Revisiting the fragility of influence functions.

In the last few years, many works have tried to explain the predictions of deep learning models. Few methods, however, have been proposed to verify the accuracy or faithfulness of these explanations. Recently, influence functions, which is a method that approximates the effect that leave-one-out training has on the loss function, has been shown to be fragile. The proposed reason for their fragility remains unclear. Although previous work suggests the use of regularization to increase robustness, this does not hold in all cases. In this work, we seek to investigate the experiments performed in the prior work in an effort to understand the underlying mechanisms of influence function fragility. First, we verify influence functions using procedures from the literature under conditions where the convexity assumptions of influence functions are met. Then, we relax these assumptions and study the effects of non-convexity by using deeper models and more complex datasets. Here, we analyze the key metrics and procedures that are used to validate influence functions. Our results indicate that the validation procedures may cause the observed fragility.

The complex etiology of multiple sclerosis.

Multiple sclerosis is a demyelinating disease which is presumed to be a consequence of infiltrating lymphocytes autoreactive to myelin proteins. This is substantiated by several lines of clinical evidence and supported by correlative studies in preclinical models. The development of new therapeutics for MS has been guided by this perspective; however, the pathogenesis of MS has proven to be quite complex as observations exist which question the role of autoreactive lymphocytes in the etiology of MS. In addition the current immunomodulatory therapeutics do not prevent most patients from progressing into more serious forms of the disease. The development of truly transformational therapeutics for MS will likely require a broad assault that expands beyond the concept of MS being an autoimmune disease.

Effects of orthopaedic wear particles on osteoprogenitor cells.

Wear particles from total joint arthroplasties are constantly being generated throughout the lifetime of an implant. Since mesenchymal stem cells and osteoprogenitors from the bone marrow are the precursors of osteoblasts, the reaction of these cells to orthopaedic wear particles is critical to both initial osseointegration of implants and ongoing regeneration of the periprosthetic bed. Particles less than 5 microm can undergo phagocytosis by mature osteoblasts, with potential adverse effects on cellular viability, proliferation and function. The specific effects are dependent on particle composition and dose. Metal and polymer particles in non-toxic doses stimulate pro-inflammatory factor release more than ceramic particles of a similar size. The released factors inhibit markers of bone formation and are capable of stimulating osteoclast-mediated bone resorption. Mesenchymal stem cells and osteoprogenitors are also profoundly affected by wear particles. Titanium and polymethylmethacrylate particles inhibit bone cell viability and proliferation, and downregulate markers of bone formation in a dose- and time-dependent manner. Future studies should delineate the molecular mechanisms by which particles adversely affect mesenchymal stems cells and the bone cell lineage and provide strategies to modulate these effects.

Staking out novelty on the genomic frontier.

Recent advances in genomics include global assessment and classification of genome content, high-throughput biological pathway construction, systematic identification of previously unpredicted genes and the in vitro creation of novel motifs with biological function not found in nature (extra-genomic gene discovery). The ability to make global surveys of transcriptomes has given rise to fields such as pharmacogenomics and toxicogenomics. These applications of genomics technologies, with conventional drug assessment methodologies, will lead to more tolerable drugs and a better understanding of clinical populations. Integration of pathway mapping, using proteomics married to expression, will also significantly affect how new therapeutics are discovered as cross-biological cross-pathway interactions lead to novel drug targets and better predictions of drug tolerance.

Evaluating more naturalistic outcome measures: A 1-year smartphone study in multiple sclerosis.

OBJECTIVE: In this cohort of individuals with and without multiple sclerosis (MS), we illustrate some of the novel approaches that smartphones provide to monitor patients with chronic neurologic disorders in their natural setting. METHODS: Thirty-eight participant pairs (MS and cohabitant) aged 18-55 years participated in the study. Each participant received an Android HTC Sensation 4G smartphone containing a custom application suite of 19 tests capturing participant performance and patient-reported outcomes (PROs). Over 1 year, participants were prompted daily to complete one assigned test. RESULTS: A total of 22 patients with MS and 17 cohabitants completed the entire study. Among patients with MS, low scores on PROs relating to mental and visual function were associated with dropout (p < 0.05). We illustrate several novel features of a smartphone platform. First, fluctuations in MS outcomes (e.g., fatigue) were assessed against an individual's ambient environment by linking responses to meteorological data. Second, both response accuracy and speed for the Ishihara color vision test were captured, highlighting the benefits of both active and passive data collection. Third, a new trait, a person-specific learning curve in neuropsychological testing, was identified using spline analysis. Finally, averaging repeated measures over the study yielded the most robust correlation matrix of the different outcome measures. CONCLUSIONS: We report the feasibility of, and barriers to, deploying a smartphone platform to gather useful passive and active performance data at high frequency in an unstructured manner in the field. A smartphone platform may therefore enable large-scale naturalistic studies of patients with MS or other neurologic diseases.

Do presence and location of annular tear influence clinical outcome after lumbar total disc arthroplasty? A prospective 1-year follow-up study.

BACKGROUND: Lumbar total disc arthroplasty is often performed in patients with axial back pain. There are multiple etiologies for axial back pain, including disc degeneration and annular tears. The location of these annular tears can vary, producing differing preoperative symptomatology. Intraoperatively, disruptions in the annulus are identifiable, and it has been suggested that patients with discrete annular tears may have better clinical outcomes after surgery. The purpose of this study was to investigate whether the presence and location of annular tears have an effect on clinical outcomes after lumbar total disc arthroplasty. METHODS: Patients undergoing a single-level anterior disc replacement from L3-S1 at a single site by a single surgeon were evaluated preoperatively for the presence or absence of annular tears with magnetic resonance imaging. All patients were part of either the ProDisc (n = 41) (Synthes, Paoli, Pennsylvania) or Activ-L (n = 19) (Aesculap [B. Braun Melsungen AG], Tuttlingen, Germany) lumbar prospective clinical trials. In those patients with annular tears, the location of the tear (central, paracentral, or lateral) was documented. Patients were assessed at 6 and 12 months after lumbar total disc arthroplasty with the Oswestry Disability Index (ODI), visual analog scale (VAS) score for back pain, VAS score for leg pain, and radiographic imaging. All radiographic evaluations were conducted by an attending neuroradiologist and an attending spinal surgeon, and reliability testing was performed. An analysis of variance was performed among the 3 anatomic locations of annular tears. RESULTS: A total of 60 patients were included and had complete 12-month follow-up. The prevalence of annular tears among all patients was 42% (n = 25). Outcome data in patients without annular tears were as follows: ODI, 66% preoperatively and 26% postoperatively; VAS score for back pain, 8.0 preoperatively and 2.6 postoperatively; and VAS score for leg pain, 2.9 preoperatively and 1.2 postoperatively. Among those patients with tears, the prevalence of central tears was 80%, the prevalence of paracentral tears was 12%, and the prevalence of lateral tears was 8%. Outcome data in patients with central tears were as follows: ODI, 66% preoperatively and 26% postoperatively; VAS score for back pain, 7.8 preoperatively and 2.6 postoperatively; and VAS score for leg pain, 5.2 preoperatively and 0.5 postoperatively. Outcome data in patients with paracentral tears were as follows: ODI, 86% preoperatively and 59% postoperatively; VAS score for back pain, 8.8 preoperatively and 3.3 postoperatively; and VAS score for leg pain, 5.0 preoperatively and 5.4 postoperatively. Outcome data in patients with lateral tears were as follows: ODI, 6.5 preoperatively and 2.6 postoperatively; VAS score for back pain, 9.2 preoperatively and 0.2 postoperatively; and VAS score for leg pain, 1.4 preoperatively and 0.7 postoperatively. In those patients with paracentral tears, there was a significantly higher incidence of postoperative radicular symptoms both from an intensity standpoint and from a duration standpoint. Other complications did not vary among those patients with or without annular tears. CONCLUSIONS: Although patients with annular tears and patients without annular tears improve after lumbar artificial disc replacement, those with central annular tears or without tears have significantly lower disability scores than those with paracentral tears or lateral tears, whose outcome scores showed significantly less improvement (P ≤ .03). In particular, patients with central tears have less postoperative leg pain than those with paracentral annular tears. In this study the presence or absence of an annular tear on magnetic resonance imaging was not a significant predictive factor for clinical outcome. Further investigation regarding the effects of paracentral annular tears and surgical techniques should be explored.

Total hip arthroplasty using the miniature Anatomic Medullary Locking stem.

We report the outcome of a prospective consecutive series of 52 primary total hip arthroplasties using the miniature porous-coated Anatomic Medullary Locking stem in patients with small anatomic proportions because of hip dysplasia or juvenile chronic arthritis. The mean age of the patients at the time of surgery was 28.7 years (range 14-56 years). The average body weight and height of the patients were 51.8 kg (range 38.5-78.3 kg) and 157.1 cm (range 142.2-183 cm), respectively. The stem was cementless in 40 hips and cemented in 12 hips because of poor bone stock. A cementless acetabular cup with screw was used in all hips. The average followup was 7.1 years (range, 3-15.6 years). The Harris hip scores improved from an average of 31.2 points (range, 3.1-68.8 points)preoperatively to 82.8 points (range, 61.1-96.6 points) at latest followup. Three of 12 (25%) cemented and two of 40 (5%) cementless stem were revised. Four of seven 42-44-mm cups were revised. The miniature Anatomic Medullary Locking cementless femoral stem provides a satisfactory outcome in patients with small anatomic proportions. However, wear and osteolysis with the use of a small cementless polyethylene liner remain challenges.