Inclusion and Exclusion Criteria
All studies included in this review were RCTs.Data Synthesis and Statistical Analysis
In this study, the mean value of the AS group minus the mean 
value of the AT group (mean difference, MD) with the 95% confidence interval (CI) was used to estimate continuous outcomes.Assessment on Risk of Bias and Article Quality
We used the risk of bias tool recommended by the Cochrane Collaboration [16] to access the risk of bias on the following six aspects: 
random sequence generation, allocation concealment, blinding of 
participants and personnel, blinding of outcome assessors, incom
-
plete outcome data, and selective outcome reporting.Databases and Search Strategy
The following databases were analyzed: PubMed, Science Direct, the Cochrane Library, the Chinese National Knowledge Infrastructure, and the Wanfang Database, with one or a combination of the following terms: autologous serum, artificial tears, com
-
parison, random, and randomized.The evaluation indicators must contain at least one of Ocular Surface Disease Index 
(OSDI), Schirmer I test, tear break-up time (TBUT), fluorescein 
staining, and rose bengal staining.Outcome Measurement
Outcome evaluation included the analysis about subjective 
symptoms, Schirmer I test, TBUT, fluorescein staining, and rose 
bengal staining.In this meta-analysis, all extracted data, including OSDI, 
TBUT, Schirmer I test, fluorescein staining score, and rose bengal 
staining score, were continuous data and we extracted the mean 
value and standard deviation (SD).Specifically, we 
carefully reviewed the studies included and checked whether there 
was a description of how the random sequence was generated and allocated, whether and how the blinding method was applied, how the 
outcomes were measured, and whether there were missing data.We 
defined resolution of DED as the increase of Schirmer I, extension 
of TBUT, the decrease of OSDI, fluorescein staining score, and 
rose bengal score.For 
example, if the random sequence was generated using a random number table or pseudo-random numbers using statistical software, the 
risk of bias was assessed as "low" level; otherwise, if there was evidence 
showing that the random sequence was generated in an inappropriate 
way, the risk of bias on random sequence generation was assessed as 
a "high" level. If no information could be obtained or it was difficult 
to make a judgment, the risk of bias was assessed as "unclear."Studies would be excluded if the study design was not RCT or 
subjects or intervention measurement did not meet the inclusion 
criteria.Besides, if the original quantitative data of effectiveness of 
AS or AT was not extractable or the appraisal system could not be 
combined with other studies, the study would be excluded.Subjective symptoms were measured by OSDI, 
which is the most extensively used questionnaire for DED.Schirmer I test was performed using a strip placed in the lower conjunctival sac for 5 min without anesthesia.Briefly, the cornea was 
horizontal divided into three equal compartments, and each zone 
was graded from 0 to 3.Details of the search strategy 
are available in online supplementary Appendix 1 (see www.karger.com/doi/10.1159/000505630).The subjects 
were DED patients, regardless of the participants' age or sex or etiology of disease.For fluorescein staining, we analyzed 
studies which used same graded system.For rose bengal staining, we analyzed stud
-
ies using rose bengal 1% and a scale of 9 points.The publication date was from 
the creation of the databases to December 31, 2018.Data Extraction
We extracted data of outcomes measured at the last follow-up 
time.All the analyses were conducted with the opensource R program (version 3.4.4).The intervention measurement should be AS 
therapy, and the control group should be given AT treatment, with 
or without combination of other therapy.TBUT was the average of 
two or three measurements.Before estimating the p