PICO Portal is proud to be attending ISPOR 2025 in Montreal from May 13 -16 and featured in two poster presentations showcasing how AI, when designed responsibly, can meaningfully accelerate and enhance evidence synthesis.
1) POSTER: Contextual Prioritization of Titles and Abstracts
Using to reprioritize screening tasks daily, this research demonstrates how AI can streamline citation screening and enable truly living systematic reviews—without compromising recall or review rigor.
May 14, 4:00–7:00 PM
Poster MSR39
Add to Google Calendar | iCal
2) POSTER: Data Extraction from Full-Text PDFs Using LLMs
This study explores a hybrid approach where large language models (LLMs) assist human methodologists in extracting structured data directly from full-text PDFs. Learn about the serious gains teams saw in productivity without any compromise to the protocol or data quality.
May 15, 10:30 AM–1:30 PM
Poster MSR74
Add to Google Calendar | iCal
POSTER-1: Contextual Prioritization of Titles and Abstracts
SR1 Prompt:
We are conducting a systematic review and network meta-analysis to evaluate the effectiveness of different antibiotic types and regimens on treating bacterial keratitis (BK). For this review, we have the following eligibility criteria:
Population: We will include studies of all types of participants who were diagnosed (clinically and/or microbiologically) and treated for BK, with no restriction to the age, gender, and ethnicity. Participants must have BK and cannot have any other forms of keratitis. We will exclude studies that include participants who are treated for other types of infectious keratitis such as fungal, viral or parasitic infection – either separately or in conjunction with BK.
Intervention: We will only include studies that examined and compared different types of topical antibiotics. We will include both active and inactive (e.g., placebo) controlled trials to increase the amount of evidence that we are likely to find and strengthen the inferences (both direct and indirect) that will be made through our NMA. Specific interventions for which we anticipate finding evidence include, but are not limited to:
- Dual fortified antibiotic therapy (e.g., combined cephalosporin-aminoglycoside or combined cephalosporin-fluoroquinolone )
- Fluoroquinolone monotherapy (which comprises 2nd, 3rd or 4th generation fluoroquinolone)
- Glycopeptides (e.g., Vancomycin)
- Polypeptides (e.g., Polymyxin)
- Cyclic peptides (e.g., Gramicidin)
- Antiseptic agents (e.g., Povidone-iodine)
- Placebo (i.e., control)
Comparator: As we are planning a network meta-analysis, we do not have any restrictions on the types of controls. The only consideration is that the control should also be an eye drop to match the same kind of treatment as the topically delivered antibiotics that are of interest.
Study design: We will include only randomized controlled trials that are parallel in design. We will include study protocols and registrations of randomized controlled trials that have not yet completed if they are relevant to our population and intervention. We will exclude all crossover trials, and all observational studies, including: pseudo or non-randomized controlled trials, cohort studies (retrospective or prospective), case-control studies, and before-and-after studies. We will exclude reviews, case reports, letters to the editor, and editorials.
The file that I will upload is a .csv containing 12103 records that were retrieved from multiple databases and de-duplicated. The first column is a record ID, the second column is the title, and the third column is the abstract (the first row is the names of these columns). I want you to look through each record’s title and abstract and make a judgement for whether it is likely to be relevant or not to our review question. Apply the criteria that I specified above and return a new .csv that contains the same three columns, but only the records that are relevant for our systematic review. If a record is missing the title or abstract, ignore the missing information and base your decision on the present information. Do you understand the task?
SR2 Prompt:
We are conducting a systematic review to evaluate the potential ophthalmic and periorbital harms of dermatologic and ophthalmic topically administered vitamin A based products on the health of the ocular surface, particularly the eyelids and the meibomian glands. For this review, we have the following eligibility criteria:
Population: Studies will include any participants (no restriction on age, sex, country, or race/ethnicity) who have applied, prescription or non-prescription vitamin A derivatives in drop, gel or ointment form topically to the eye (e.g., through eye drops or ointment), eyelids, or to the periocular area (e.g., on the skin surrounding the eye) for any frequency or duration. We will exclude studies in which participants concurrently use another eye drop known to induce dry eye signs or symptoms (e.g., glaucoma drops). We will also exclude studies in which there may be a confounding factor such as concomitant use of another treatment for dry eye or ocular surface disease. Animal studies and other lab-based pre-clinical studies also will be excluded.
Intervention: We will include prescription or non-prescription eye drops, lotions, ointments, or creams containing any amount of vitamin A or vitamin A derivatives as an active ingredient. Such products could include tretinoin cream, vitamin A creams, retinol moisturizers, pro-retinol creams, retin-A creams, vitamin-A containing anti-wrinkle creams, etc. Please ensure that the intervention is applied ocularly or periorbitally (i.e., to the eyes or surrounding area)
Comparator: Comparators will include placebo, no treatment, active comparators (non-vitamin A containing topical treatments). Comparators will be in any form—drop, lotion, ointment or cream.
Study design: We will include observational studies, randomized and non-randomized controlled trials, cohort studies (retrospective or prospective), case-control studies, and before-and-after studies. For interventional studies, we will include any phase of study (Phase 1, Phase 2, Phase 3, or Phase 4). We will accept any primary study design that compares participants who have been exposed with ones who are not exposed. We will include study protocols and registrations of studies that have not yet completed if they are relevant to our population and intervention. We will exclude reviews, case reports, letters to the editor, and editorials.
The file that I will upload is a .csv containing 8900 records that were retrieved from multiple databases and de-duplicated. The first column is a record ID, the second column is the title, and the third column is the abstract. I want you to look through each record’s title and abstract and make a judgement for whether it is likely to be relevant or not to our review question. Apply the criteria that I specified above and return a new .csv that contains the same three columns, but only the records that are relevant for our systematic review. If a record is missing the title or abstract, ignore the missing information and base your decision on the present information. Do you understand the task?
Inclusion Prompt:
Our objective is to evaluate the potential ophthalmic and periorbital harms of dermatologic and ophthalmic topically administered vitamin A based products on the health of the ocular surface, particularly the eyelids and the meibomian glands. A record should be included if the title and/or abstract describes the following characteristics: the intervention/exposure must be an eye drop, lotion, ointment, or cream with any vitamin A or vitamin A derivative as the active ingredient (e.g., tretinoin cream, vitamin A creams, retinol moisturizers, pro-retinol creams, retin-A creams, vitamin-A containing anti-wrinkle creams); the intervention/exposure must be applied to the eye or surrounding area (i.e., ocularly or periorbitally); the study must include a comparison group but can be any of the following designs: randomized controlled trial, non-randomized trial, cohort study, case-control study, or before-and-after study.
Eligibility prompts for PICO Portal
Exclusion Prompt:
Our objective is to evaluate the potential ophthalmic and periorbital harms of dermatologic and ophthalmic topically administered vitamin A based products on the health of the ocular surface, particularly the eyelids and the meibomian glands. A record should be excluded if the title and/or abstract describes any of the following characteristics: any kind of eye drop, lotion, ointment, or cream that is applied to the hands, non-eye body part, or ingested (e.g., vitamin supplementation); if the study design is a review article, case report, case series, editorial/commentary, lab-based, or animal study.
INC_CRITERIA = “the intervention or exposure must be an eye drop, lotion, ointment, or cream with any vitamin A or vitamin A derivative as the active ingredient (e.g., tretinoin cream, vitamin A creams, retinol moisturizers, pro-retinol creams, retin-A creams, vitamin-A containing anti-wrinkle creams); the intervention/exposure must be applied to the eye or surrounding area (i.e., ocularly or periorbitally); the study must include a comparison group but can be any of the following designs: randomized controlled trial, non-randomized trial, cohort study, case-control study, or before-and-after study.”
EXC_CRITERIA = “lotion, ointment, or cream that is applied to the hands, or ingested (e.g., vitamin supplementation); if the study design is a review article, case report, case series, editorial or commentary, lab-based, or animal study.”
POSTER-2: Data Extraction from Full-Text PDFs Using LLMs