Background & Aims

A variety of fecal immunochemical tests (FITs) for hemoglobin (Hb) are used in colorectal cancer screening. It is unclear to what extent differences in reported sensitivities and specificities reflect true heterogeneity in test performance or differences in study populations or varying pre-analytical conditions. We directly compared the sensitivity and specificity values with which 9 quantitative (laboratory-based and point-of-care) FITs detected advanced neoplasms (AN) in a single colorectal cancer screening study.

Methods

Pre-colonoscopy stool samples were obtained from participants of screening colonoscopy in Germany from 2005 through 2010 and frozen at –80°C until analysis. The stool samples were thawed, homogenized, and used for 9 different quantitative FITs in parallel. Colonoscopy and histology reports were collected from all participants and evaluated by 2 independent, trained research assistants who were blinded to the test results. Comparative evaluations of diagnostic performance for AN were made at preset manufacturers′ thresholds (range, 2.0–17.0 μg Hb/g feces), at a uniform threshold (15 μg Hb/g feces), and at adjusted thresholds yielding defined levels of specificity (99%, 97%, and 93%).

Results

Of the 1667 participants who fulfilled the inclusion criteria, all cases with AN (n = 216) and 300 randomly selected individuals without AN were included in the analysis. Sensitivities and specificities for AN varied widely when we used the preset thresholds (21.8%–46.3% and 85.7%–97.7%, respectively) or the uniform threshold (16.2%–34.3% and 94.0%–98.0%, respectively). Adjusting thresholds to yield a specificity of 99%, 97%, or 93% resulted in almost equal sensitivities for detection of AN (14.4%–18.5%, 21.3%–23.6%, and 30.1%–35.2%, respectively) and almost equal positivity rates (2.8%–3.4%, 5.8%–6.1%, and 10.1%–10.9%, respectively).

Conclusions

Apparent heterogeneity in diagnostic performance of quantitative FITs can be overcome to a large extent by adjusting thresholds to yield defined levels of specificity or positivity rates. Rather than simply using thresholds recommended by the manufacturer, screening programs should choose thresholds based on intended levels of specificity and manageable positivity rates.

中文翻译：

---

直接比较9种定量粪便免疫化学检测对大肠癌筛查的诊断性能

背景与目标

各种针对血红蛋白（Hb）的粪便免疫化学测试（FIT）用于结直肠癌筛查。目前尚不清楚所报道的敏感性和特异性的差异在多大程度上反映了测试性能的真正异质性，研究人群的差异或分析前条件的变化。我们直接比较了9个定量（基于实验室和现场护理）FIT在单个结肠直肠癌筛查研究中检测到的晚期肿瘤（AN）的敏感性和特异性值。

方法

结肠镜检查前的粪便样本是从2005年至2010年从德国的结肠镜检查参与者中获得的，并在–80°C下冷冻直至进行分析。将粪便样品解冻，匀浆，并平行用于9种不同的定量FIT。从所有参与者中收集结肠镜检查和组织学报告，并由对测试结果不知情的2位受过训练的独立研究助理进行评估。在预设的制造商阈值（范围2.0–17.0μgHb / g粪便），统一阈值（15μgHb / g粪便）和调整后的阈值（产生定义的特异性水平）下，对AN的诊断性能进行了比较评估。 99％，97％和93％）。

结果

在满足纳入标准的1667名参与者中，所有患有AN的病例（n = 216）和300名随机选择的没有AN的个体都纳入了分析。当使用预设阈值（分别为21.8％–46.3％和85.7％–97.7％）或统一阈值（分别为16.2％–34.3％和94.0％–98.0％）时，AN的敏感性和特异性差异很大。调整阈值以产生99％，97％或93％的特异性导致对AN的检测灵敏度几乎相等（分别为14.4％–18.5％，21.3％–23.6％和30.1％–35.2％）并且几乎相等阳性率（分别为2.8％–3.4％，5.8％–6.1％和10.1％–10.9％）。

结论

通过调整阈值以产生确定水平的特异性或阳性率，可以在很大程度上克服定量FIT诊断性能中的表观异质性。筛查程序不仅应使用制造商建议的阈值，还应根据预期的特异性水平和可控制的阳性率选择阈值。