Background Placebo effect is one of the methodological difficulties in dry eye clinical trials. If we could elucidate the tendencies of the placebo response and find predictors, we could reduce the placebo response in clinical trials for dry eye. In this study, we investigated the predictive factors for the placebo effect in dry eye clinical trials.
Methods A total of 205 patients with dry eye assigned to the placebo arms of three placebo-controlled randomised clinical trials were analysed by simple and multivariable regression analysis. The corneal fluorescein (FL) staining score and dry eye symptoms were studied at week 4. The variables of interest included gender, age, complications of Sjögren’s syndrome, Schirmer’s test I value, tear break-up time and conjunctival hyperaemia score. We also conducted a stratified analysis according to the patients’ age.
Results Among all the studied endpoints, the baseline scores were significantly related to the corresponding placebo response. In addition, for the FL score and the dryness score, age was a significant predictor of the placebo response (p=0.04 and p<0.0001, respectively). Stratified analysis by age showed that patients more than 40 years of age are more likely to have a stronger placebo response in the FL and dryness scores.
Conclusion The baseline scores and age were predictive factors of the placebo response in frequently used endpoints, such as FL score or dryness symptoms. These patient characteristics can be controlled by study design, and our findings enable the design of more efficient placebo-controlled studies with good statistical power.
- clinical Trial
Statistics from Altmetric.com
Dry eye is a multifactorial disease of the tear and ocular surface, which results in subjective symptoms, visual disturbance and tear film instability in addition to damage to the cornea and conjunctiva.1 The typical initial therapy for dry eye is artificial tears and the second recommendation is ophthalmic solutions of anti-inflammatory agents or secretagogues.2 Recently, the trend of drug development for dry eye has shifted from simple artificial tears to pharmacological solutions that stimulate natural production of tear constituents, maintain ocular surface epithelial health and barrier function, and inhibit inflammatory factors.2 A crucial issue in drug development is the choice of the primary endpoint in late-phase clinical trials. Regulatory agencies in some countries require efficacy in terms of both symptoms and signs,3 but in practice a wide range of endpoints, including both objective and subjective measures, have been used (eg, corneal fluorescein (FL) staining score, dryness score and other subjective dry eye symptoms).
Another methodological difficulty in dry eye clinical trials is placebo effects. Placebo responses in placebo-controlled trials are often attributable to regression to the mean, but placebo effects can be more pronounced in dry eye because a placebo involves adding water to the ocular surface with inherent therapeutic effects. Furthermore, as in other disease areas, mental effects might be added onto the real therapeutic effects. To reduce the placebo response during the treatment period, patients are administered a placebo prior to the randomisation visit in clinical trials.3–10 However, placebo responses were seen in these placebo run-in studies and we cannot yet control the placebo response in dry eye clinical trials.
It is well known that there are placebo responses in placebo-controlled trials of various diseases, including subjective endpoints and objective signs.11–13 When the effect of the placebo response is large, the statistical power is lowered, which might cause the failure of trials.14 15 The placebo response is typically considered as a combination of the ‘placebo effect’ and the “effect of being in the ‘no treatment group’”, which are difficult to distinguish in clinical trials.15 16 The former consists of both physical and mental effects, and the latter consists of regression to the mean or natural healing effects.
The aim of this study was to determine predictors including patient characteristics that specifically influence placebo response using the pooled data of placebo arms in three clinical trials for dry eye.
Study design and setting
We conducted a pooled data analysis of placebo response among patients with dry eye assigned to placebo groups in three clinical trials in Japan. The dataset was composed of three phase II clinical trials of Diquafosol (NCT01189032)5 and DE-101 (JapicCTI-090877 (phase IIa) and JapicCTI-090916 (phase IIb)) conducted by Santen Pharmaceutical Co., Ltd. All data were provided by Santen Pharmaceutical Co., Ltd. Full analysis set (FAS) data sets of each clinical trial were used in this study. The Ethics Committee of the Kyoto University School of Medicine approved the protocol for this study, and the study was conducted in accordance with the guidelines of the Declaration of Helsinki.
Participants in original clinical trials
All clinical trials were multicentre, placebo-controlled, double-masked studies conducted under the principles of good clinical practice. The three original clinical trials each had a 2-week placebo run-in period followed by a 6- or 12-week treatment period. During the run-in period, a placebo was administered four (DE-101) or six (Diquafosol) times per day. The placebo used in each study was the vehicle solution of the investigational drug. All clinical trials were approved by the institutional review board in each study site and written informed consent was obtained for all subjects enrolled in these studies. The enrolled patients were patients all with dry eye diagnosed by the Japanese diagnostic criteria, with Schirmer’s I test values <5 mm in 5 min. Study subjects were more than 20 years old and had mild-to-moderate FL scores (more than three points/maximum nine points in the Diquafosol and DE-101 phase IIa study, or five points/maximum fifteen points in the DE-101 phase IIb study) at the randomisation visit. Patients who had active allergic conjunctivitis or other ocular infections were excluded.
Primary and secondary outcomes in original clinical trials
The primary endpoint was the change from baseline of the FL score. Primary outcome measures were the FL score and subjective symptoms related to dry eye. The secondary endpoints were evaluations of subjective dry eye symptoms and fluorescein tear break-up time (BUT). The FL score was evaluated by a 9-point scale in the Diquafosol study and DE-101 phase IIa study, and a 15-point scale in the DE-101 phase IIb study. In the 9-point scale, the cornea was divided into three equal zones and each zone had a staining score ranging from 0 to 3 points. In the 15-point scale, the cornea was divided into five equal zones and each zone had a staining score ranging from 0 to 3 points (the National Eye Institute’s recommended scale). Subjective dry eye symptoms were scored across 11 parameters: foreign body sensation, photophobia, itching, eye pain, dryness, heaviness, blurred vision, eye fatigue, eye discomfort, eye discharge and lacrimation. These parameters were scored on a 4-point scale from 0 to 3 (0 no symptoms, 1=mild, 2=moderate, 3=severe) at each study visit by interviews with investigators.
Assessment of predictors of placebo effect
Definition of placebo response
We defined the placebo response as the change of the FL score and dry eye symptom scores from baseline to a selected time point. We selected week 4 or the last observation carried forward within 4 weeks, which was the primary endpoint in the Diquafosol clinical trial. For the FL score, the scores in the 15-point scale were multiplied by 9/15 to adjust to the 9-point scale. For dry eye symptom scores, the scores of each item and the total were evaluated.
We recorded the following patient characteristics: gender, age, complications of Sjögren’s syndrome (SS), Schirmer’s test I value, BUT and conjunctival hyperaemia score. In addition, we calculated the amount of score change from baseline in each of the FL score and the dry symptom score.
Simple regression analysis was used to narrow down the variables for multivariable regression analysis based on two-sided tests and an alpha level of 0.1. We evaluated the patient characteristics as mentioned above, the presence of pretreatment by dry eye medication and the difference in protocols (Diquafosol, DE-101 phase IIa and DE-101 phase IIb). After selecting significant variables in the univariate analyses, multivariable regression analysis was conducted to determine the predictive factors of placebo response for each variable: the FL score, the total dry eye symptom score and the individual score of each dry eye symptom. The identification of each protocol was also included in the model to determine potential differences in the protocols. The threshold for statistical significance in the two-sided tests was an alpha level of 0.05. Moreover, to verify the validity of the predictive variables, we conducted a stratified analysis of the FL score and dryness score by age. Statistical analysis was performed using JMP Pro V.12.0.1 (SAS Institute, 2015).
Demographics and baseline characteristics
Our study data set was composed of a total of 205 patients who were assigned to the placebo arm in three clinical trials (Diquafosol: 96, DE-101 phase IIa: 46, DE-101 phase IIb: 68, excluded from FAS: 5) (figure 1). The patients’ characteristics and dry eye parameters are shown in table 1.
The variables that were found to influence the placebo response of the FL score in the simple regression analysis included age, pretreatment by dry eye drugs, baseline FL score, dryness, lacrimation, eye discomfort and eye fatigue. In the multiple regression analysis, statistical significance was shown in age (β, −0.01 (95% CI, −0.02 to 0.00); p=0.039) and the baseline FL score (β, −0.20 (95% CI, −0.32 to −0.09); p<0.0001) (table 2).
Dry eye symptom total score
The variables that were found to influence the total dry eye symptom score in the simple regression analysis included baseline total dry eye symptom score and baseline conjunctival hyperaemia score. In the multiple regression analysis, there were no statistically significant factors.
Individual dry eye symptom scores
In each dry eye symptom score, 1–11 variables were found to influence the placebo response in the simple regression analysis. In the multiple regression analysis, the score of each dry eye symptom was significantly associated with its own baseline score. The other factors that were found to be significantly associated with the placebo response were ‘dryness,’ ‘heaviness,’ ‘lacrimation’ and ‘eye discomfort.’ For ‘dryness,’ age was significantly associated with the placebo response (β, −0.01 (95% CI, −0.41 to −0.22); p<0.0001) (table 3). For the other items, ‘heaviness,’ ‘lacrimation’ and ‘eye discomfort’ were influenced by ‘SS’ (β, −0.13 (95% CI, −0.19 to −0.06); p<0.0001), ‘the FL score’ (β, 0.03 (95% CI, 0.01 to 0.05); p<0.0001) and ‘the baseline score of foreign body sensation in the eye’ (β, 0.08 (95% CI, 0.00 to 0.15); p=0.045), respectively (online supplementary etables 1, 2 and 3).
Stratified analysis by age in the FL score and dryness score
We conducted a stratified analysis by age of the FL score and dryness score (table 4). The differences in FL scores before and after treatment were significantly larger in patients more than 40 years of age. There were no significant relationships in the dryness scores before and after treatment, although the amount of change tended to increase with ageing.
We studied the predictors of placebo response in dry eye clinical trials using multivariate linear regression models and found that high baseline scores and ageing affected the placebo responses of the FL score and dryness score. It is notable that age is a predictive factor beyond regression to the mean because these models were adjusted for the baseline scores. These placebo responses increased with age and they were more pronounced in patients aged 40 years and older. Ageing is a well-known key risk factor for dry eye. Aged populations have a relatively high incidence of dry eye.17 18 However, the influence of ageing in clinical trials is currently unclear. One of the age-related risk factors is the complication of conjunctivochalasis, which was reported to increase with ageing, starting from 40 years, at the nasal and temporal conjunctiva.19 20 Conjunctivochalasis decreases the tear clearance rate and increases the concentration of inflammatory cytokines.21 22 It can be speculated that the washout effect can improve eye conditions by decreasing the concentration of inflammatory cytokines. Thus, researcher should consider that treating with a placebo could improve the eye condition of patients with dry eye, especially among those aged 40 years and older.
Each symptom score was found to be significantly influenced by its own baseline score. Patients with high baseline scores had a high placebo response in each corresponding score. A possible cause is ‘regression to the mean,’ which has been observed in various clinical trials. Therefore, most improvements in the placebo arms might be instances of regression to the mean.11 Although it is considered to be caused by regression to the mean, this result is inconsistent with previous reports of psychiatric disorders such as depression or attention-deficit/hyperactivity disorder.23–25 In these diseases, disease severity is negatively correlated with the placebo response. The reason might be that the milder group has subjects enrolled by a false-positive diagnosis and that severe patients might be resistant to the placebo treatment. In contrast, dry eye requires objective signs for diagnosis; therefore, the false-positive rate might be low. In addition, the placebo might have a therapeutic effect in severe patients by acting as artificial tears. Considering the contribution of each baseline score to each placebo response, we should avoid setting strict cut-off values related to endpoints to prevent high baseline scores.
Interestingly, we also revealed that the baseline FL score was a significant factor affecting the lacrimation scores after the placebo treatment. In dry eye, it is known that subjective symptoms are not correlated with objective signs.26 However, this result suggests that lacrimation (the subjective symptom) might be related to the FL score (the objective sign). Further study will be needed to verify this relationship. We found similar relationships in other symptoms: between heaviness and complications of SS, and between eye discomfort and the baseline score of foreign body sensation. However, little is known about these symptoms, and further research is needed to understand more about these phenomena.
Our findings have some implications for the study design and data analysis of placebo-controlled clinical trials for dry eye. First, we should note that the placebo response might be large in severe patients and that the baseline scores of symptoms should be considered when analysing the data and interpreting the results. Second, the patients’ age must be considered when dealing with the FL score and dryness as endpoints. We suggest that stratified analyses should be conducted, with subjects aged more than 40 years analysed as a separate group. Furthermore, for a proof of concept study, restricting the study population to subjects aged less than 40 years might lead to a more precise evaluation of drug efficacy by suppressing the placebo effect. Third, it might be necessary to collect a detailed history of conjunctivochalasis. Finally, when analysing dry eye symptoms, the scores of individual symptoms should not be summed to create a single endpoint, that is, a ‘total symptom score.’ Some questionnaires involve the summation of separate symptom scores in their algorithm,27 28 and these questionnaires are frequently used as the endpoints for symptoms. However, we found that each symptom had different factors predicting placebo responses and that the total symptom score did not have any predictive factors. Therefore, it might be difficult to control the placebo response on total symptom scores indicating each symptom should be evaluated independently.
Our study had some strong points. The quality of this study data is high, with no missing data, as all the data were from clinical trials conducted with strict regulations and appropriate source data verification. We also evaluated various endpoints, including 11 subjective symptoms, which enabled us to discuss the fine details of signs and symptoms. However, this study had some limitations. First, because the data set we used consisted of data from a single sponsor and Japanese population, the generalisability of our findings might be limited. To mitigate this, we used a pooled placebo data set of three randomised clinical trials, rather than the data set of a single clinical trial. Second, information about concomitant diseases was limited. Further study will be needed to evaluate concomitant diseases other than SS, especially diseases that have known relationships with dry eye.
In this study, we identified baseline scores and age as predictive factors of the placebo response in dry eye clinical trials. Patients with mild-to-moderate symptoms or high baseline scores in the assessed endpoints and patients aged more than 40 years are more likely to have a high placebo response. Because these patient characteristics can be easily controlled via inclusion and exclusion criteria, our findings could enable the design of more efficient placebo-controlled studies with good statistical power.
Acknowledgements The data set used in this research was provided by Santen Pharmaceutical Co., Ltd. This research was supported by the Keihanshin Consortium for Fostering the Next Generation of Global Leaders in Research (K-CONNEX), established by Human Resource Development Program for Science and Technology, MEXT.
Contributors TI obtained the data, wrote the study protocol and the statistical analysis plan, cleaned and analysed the data, and drafted and revised the paper. IS, ST and KK analysed the data, and drafted and revised the paper. KK is guarantor.
Competing interests TI is an employee of Santen Pharmaceutical Co., Ltd. KK received research funds from Medical Platform Co., Novartis Pharmaceutical K.K. and Bayer; honorarium from Astellas, Novartis Pharmaceutical K.K. and Sanofi K.K. There are no patents, products in development or marketed products to declare relevant to those companies. No other disclosures were reported.
Patient consent This was the secondary analysis using anonymised data of clinical trials. The original clinical trials were conducted on Good Clinical Practice.
Ethics approval The Ethics Committee of the Kyoto University School of Medicine.
Provenance and peer review Not commissioned; externally peer reviewed.
If you wish to reuse any or all of this article please use the link below which will take you to the Copyright Clearance Center’s RightsLink service. You will be able to get a quick price and instant permission to reuse the content in many different ways.