Rethinking PICO in the Machine Learning Era: ML-PICO

Background  Machine learning (ML) has captured the attention of many clinicians who may not have formal training in this area but are otherwise increasingly exposed to ML literature that may be relevant to their clinical specialties. ML papers that follow an outcomes-based research format can be assessed using clinical research appraisal frameworks such as PICO (Population, Intervention, Comparison, Outcome). However, the PICO frameworks strain when applied to ML papers that create new ML models, which are akin to diagnostic tests. There is a need for a new framework to help assess such papers. Objective  We propose a new framework to help clinicians systematically read and evaluate medical ML papers whose aim is to create a new ML model: ML-PICO (Machine Learning, Population, Identification, Crosscheck, Outcomes). We describe how the ML-PICO framework can be applied toward appraising literature describing ML models for health care. Conclusion  The relevance of ML to practitioners of clinical medicine is steadily increasing with a growing body of literature. Therefore, it is increasingly important for clinicians to be familiar with how to assess and best utilize these tools. In this paper we have described a practical framework on how to read ML papers that create a new ML model (or diagnostic test): ML-PICO. We hope that this can be used by clinicians to better evaluate the quality and utility of ML papers.     

Oral Valganciclovir Is Noninferior to Intravenous Ganciclovir for the Treatment of Cytomegalovirus Disease in Solid Organ Transplant Recipients

Intravenous ganciclovir is the standard treatment for cytomegalovirus disease in solid organ transplant recipients. Oral valganciclovir is a more convenient alternative. In a randomized, international trial, recipients with cytomegalovirus disease were treated with either 900 mg oral valganciclovir or 5 mg/kg i.v. ganciclovir twice daily for 21 days, followed by 900 mg daily valganciclovir for 28 days. A total of 321 patients were evaluated (valganciclovir [n = 164]; i.v. ganciclovir [n = 157]). The success rate of viremia eradication at Day 21 was 45.1% for valganciclovir and 48.4% for ganciclovir (95% CI -14.0% to +8.0%), and at Day 49; 67.1% and 70.1%, respectively (p = NS). Treatment success, as assessed by investigators, was 77.4% versus 80.3% at Day 21 and 85.4% versus 84.1% at Day 49 (p = NS). Baseline viral loads were not different between groups and decreased exponentially with similar half-lives and median time to eradication (21 vs. 19 days, p = 0.076). Side-effects and discontinuations of assigned treatment (18 of 321 patients) were comparable. Oral valganciclovir shows comparable safety and is not inferior to i.v. ganciclovir for treatment of cytomegalovirus disease in organ transplant recipients and provides a simpler treatment strategy, but care should be taken in extrapolating to organ transplant recipients not properly represented in the present study.     

A Prospective Randomized Trial of Steroid-Free Maintenance Regimens in Kidney Transplant Recipients—An Interim Analysis

We compared three maintenance immunosuppressive regimens in a rapid discontinuation of prednisone protocol. From March 1, 2001, through December 31, 2003, 239 first and second kidney transplant recipients (166 LD; 73 DD) were randomized. All recipients were treated with Thymoglobulin; all received steroids intraoperatively and for 5 days postoperatively. Randomization was to cyclosporine-mycophenolate mofetil (n = 85); high-level tacrolimus (TAC) (8-12 ng/mL)-low-level sirolimus (SRL) (3-7 ng/mL) (n = 72); or low-level TAC (3-7 ng/mL)-high-level SRL (8-12 ng/mL) (n = 82). We found no difference at 24 months between groups in patient, graft, death-censored graft, or acute rejection-free graft survival, or in kidney function. Wound complications were more common in SRL-treated recipients (p = 0.02); we found no other differences between groups in complication rates. Our data suggest that excellent patient and graft survival and low rejection rates can be obtained using a variety of maintenance protocols without prednisone.     

A novel model to study bacterial adherence to the transplanted airway: Inhibition of Burkholderia cepacia adherence to human airway by dextran and xylitol

BACKGROUND: Lung infection with Burkholderia cepacia complex before lung transplantation in patients with cystic fibrosis is a major risk factor for decreased post-operative survival rates compared with those of patients colonized with the more common opportunistic pathogen Pseudomonas aeruginosa. Because adherence to mucosal surfaces is an important initial step in infection, we investigated the use of non-toxic neutral polysaccharides and a sugar alcohol to prevent adherence of B cepacia complex to allograft airway epithelium. METHODS: We used human airway explants prepared from donor tracheobronchial tissue to test the effect of dextrans and xylitol in inhibiting the binding of Burkholderia cepacia complex. We used immunofluorescence and electron microscopy to determine the distribution of bacteria in the explants. RESULTS: Burkholderia cepacia complex bound to the explants and was found only in the surface mucus layer. Dextran 40 kd applied before adding the bacteria decreased the number of bound organisms by 80% to 99%. Smaller molecular mass dextrans (4 and 20 kd) were ineffective. Xylitol inhibited bacterial binding by 67% to 85%. Both agents seemed to decrease the thickness of the surface mucus, suggesting that they may indirectly inhibit bacterial binding by removing adherent surface mucus. CONCLUSIONS: Treating donor lungs with dextran 40 kd or xylitol before (and possibly after) surgery may inhibit the adherence of Burkholderia cepacia complex to airways and may prevent or decrease subsequent infection of the allografts.     

Evaluation of specialized laparoscopic suturing and tying devices.

BACKGROUND: Laparoscopic suturing and tying constitute advanced minimally invasive surgery skills. Developing proficiency in the standard methods with needle drivers is often an arduous process. Recent advances in laparoscopic instrumentations has allowed for easier methods of suturing and tying. This study investigated the hypothesis that the use of a specialized suturing device and a specialized tying device allows inexperienced medical students to suture and tie laparoscopically. METHODS: Preclinical medical students who had not received any training in open or laparoscopic surgery were included in this investigation. Each student was given a 5-minute demonstration of a specialized suturing device and a specialized tying device. The medical students were not allowed to deploy either device before actual use. After the demonstration, each student was given the device to use in a porcine model. Times were recorded and a subjective grade was given for each student. RESULTS: Twenty medical students were involved in this study. All medical students were able to complete the task of suturing and tying. The average time to suture was 104.6 seconds and the average time to tying was 31.2 seconds. The average subjective performance grade was 90 (out of 100). CONCLUSION: Specialized devices are easy to learn and use for laparoscopic suturing and tying with minimal instruction even for inexperienced medical students. Even surgeons who are not well versed in laparoscopic surgery should be able to suture and tie with certain laparoscopic instruments.