studies have shown that if health state changes in different directions, the MCID may also be different [53], whereas because there were no patients who responsed to GRCQ with “worsen” in this study, MCID could not be estimated for this group of patients. Future studies could develop MCID for CIN patients with worsen health states to determine whether it is differ from patients with improvement. It is well known that MCID changes on account of demographic characteristics, interventions and so on [44, 54], so the results of the study cannot be generalized to other clinical settings. Furthermore, small sample size may affect the accuracy of MCID, though it has met the basic requirements for MCID estimation [55].