1. Ethical considerations

This study was classified as risk-free, in accordance with Colombia’s Ministry of Health Resolution No. 8430/1993. A Participation Agreement, signed prior to the study, highlighted the protection of participants’ dignity and the integrity of data, assuring participants the right to withdraw at any time without repercussions. The research emphasized adherence to inclusion criteria to ensure objectivity and minimize biases, without offering financial compensation. In line with the Helsinki Declaration, the study mandated transparent result disclosure and a commitment to reveal any conflicts of interest or funding sources that might compromise ethical integrity.

2. Establishment of the review committee

The ACE formed a review committee comprising six experienced endocrinologists, three of whom specialized in epidemiology, tasked with evaluating the submitted recommendations. The primary responsibility of this committee was to preliminarily assess the quality and relevance of these recommendations before they were advanced to the expert panel for the consensus process. To safeguard the integrity and clinical pertinence of this process, the committee excluded individuals engaged in non-clinical functions within government health departments, Health Promotion Entities (Entidades Promotoras de Salud, EPS), payer organizations, or those affiliated with companies involved in the production or distribution of medical technologies. This exclusion was aimed at reducing potential conflicts of interest.

Figure 1: Flowchart of the Delphi Process for Developing Decisiones Acertadas Recommendations in Endocrinology

Source: Authors' own elaboration

3. Receipt and initial assessment of recommendations

ACE members were invited to submit ideas and recommendations regarding “do not do” actions using the Google Forms platform. They were encouraged to ground their suggestions on uncertainty regarding utility or the lack of conclusive evidence of effectiveness or potential harm. Subsequently, a review was undertaken, applying exclusion criteria to eliminate ambiguous suggestions, opinions lacking scientific support, personal criteria without substantial basis, remarks targeting other specialties, and recommendations to address underuse, such as “Remember to...”.

4. Comprehensive review and categorization of recommendations

After applying the exclusion criteria, the review committee performed a detailed evaluation of the remaining recommendations, organizing them into categories like medication classes, diagnostic support, surgical and non- surgical procedures, devices, and others. This classification, determined by the frequency of proposals from ACE members, was designed to enhance understanding and organization.

5. Systematic literature review and synthesis of evidence

A thorough literature review was conducted using PubMed, LILACS, Embase, and Google Scholar, focusing on articles published in English, Spanish, and Portuguese from 2013 to 2023, and augmented by relevant studies provided by the review committee members. The review committee utilized the “Levels of Evidence 2011” framework from the Oxford Centre for Evidence- Based Medicine to assess the evidence backing the shortlisted recommendations. Recommendations were graded with evidence levels from I to V, reflecting the strength and depth of the available research. To avoid information bias, these evidence levels were not initially revealed to the Delphi panel but were made available during the deliberation rounds upon the panelists’ requests.

6. Development and pilot testing of an initial questionnaire

A questionnaire incorporating the first 20 preselected recommendations was crafted and piloted with three endocrinologists, each possessing over five years of clinical experience spanning academia, public hospitals, and private practice, who were not involved in the study. Feedback from this pilot was used to enhance the wording and structure of the questionnaire.

7. Selection of the expert panel

Expert panel participants from various regions of Colombia were selected via convenience sampling, based on criteria including over five years of clinical endocrinology experience with a minimum caseload of 50 patients weekly. Additionally, participants were required to demonstrate engagement in continuous education with relevant certifications, contribute to scientific dialogue through publications and conference participation, and receive peer recognition through awards and endorsements (11,12). To ensure anonymity and impartiality, participants were assigned unique identifiers, P#R#, where “P#” represents the participant number from 1 to 11 and “R#” denotes the region of origin from 1 to 6. This system aimed to minimize bias and foster an environment conducive to the unbiased exchange of professional insights and critiques.

8. Collection of responses and statistical analysis

After the pilot test, the review process continued with a two-round online survey on Google Forms. In the first round, experts rated the relevance of each recommendation on an ordinal Likert scale from 1 (completely disagree) to 9 (completely agree), with 5 indicating neutrality (12,13). A consensus for a recommendation was established at an agreement level of ≥ 70%, with disagreement noted between 31% and 69%, and a consensus against recognized at < 30%. If the first round failed to achieve sufficient consensus, the review committee advanced the top nine recommendations to a second iteration round.

During this phase, participants re-evaluated these recommendations, applying the Likert scale for new ratings. This method led to the identification of the final five recommendations that garnered the highest consensus.

9. Concluding meeting for dissemination and discussion of results

A concluding meeting was held to disseminate the study’s findings, where detailed results were shared and deliberated with the review committee, peer reviewers, ACE members, and participating scientific societies. This session solidified the consensus and collected further input. Subsequently, an action plan for the implementation and distribution of the recommendations was developed, engaging experts, the scientific community, and the target patient groups in an interactive manner. The final five recommendations, along with contributions from other scientific societies, will be made available on the following website: http://decisionesacertadas.sociedadescientificas.com

10. External validation of the final document

After finalizing the draft containing the top five recommendations, which were collaboratively shaped by all consensus participants, it underwent review by an external expert in shared decision- making and overdiagnosis from the Mayo Clinic (JPBC). This review aimed to validate the quality and relevance of the recommendations.

1. Do not order routine thyroid ultrasounds in the general population or patients with hypothyroidism when the physical examination is normal.

Results of the second iteration: Median of 9.0 and 100% agreement rate.

Justification: The recommendation against routine thyroid ultrasounds in patients without detectable abnormalities during clinical examination is grounded in the phenomenon of overdiagnosis in thyroid cancer. This has led to a significant increase in the detection of small tumors, particularly those measuring less than 1 cm, without a corresponding rise in mortality rates (14-16). Consequently, unnecessary treatments such as fine needle aspirations, thyroidectomies, and radioactive iodine therapies have become prevalent (17,18). However, these interventions not only lack significant impact on recurrence or mortality rates (19) but also have the potential to negatively impact the long-term well-being of patients (20,21).

Studies indicate that patients may experience emotional responses similar to those observed with other malignancies, despite low-risk thyroid cancer typically having a less aggressive clinical course (22-26). Additionally, it is crucial to consider the additional economic burden, as managing thyroid cancer entails substantial costs in the Colombian context (27). Therefore, it is imperative to refrain from indiscriminate thyroid ultrasounds and instead utilize them judiciously for cases with suitable indications (18). This recommendation does not apply to patients with high-risk conditions, such as mutations in the RET gene for multiple endocrine neoplasia type 2 (MEN 2), patients with three or more family members with thyroid carcinoma, and patients with prior neck radiation (28-30).

2. Do not request markers of bone turnover in patients with osteoporosis.

Results of the second iteration: Median of 9.0 and 100% agreement rate.

Justification: The recommendation against using bone turnover markers in patients with osteoporosis, whether to assess treatment efficacy and adherence or before undergoing invasive dental procedures, is based on various factors. These factors include the variability in bone turnover marker results (31,32), lack of standardization in their measurement (31), and weak correlation with fracture risk and mandibular osteonecrosis (33,34). Additionally, research suggests that these markers lack reliable predictive ability in the context of invasive dental procedures (35,36).

A meta-analysis of seven observational studies revealed that a serum type 1 collagen C-terminal telopeptide (sCTX) cutoff point of 150 pg/mL had a sensitivity of 57% (95% CI: 41-71%) and a specificity of 72% (95% CI: 64-79%). The positive likelihood ratio (LR+) was 2 (95% CI: 1.3-3.1), the negative likelihood ratio (LR-) was 0.6 (95% CI: 0.4-0.9), and the diagnostic odds ratio (DOR) was 3.4 (95% CI: 1.5-7.7). These findings suggest that as a preoperative marker, sCTX is not adequate for predicting the risk of developing drug-related osteonecrosis of the jaws (36).

Despite this recommendation, it’s crucial to acknowledge that it contradicts certain guidelines for postmenopausal osteoporosis management (37-39). Although some guidelines classify certain recommendations related to bone turnover markers as A or B, suggesting their use for assessing therapeutic compliance and efficacy of osteoporosis treatment, the evidence supporting this is limited. Hence, it’s essential to consider these factors when making clinical decisions and to recognize the challenges in practically applying biochemical bone turnover markers in our specific context.

3. Do not order basal insulin, post-load glucose, or use the HOMA index to assess insulin resistance in patients who are overweight, obese, or show clinical signs of insulin resistance.

Results of the second iteration: Median of 9.0 and 87.5% agreement rate.

Justification: Various indices, such as Homeostasis Model Assessment (HOMA), Quantitative Insulin Sensitivity Check Index

(QUICKI), Matsuda Index, and Insulin Secretion- Sensitivity Index-2 (ISSI-2), are utilized in clinical research to quantify insulin resistance, adjusting for variables like age, sex, and ethnicity (40-43). Elevated levels of these indices in research settings have been linked to clinical conditions like abdominal obesity, low HDL levels, hypertriglyceridemia, hyperglycemia, fatty liver, polycystic ovary syndrome, and hypertension (40,41,44,45). However, certain limitations exist in their clinical application:

• ◾ Lack of universal validation for clinical use (40).

• ◾ Variability in insulin measurement due to lack of standardization in immunoassay tests (42,46).

• ◾ Inconsistencies in reference intervals among laboratories (42,46).

• ◾ Potential inaccuracies due to changes in beta cell function over time (40).

• ◾ The post-load glucose insulin test may overestimate insulin resistance in as many as 25% of individuals evaluated (40,47).

• ◾ Low intraindividual reproducibility has been observed in insulin measurement (46).

These limitations in the clinical application of basal and post-load insulin support the recommendation to avoid their use in overweight, obese, or clinically insulin-resistant patients (48). Consequently, the presence of insulin resistance is typically inferred from clinical criteria like the presence of metabolic syndrome and insulin resistance syndrome (49).

Throughout this research, some experts discussed the potential utility of triglyceride levels as indirect markers of insulin resistance, either individually or in conjunction with glucose or HDL cholesterol. It was suggested that in patients with prediabetes and triglycerides equal to or greater than 150 mg/dL, there might be a higher probability of insulin resistance (50). Additionally, it was noted that the triglyceride-glucose index exhibited a more robust correlation with adiponectin levels in individuals with insulin resistance compared to other indirect indices such as HOMA-IR and QUICKI (51). Among Caucasians, an elevated triglyceride-HDL ratio, notably surpassing 3.0 for men and 2.5 for women, has been correlated with insulin resistance (52,53). However, these indices discussed by certain study participants did not undergo a Delphi process or systematic searches, unlike the focused analysis conducted for basal and post-load insulin levels in assessing insulin resistance. Consequently, due to this consensus, a routine recommendation for evaluating insulin resistance based on triglyceride levels or related indices cannot be established.

4. Do not perform routine measurements of vitamin D (25 (OH) vitamin D) in the general population.

Results of the second iteration: Median of 9.0 and 87.5% agreement rate.

Justification: Vitamin D has been associated with various effects, both skeletal and extra- skeletal, as evidenced by preclinical and observational studies. However, findings from randomized clinical trials like VITAL, ViDA, and D2d suggest that vitamin D supplementation in individuals with 25-hydroxyvitamin D (25OHD) levels at or above 20 ng/ml does not significantly impact cancer prevention, cardiovascular events, falls, or the onset of type 2 diabetes mellitus (54-58). Additionally, over 60 Mendelian randomization studies, designed to mitigate confounding biases, have yielded null effects regarding the association between genetically reduced 25OHD levels and disease risk (59).

Conversely, correcting severe vitamin D deficiency, defined as a serum 25OHD concentration below 12 ng/ml, has shown benefits. Thus, specific high-risk groups, such as those with limited sun exposure, malabsorption syndromes, or undergoing osteoporosis treatments that may predispose to hypocalcemia, may benefit from assessing their vitamin D levels. In such cases, measuring 25-hydroxyvitamin D levels can offer both cost-effectiveness and clinical relevance.

4.1. Fractures. Over the past two decades, placebo-controlled prospective studies with follow-ups of up to 5.3 years have not been able to demonstrate the efficacy of high doses of vitamin D, administered annually, quarterly, or monthly, as well as low daily doses, in preventing fractures (55,56,60-68) Table (7). Furthermore, two studies using single high doses of vitamin D reported significant increases in fracture incidents. The first study, which administered 300,000 IU of vitamin D2 annually via intramuscular injections over three years, observed a substantial increase in hip fractures among women (hazard ratio (HR) 1.82; 95% confidence interval (CI) 1.12-2.99), but not among men (61). The second study, which administered an annual oral dose of 500,000 IU of cholecalciferol, noted an overall increase in fractures (HR 1.26; 95% CI 1.00-1.59), attributed to falls (62).

Table 7: Fracture Incidence in Placebo-Controlled Clinical Trials of Vitamin D

Source: Authors’ own elaboration

Recent large-scale clinical trials, such as the VITAL trial and the D-Health study, have highlighted the ineffectiveness of vitamin D alone in reducing fracture risks (55,63). In the VITAL trial, supplementation with cholecalciferol did not significantly reduce the risk of fractures (total, non-vertebral, and hip) compared to placebo among healthy middle-aged and older adults not selected for vitamin D deficiency, low bone mass, or osteoporosis (55). Notably, a subgroup of 401 participants with serum 25(OH)D levels below 12 ng/mL, showed a lower fracture rate compared to those with higher levels. In the D-Health study, large monthly doses of 60,000 IU did not significantly alter the general fracture risk (63).

In addition, the co-administration of vitamin D with calcium has proven beneficial, reducing the risk of hip fractures (relative risk (RR) 0.61- 0.84) and all fractures (RR 0.74-0.95), according to a recent umbrella review of meta-analyses of randomized controlled trials of vitamin D (69). However, this reduction in fracture risk was not seen in studies that only assessed community- dwelling individuals or those receiving only vitamin D compared to a placebo or control, indicating that the benefits of vitamin D and calcium supplementation on fracture prevention are primarily limited to institutionalized populations (69).

4.2. Falls. In five clinical trials spanning 2 to 5 years (54,62,70-72) Table (8), various doses of vitamin D were investigated. Despite higher serum 25(OH)D concentrations in the treatment group compared to the placebo group (39 ng/Ml versus 27 ng/mL), no significant differences in fall incidence were observed between the groups.

Table 8: Falls Incidence in Placebo-Controlled Clinical Trials of Vitamin D

Source: Authors’ own elaboration

4.3. Cancer. Clinical trials using low daily doses ranging from 2000 to 4000 IU or high doses of 100,000 IU of oral cholecalciferol every four months did not demonstrate a reduction in cancer incidence over a follow-up period of up to 5.3 years (56,57,64,73-75). For instance, in the FIND study (74), which included 2,495 participants, supplementation with either 1600 IU daily or 3200 IU daily for 5 years did not decrease the incidence of invasive cancer (1600 IU/d (HR: 1.14; 95% CI: 0.75-1.72; P = 0.55) and 3200 IU/d (HR: 0.95; 95% CI: 0.61-1.47; P = 0.81)). Additionally, the D-Health trial (75), using monthly doses of 60,000 IU of cholecalciferol, did not impact cancer mortality, consistent with previous findings from Mendelian randomization studies (59) Table (9). Further analysis of the VITAL study, a large-scale randomized clinical trial, revealed a decreased cancer risk among individuals with a normal BMI (less than 25 kg/ m²), although this analysis was not adjusted for multiple comparisons. Additionally, this analysis revealed a potential reduction in cancer risk among African Americans, alongside an observed increase in cancer mortality starting from the fourth year of follow-up (57). These findings suggest that extending the follow-up period beyond four years could yield modest mortality benefits.

Table 9: Cancer Incidence in Placebo-Controlled Clinical Trials of Vitamin D

Source: Authors’ own elaboration

4.4. Cardiovascular events. The VITAL, FIND, and ViDA clinical trials failed to demonstrate a reduction in major cardiovascular events with vitamin D supplementation Table (10) (56,57,74). However, in the D-Health trial (76), the vitamin D group showed a lower incidence of these events compared to the placebo group (HR: 0.91, 95% CI 0.81-1.01), particularly among participants taking cardiovascular medications at baseline (HR: 0.84, 95% CI: 0.74-0.97; P for interaction=0.12). Additionally, the vitamin D group exhibited lower rates of myocardial infarction (HR: 0.81, 95% CI: 0.67-0.98) and coronary revascularization (HR: 0.89, 95% CI: 0.78-1.01). While no differences were noted in stroke rates (HR: 0.99, 95% CI: 0.80-1.23), these findings suggest a potential cardiovascular benefit from vitamin D supplementation, albeit with a small absolute risk difference and a confidence interval consistent with no effect.

Table 10: Cardiovascular Events Incidence in Placebo-Controlled Clinical Trials of Vitamin D

Source: Authors’ own elaboration

4.5. Diabetes. In the Tromsø and D2d studies, the administration of 20,000 IU of cholecalciferol weekly for five years and 4,000 IU daily for four years, respectively, did not significantly prevent the transition from prediabetes to type 2 diabetes Table (11) (58,77). However, detailed analysis of a subset of 103 D2d participants with initial 25-hydroxyvitamin D levels below 12 ng/mL showed a hazard ratio of 0.38 (95% CI, 0.18- 0.80) in the vitamin D supplemented group, suggesting a protective effect in those with severe deficiency (58). Moreover, a recent meta-analysis involving data from three clinical trials reported a 15% decrease in the risk of progressing from prediabetes to diabetes (58,77-79). Nonetheless, the benefits from lifestyle modifications and metformin, which decrease the progression rates by 58% and 31% respectively, are more substantial (80). Thus, the slight benefits derived from vitamin D supplementation do not support its widespread recommendation as a preventive measure for diabetes progression.

Table 11: Diabetes Incidence in Placebo-Controlled Clinical Trials of Vitamin D

Source: Authors’ own elaboration

4.6 Other outcomes. During the final consensus meeting, various outcomes were assessed, including asthma (81,82), respiratory infections (83,84), COVID-19 infection (85,86), autoimmune diseases (87), and infertility (88). Like the previously analyzed outcomes, a potential benefit for individuals with severe vitamin D deficiency was identified. However, it is critical to acknowledge that these outcomes did not undergo a systematic literature review or application of the Delphi method, therefore the conclusions do not completely encompass these results.

5. Do not routinely prescribe vitamin D, except in cases of deficiency (25 (OH) vitamin D < 12 ng/ml), osteomalacia, secondary hyperparathyroidism due to vitamin D deficiency, or in patients with osteoporosis at risk of hypocalcemia.

Results of the second iteration: Median of 9.0 and 88.8% agreement rate.

Justification: Clinical trials have shown that vitamin D supplementation does not provide significant benefits to individuals with levels above 20 ng/mL but does offer more reliable and positive results for those below 12 ng/mL. Thus, there is insufficient evidence to advocate for routine vitamin D supplementation across the general population. Recommendations should be specifically directed toward high-risk groups, including individuals with 25-hydroxyvitamin D levels below 12 ng/mL at risk of osteomalacia, those with secondary hyperparathyroidism due to vitamin D deficiency, or osteoporosis patients susceptible to medication-induced hypocalcemia (59,89,90).

During the consensus discussions, the precise definition of vitamin D deficiency was debated. Although there is consensus that levels below 12 ng/mL denote deficiency and levels above 30 ng/ mL denote sufficiency, the classification for levels between 12 and 30 ng/mL remains ambiguous (91). Factors such as racial (50) and age- related differences (51), variable sun exposure (51), differences in fat distribution, pregnancy (92,93), protein carriers, (94,95) and, the lack of standardized assays (96,97), contribute to this uncertainty. These factors hinder the ability to establish a clear consensus on what constitutes deficiency within this range.

Further deliberations emphasized the need to update existing management guidelines based on our study’s findings and a thorough literature review (98,99). A proposal was made to redefine the criteria for vitamin D insufficiency and deficiency to avoid unnecessary supplementation in cases where it is not clearly indicated. This change would allow for a more personalized and evidence-based approach to managing vitamin D levels in patients with endocrine disorders, thereby enhancing patient care and outcomes.