The Influence of Incidental Detection of Thyroid Nodule on Thyroid Cancer Risk and Prognosis

A Systematic Review

Je Ern Chooi; Abiramie Ravindiran; Saba P. Balasubramanian


Clin Endocrinol. 2022;96(2):246-254. 

In This Article


The initial searches yielded 3095 manuscripts. After removing non-English, nonhuman studies and duplications, 2159 manuscripts were assessed for eligibility using the inclusion and exclusion criteria (shown in Table S2). Eighteen studies reporting comparison between the incidental and nonincidental arms were included; four studies evaluated cancer risk; nine reported on prognosis; and five reported on both risk and prognosis. Due to the duplication of the patients included in the 2017[18] and 2020[19] studies by Marina et al.,[18,19] only the latter one is including in the analysis. No additional manuscripts were identified on screening the bibliographies of these eighteen studies.

Risk of Cancer in Thyroid Incidentalomas

Characteristics of Studies on Thyroid Cancer Risk. Six case-control and three retrospective cohort studies with study periods ranging from 1991 to 2016 were included. The patient eligibility criteria as reported in the individual studies were presented in Table S3. Eight were single centre studies[20–27] and one study[28] involved two centres. Six studies[20,22,23,25,26,28] were carried out in the United States whilst the rest were from the United Kingdom,[21] Australia[24] and Israel.[27] Eight studies[20–22,24–28] were carried in the adult population and one[23] in individuals under the age of 18. The definitions of modes of detection (incidental and nonincidental) as reported in the individual studies are summarized in Table S4. For the purpose of our analysis, patients with nodules discovered on thyroid cancer screening and surveillance for previously detected thyroid nodules were excluded from the analysis.[28] However, in data extraction from two manuscripts,[23,26] patients with nodules detected on thyroid cancer screening or imaging for thyroid-related reasons could not be excluded from the analysis. The nonincidental detection group included patients with symptomatic or palpable thyroid nodules (noticed by either patient or clinician) and lesions discovered in the workup of abnormal thyroid function.

Case-control Studies on Thyroid Cancer Risk. A total of 738 malignant thyroid nodules and 2508 benign thyroid nodules were included in the six case-control studies.[20–24,26,28] The odds of incidental detection in the cancer and benign groups ranged from 0.16 to 0.5 and 0.06 to 0.38, respectively (odds ratio = 0.64–2.86; statistical significance testing was not performed; shown in Table 1). The results of the individual case-control studies were described in Table S5.

Cohort Studies on Thyroid Cancer Risk. A total of 91 patients in the incidental group and 398 patients in the nonincidental group were included in the three cohort studies.[23,25,27] The risk of malignancy for thyroid nodules ranged from 4% to 23.5% in the incidental and 3.8% to 28.7% in the nonincidental groups (statistical significance testing was not performed; shown in Table 2). The results of the individual cohort studies were described in Table S5.

Quality of Studies on Thyroid Cancer Risk. The scores of study quality assessed using the modified NOS ranged from low to moderately high in the case-control studies (n = 6)[20–24,26,28] and cohort studies (n = 3)[23,25,27] with the range of 22.2%–66.7% and 44.4%–66.7%, respectively (Figure S1A,B). Major issues affecting study quality were lack of adjustment for confounding factors, exclusion of patients due to incomplete medical records, inadequate outcome assessment (cancers were diagnosed on cytology instead of histopathology) and inadequate or unclear follow up.

Meta-analysis of Thyroid Cancer Risk Studies. Meta-analysis was carried out for the case-control studies with comparable study populations and definition of incidental detection (n = 4)[21,22,24,28] (shown in Figure 2A). The summated odds ratio was 1.11 (95% confidence interval [CI] = 0.74–1.68) and I2 of 28%. Previous studies have suggested that studies scoring 5 out of 9 on NOS were considered of high risk of bias.[29,30] A second meta-analysis looking at good quality case-control studies (NOS > 50%; n = 3)[22,24,28] showed a summated odds ratio of 1.04 (95% CI = 0.63–1.70; p = .88) and I 2 of 46% (shown in Figure 2B). Publication bias was not assessed due to the low number of studies in the meta-analysis.

Figure 2.

Results of meta-analysis for (A) eligible case-control studies (B) eligible case-control studies with NOS score of more than 50%. CI. confidence interval; NOS, Newcastle–Ottawa Scale

Prognosis of Thyroid Cancers in Thyroid Incidentalomas

Characteristics of Studies on Thyroid Cancer Prognosis. Of the 14 studies included, 12 were retrospective cohort studies[19,20,22,23,27,28,31–36] and 2[37,38] were cross-sectional studies; the study periods ranged from 1998 to 2020. The inclusion and exclusion criteria for the individual studies are described in Table S6. Eight studies were carried out in the United States and the rest were from Italy,[19] Israel,[27] Korea,[32] Canada,[33] Ecuador[37] and Australia.[38] One study[23] only included individuals under the age of 18 while the others were in the adult population. Most retrospective cohort studies are single centred studies except for the study by Davies et al.[28] which was carried out in two centres. Among these cohort studies, five studies[20,22,23,27,28] also evaluated the risk of thyroid cancer.

For prognosis, incidental detection refers to cancer detected in a thyroid incidentaloma. The definitions of incidental and non-incidental detection for individual studies are shown in Table S7. Nonincidental detection includes cancer detected in a palpable or symptomatic thyroid nodule or nodule discovered through an abnormal thyroid function test. 'Unsuspected nodules found on surgery for benign thyroid disease' were considered as nonincidental nodules in this analysis regardless of how it was classified in the reported studies. However, there were a few exceptions: The study by Shakil et al.[31] where patients with thyroid cancer were discovered histologically after thyroidectomy for benign thyroid disease, the study by Marina et al.[19] where nodules were detected on the investigation of thyroid disease or during pathological examination for a benign lesion and the study by Hagag et al.[27] where patients with thyroid cancer detected on ultrasound for nonnodular thyroid disease could not be excluded from the incidental detection group.

A total of 5164 patients (2424 in the incidental group and 2740 in the nonincidental group) were identified from the 14 studies (shown in Table S8). The type of thyroid cancer was not mentioned in the studies by Iwata et al.[20] and Gupta et al.;[23] possibly because not all nodules underwent histological assessment. Two studies[32,33] focussed exclusively on PTC whilst two studies[19,35] focussed on differentiated thyroid cancer (DTC) which included papillary, follicular and Hurtle cell carcinomas. The rest of the studies[22,27,28,31,34,36–38] included patients with DTC, medullary and anaplastic carcinomas. Three direct and thirteen indirect prognostic markers were discussed in respective studies (shown in Table S8). Only 3[19,31,32] of the 13 studies provided information on patients' follow up; the median follow-up period ranged from 26.5 to 114 months.

Quality of Studies on Thyroid Cancer Prognosis. The risks of bias for the thirteen cohort studies assessed on the modified NOS ranged from moderately low to very low (66.7%–100%) while one study[32] scored 100% (shown in Figure S1C). Some criteria were not applicable to some cohort studies;[20,22,23,27,28,33–36] an example is where the main prognostic marker reported in the study is not a time-dependent marker (i.e., an indirect marker of prognosis). For the cross-sectional studies, Solis et al.[37] scored 77.8% whilst the study by Kahn et al.[38] scored the highest score of 100% indicating a low risk of bias (shown in Figure S1D). Despite the high NOS score, a meta-analysis was not done for the prognostic markers due to the heterogeneity in definitions of incidental detection, age of the population studied, subtypes of cancer included, differences in the prognostic outcomes reported and the assessment used for these prognostic markers.

Direct Markers of Prognosis. Solis et al.[37] showed that the nonincidental group had a higher risk of recurrence (16.5% intermediate and 10% high risks) as defined by the ATA 2009 risk stratification system when compared to the incidental group (6.8% intermediate and 7.6% high risks). However, the prevalence ratios of both risks were not statistically significant among both arms in the multivariate analyses (prevalence ratio [PR] = 0.98, 95% CI = 0.67–1.43; p = .902 and PR = 1.04, 95% CI = 0.82–2.16; p = .239, respectively; shown in Table S9). The study by Marina et al.[19] looking at DTC only showed that the disease-free survival was higher in the incidental group when comparing the number of events over 15 years and the time to event (hazard ratio [HR]: not available). One study[31] found that the rate of residual disease or recurrence (R/R) was significantly higher in the nonincidental detection group (21%) compared to the incidental detection group (7%) (p = .04); the incidental group also had longer progression-free survival compared to the nonincidental group (log-rank test p = .08). In the univariate analysis, thyroid cancer recurrence was lower (but not significant) in the incidental detection group (HR = 0.36, 95% CI = 0.11–1.26; p = .1). The fourth study[32] reported that more patients with nonincidentally detected PTC had a relapse when compared to the incidental group (11% and 3%, respectively; p < .001) regardless of the site of relapse but the time to recurrence was not significant (p = .681). The 5-year recurrence-free survival rates and the overall survival were higher in the incidental detection group (97% and 99%, respectively) when compared to the nonincidental group (91% and 97%, respectively; log-rank test p < .001). In the multivariate analyses, nonincidental detection was found to be an independent predictor of recurrence-free survival (HR = 2.02, 95% CI = 1.02–4.01; p = .043) and overall survival (HR = 5.84, 95% CI = 2.07–16.44; p = .001), respectively.

Indirect Markers of Prognosis.Tables S10–S12 show the 13 indirect prognostic markers (size of the nodule, histological subtypes, lymph node and distant metastasis, disease spread at the time of diagnosis, lymphovascular and capsular invasions, extra-thyroidal and extra-nodal extensions, bilaterality, multifocality/multicentricity, cancer staging, lymphocytic infiltration, MACIS and AMES score) that were assessed and compared between the incidental and nonincidental groups. Nonincidentally detected thyroid nodules were significantly more likely to be larger[32,34,35,37,38] and have higher rates of extra-thyroidal and extra-nodal extensions,[32] multifocality[37] and lymph node metastasis.[22,31,32] Interestingly, incidental thyroid nodules were reported to have more advanced disease (Stage III and Stage IV) in one study,[36] this could be explained by the lack of adjustment of age as a confounding factor. Other indirect prognostic markers were not shown to be significantly different between the two groups. The results of the individual studies were described in Table S13.