Focal colorectal uptake in ¹⁸FDG-PET/CT: maximum standard uptake value as a trigger in a semi-automated screening setting

We found no correlation between the SUV_max and a priori known extrinsic factors that could bias the SUV_max measurements (Additional file 1: Table S1). This suggests that the SUV_max does not need to be further normalised for these a priori known extrinsic factors. The lack of correlation between the SUV_max and the time to scan suggests that the SUV_max is independent of the time of scanning in the tested range 59–112 min (mean: 75 ± 14 min) after injection. Further study is needed using dynamic PET to test intra-individually at 30, 60, 90 and 120 min to ascertain which time interval between injection and imaging is optimal for the detection of malignant colorectal uptakes. It has recently been found that normalisation to the blood pool in the aorta provides a means of correcting for scan time dependence [9]. This requires further verification. The lack of correlation between the SUV_max and activity in the tested range 94–395 MBq (mean: 329 ± 46 MBq) suggests that the activity can be reduced, possibly independent of weight, to 200 MBq. This is of interest in situations where PET/CT may be offered to asymptomatic individuals instead of screening colonoscopy where endoscopy is contraindicated, refused or not possible to complete.

To be justified, the benefit of PET/CT in early detection must compensate for its radiation risk. With 200 MBq activity the total radiation exposure can be reduced to less than two times natural background radiation (200 MBq × 6.7 mSv/350 MBq [10] + 0.8 mSv [11] = 4.6 < 2 × 2.4 = 4.8 mSv, assuming a linear relationship between MBq and mSv if 350 MBq results in 6.7 mSv [10]). The dose issue is especially significant if PET/CT is used for screening before the onset of symptoms in healthy subjects. The purely hypothetical [12] and delayed radiation risk is compensated if PET/CT detects 3 % of the colorectal cancers which occur in 0.9 % of cases (colonoscopic prevalence in a screening setting) [13] (x % sensitivity × 0.9 % prevalence >0.005 %/mSv [14] radiation risk × 4.8 mSv radiation exposure; with x % >3 %). In this calculation of the minimum required sensitivity (x % > 3 %), the concurrent detection of extra-colonic cancer entities and advanced adenomas, as well as other serious conditions such as cardiovascular disease, were not taken into account thus underestimating the overall benefit of PET/CT. Furthermore, there is a 10- to 40-year delay [15] between the hypothetical induction and development of radiation-induced cancer; the natural history of a missed cancer, had PET/CT not been performed, is more severe than the natural history of a hypothetical and delayed induced cancer, had PET/CT been performed. Additionally, the benefit-to-risk ratio of PET/CT increases with age due to the decreasing radiation risk and increasing incidence of cancer. This is in contrast to colonoscopy, where the rate of complication increases with age, while the prophylactic meaning of a polypectomy decreases. This is especially relevant as colorectal screening is recommended for individuals up to 70 years of age [16].

In our study, we found the minimum SUV_max in 54 colorectal carcinomas to be 5 thus determining the threshold. There are only a few studies which showed an SUV_max lower than 5 for colorectal cancer. Sarikaya et al. [17] reported four carcinomas that were detected with SUV_max <4.5, of which three were mucinous. Peng et al. [18] reported a range in SUV_max from 3.1 to 28 which included two mucinous carcinomas. The low cellularity of mucinous carcinomas may explain the low SUV_max. A meta-analysis regarding the SUV_max of colorectal cancer is not possible because the SUV_max was not always listed for each carcinoma. When SUV_max is used as the sole trigger, and not in combination with other factors, an SUV_max threshold of 5 would cover 96 % (215/224) of FDG-positive colorectal cancers (Figs. 1, 2, 3, 4) [17‐31], leaving only 4 % of positive cases requiring individual interpretation.

Besides the 4 % of PET-positive colorectal cancer cases that fall below the threshold, some cancer cases are completely PET negative. The rate of PET-negative cancer cases may be at least 5 % as suggested by studies looking at all patients who underwent PET/CT followed by colonoscopy within a short period of time [25, 30, 32]. On the other hand, the miss-rate of optical colonoscopy can be estimated at a worst case of 2.9 %, assuming that all 2.9 % of the so-called interval cancers occurring within 5 years of a negative colonoscopy were missed and not newly developed [33].

The issue of false negatives must be viewed in context. The vast majority of the German population currently does not come forward for screening due in large part to the invasive nature of colonoscopy. Between 2002 and 2008, 2,821,392 screening colonoscopies were performed across Germany which represents 15.5 and 17.2 % of all eligible men and women, respectively, from the age group 55–74 years [13]. Thus, approximately 80 % of the target group did not take advantage of the colonoscopy screening programme during this 6-year screening interval. Although the acceptance rate is higher in some other countries, such as the US, there is a widespread reluctance on the part of the population to come forward for colonoscopy-based screening. Shortcomings in alternative screening techniques must be balanced against the significant number of tumours which progress to a more advanced stage due to this very low acceptance of colonoscopy screening. PET/CT should not replace colonoscopy screening in the minority of individuals who assent, but provide an attractive alternative for the majority who refuse. Thus, if colonoscopy is refused, PET/CT needs to be compared with faecal occult blood test (FOBT) and not with colonoscopy.

FDG-enriched stool in the caecum is the most common cause of false-positive FDG accumulation [11]. FDG excretion into the small bowel and accumulation in the caecum during the 60-min interval between injection and imaging may explain this observation. The typical location in the caecum in conjunction with centric distribution and air-typical CT values, which indicate air inclusions, helps to differentiate FDG-enriched stool from a wall-adherent eccentric mass. Although Van Heoij et al. [31] recently found that the SUV_max in 404 focal colorectal uptakes was significantly higher for cancer (p < 0.001) than for all other types of lesions (advanced adenoma, non-advanced adenoma and benign lesions), Keyzer et al. [34] showed that the SUV_max alone does not differentiate true- from false-positive colorectal FDG foci. The metabolic volume also failed to differentiate TP from FP [34, 35].

As the SUV_max in premalignant/malignant and physiological/benign colorectal FDG accumulation is indistinct, the clear separation (cut-off) between TP and FP seems to be unattainable with SUV_max alone. We therefore defined the trigger as automating the decision above a threshold only (semi-automated analysis). Given the relatively low prevalence of focal colorectal uptakes (3.6 %) but the relatively high risk of these being malignant or premalignant (68 %) [36], the benefit of maximising the sensitivity with semi-automated analysis seems to justify a lower specificity with more FPs. If colonoscopy is the worst consequence of an FP, these patients would not be disadvantaged compared to their outcome had they taken up the current screening programme [16, 37]. In comparison to colonoscopy, the 1.5 % rate of FP in PET/CT [36] with consecutive colonoscopy is far lower than the 26.5 % rate of false-positive polypectomies, several polypectomies per person not counted [13].

Averaging within a volume pixel (voxel) of a finite edge length is a drawback of digitisation. Smaller lesions in the range of only view voxels might not be visible due to spatial and temporal averaging within one voxel (partial volume artefacts) [38]. The resultant blurring might reduce the overall contrast so that the lesion is not delineated. However, a very high uptake—the so-called hot spot phenomenon, as known from melanoma—might compensate for a larger voxel size and even depict lesions within the range of the voxel resolution [currently: 95 mm³ (=0.095 ml) based on 400 × 400 matrix reconstruction]. However, this potential inferiority in voxel resolution compared to optical colonoscopy might be compensated by a shorter screening interval (e.g. 5 years as for CT colonography [37]). This may be completely unnecessary when the long lead time of 10 years in the adenoma-to-carcinoma sequence is taken into account [39‐42] and the fact that therapy in asymptomatic (lower stage) colorectal cancer is mostly curative. Furthermore, it must be emphasised that lower voxel resolution can easily be compensated for by a shorter screening interval, in contrast to a lower screening acceptance rate which cannot be compensated for.

There is some evidence that PET/CT failed to detect around half of cases with advanced adenoma [43]. The study was performed between 2000 and 2009 using now outdated PET and PET/CT technology. Since then, the spatial resolution has improved from 4.5 mm to almost 2 mm today, for example. In an interval screening programme, it is the accuracy of the programme and not of the single test which matters. Furthermore, the consequence of a missed tiny adenoma is unclear if the cancer can still be curatively resected at the consecutive screening, if indeed the adenoma develops to cancer at all. The mismatch in prevalence between advanced adenoma (6.4 %) and colorectal cancer (0.9 %) [13] suggests that not all advanced adenomas proceed to cancer. It is assumed that a patient with advanced adenoma at age 55–65 has a greater than 50 % chance of developing colon cancer [44]. The potential lack in screening sensitivity may be compensated by reducing the interval between examinations (for example from 10 to 5 years as proposed for CT colonography [37]), but a low screening acceptance rate cannot be compensated.

To date, we have neither included advanced adenoma nor correlated the FDG uptake with the KI 67 index as markers for proliferation. We measured FDG uptake versus TNM stage, however, and found no correlation (Fig. 4). Pending further study, we might extrapolate that the trigger SUV_max ≥5 is also valid for advanced adenoma, depending on the growth rate. The hypothesis that the SUV_max correlates with the growth rate seems correct; glucose provides the energy for proliferation and is supported by the relationship between pre-operative ¹⁸FDG uptake and epidermal growth factor receptor [45]. Also, ¹⁸FDG-PET detects all cancers in patients with familial adenomatous polyposis [46]. This is still speculative, however, pending a larger study. Recently, Na et al. [47] proposed an SUV_max = 5.8 as optimal cut-off to identify a malignancy or high-grade dysplasia but warned that colonoscopy should be performed above an SUV_max = 2.5 to avoid missing a malignancy or high-grade dysplasia. This is in line with the semi-automation we propose: an SUV_max ≥5 should automatically trigger a referral for colonoscopy leaving the cases with SUV_max <5 for individual interpretation.

A great deal of expertise and resources are currently invested in establishing, testing and improving mono-organ screening methods. Screening programmes are currently in place for the early detection of oncological, cardiovascular and metabolic diseases including prostate, lung, colorectal, ovarian [48] and breast cancer, as well as arteriosclerosis, aortic aneurysm and osteoporosis. PET/CT offers the possibility of replacing most mono-organ screening methods with a single multi-organ screening exam. A single PET/CT screening appointment lasting around 1 h promises to be more accepted, efficient, effective and safe than the combined organ-specific screening techniques currently in use.

In addition, PET/CT is a promising candidate for semi-automated analysis as it acquires digital data, in vivo, at the molecular level. In the context of colorectal cancer, this cannot be said of the subjective optical interpretation at the macroscopic level required for colonoscopy. Although laxative-free CT colonography [49] and PET/CT are both non-invasive and require no bowel preparation, PET/CT seems superior for multi-organ screening and semi- or potentially full automation of the analysis, as we have discussed.