C.2.1. Literature Search

As noted in Appendix B, the current literature search was intended to update the draft toxicological profile for molybdenum released for public comment in 2017. See Appendix B for the databases searched and the search strategy.

A total of 2,508 records relevant to all sections of the toxicological profile were identified (after duplicate removal).

C.2.2. Literature Screening

As described in Appendix B, a two-step process was used to screen the literature search to identify relevant studies examining the health effects of molybdenum.

Title and Abstract Screen. In the Title and Abstract Screen step, 2,508 records were reviewed; 71 documents were considered to meet the health effects inclusion criteria in Table C-1 and were moved to the next step in the process.

Full Text Screen. In the second step in the literature screening process for the systematic review, a full text review of 92 health effects documents (documents identified in the update literature search and documents cited in older versions of the profile) was performed. From those 92 documents, 115 studies were included in the qualitative review.

C.5.1. Risk of Bias Assessment

The risk of bias of individual studies was assessed using OHAT’s Risk of Bias Tool (NTP 2015). The risk of bias questions for observational epidemiology studies, human-controlled exposure studies, and animal experimental studies are presented in Tables C-5, C-6, and C-7, respectively. Each risk of bias question was answered on a four-point scale:

• Definitely low risk of bias (++)
• Probably low risk of bias (+)
• Probably high risk of bias (−)
• Definitely high risk of bias (−−)

In general, “definitely low risk of bias” or “definitely high risk of bias” were used if the question could be answered with information explicitly stated in the study report. If the response to the question could be inferred, then “probably low risk of bias” or “probably high risk of bias” responses were typically used.

After the risk of bias questionnaires were completed for the health effects studies, the studies were assigned to one of three risk of bias tiers based on the responses to the key questions listed below and the responses to the remaining questions.

• Is there confidence in the exposure characterization? (only relevant for observational studies)
• Is there confidence in the outcome assessment?
• Does the study design or analysis account for important confounding and modifying variables? (only relevant for observational studies)

First Tier. Studies placed in the first tier received ratings of “definitely low” or “probably low” risk of bias on the key questions AND received a rating of “definitely low” or “probably low” risk of bias on the responses to at least 50% of the other applicable questions.

Second Tier. A study was placed in the second tier if it did not meet the criteria for the first or third tiers.

Third Tier. Studies placed in the third tier received ratings of “definitely high” or “probably high” risk of bias for the key questions AND received a rating of “definitely high” or “probably high” risk of bias on the response to at least 50% of the other applicable questions.

The results of the risk of bias assessment for the different types of molybdenum health effects studies (observational epidemiology, human-controlled exposure studies, and animal experimental studies) are presented in Tables C-8, C-9, and C-10, respectively.

C.6.1. Initial Confidence Rating

In ATSDR’s modification to the OHAT approach, the body of evidence for an association (or no association) between exposure to molybdenum and a particular outcome was given an initial confidence rating based on the key features of the individual studies examining that outcome. The presence of these key features of study design was determined for individual studies using four “yes or no” questions, which were customized for epidemiology, human controlled exposure, or experimental animal study designs. Separate questionnaires were completed for each outcome assessed in a study. The key features for observational epidemiology (cohort, population, and case-control) studies, human controlled exposure, and experimental animal studies are presented in Tables C-11, C-12, and C-13, respectively. The initial confidence in the study was determined based on the number of key features present in the study design:

• High Initial Confidence: Studies in which the responses to the four questions were “yes”.
• Moderate Initial Confidence: Studies in which the responses to only three of the questions were “yes”.
• Low Initial Confidence: Studies in which the responses to only two of the questions were “yes”.
• Very Low Initial Confidence: Studies in which the response to one or none of the questions was “yes”.

The presence or absence of the key features and the initial confidence levels for studies examining The presence or absence of the key features and the initial confidence levels for studies examining respiratory, gastrointestinal, renal, dermal, and ocular effects observed in the observational epidemiology, human-controlled exposure, and animal experimental studies are presented in Tables C-14, C-15, and C-16, respectively.

A summary of the initial confidence ratings for each outcome is presented in Table C-17. If individual studies for a particular outcome and study type had different study quality ratings, then the highest confidence rating for the group of studies was used to determine the initial confidence rating for the body of evidence; any exceptions were noted in Table C-17.

C.6.2. Adjustment of the Confidence Rating

The initial confidence rating was then downgraded or upgraded depending on whether there were substantial issues that would decrease or increase confidence in the body of evidence. The nine properties of the body of evidence that were considered are listed below. The summaries of the assessment of the confidence in the body of evidence for respiratory, hepatic, renal, alterations in uric acid levels, reproductive, and developmental effects are presented in Table C-18. If the confidence ratings for a particular outcome were based on more than one type of human study, then the highest confidence rating was used for subsequent analyses. An overview of the confidence in the body of evidence for all health effects associated with molybdenum exposure is presented in Table C-19.

Five properties of the body of evidence were considered to determine whether the confidence rating should be downgraded:

• Risk of bias. Evaluation of whether there is substantial risk of bias across most of the studies examining the outcome. This evaluation used the risk of bias tier groupings for individual studies examining a particular outcome (Tables C-14, C-15, and C-16). Below are the criteria used to determine whether the initial confidence in the body of evidence for each outcome should be downgraded for risk of bias:

  • No downgrade if most studies are in the risk of bias first tier
  • Downgrade one confidence level if most studies are in the risk of bias second tier
  • Downgrade two confidence levels if most studies are in the risk of bias third tier

• Unexplained inconsistency. Evaluation of whether there is inconsistency or large variability in the magnitude or direction of estimates of effect across studies that cannot be explained. Below are the criteria used to determine whether the initial confidence in the body of evidence for each outcome should be downgraded for unexplained inconsistency:

  • No downgrade if there is little inconsistency across studies or if only one study evaluated the outcome
  • Downgrade one confidence level if there is variability across studies in the magnitude or direction of the effect
  • Downgrade two confidence levels if there is substantial variability across studies in the magnitude or direct of the effect
• Indirectness. Evaluation of four factors that can affect the applicability, generalizability, and relevance of the studies:

  • Relevance of the animal model to human health—unless otherwise indicated, studies in rats, mice, and other mammalian species are considered relevant to humans
  • Directness of the endpoints to the primary health outcome—examples of secondary outcomes or nonspecific outcomes include organ weight in the absence of histopathology or clinical chemistry findings in the absence of target tissue effects
  • Nature of the exposure in human studies and route of administration in animal studies—inhalation, oral, and dermal exposure routes are considered relevant unless there are compelling data to the contrary
  • Duration of treatment in animal studies and length of time between exposure and outcome assessment in animal and prospective human studies—this should be considered on an outcome-specific basis
  Below are the criteria used to determine whether the initial confidence in the body of evidence for each outcome should be downgraded for indirectness:

  • No downgrade if none of the factors are considered indirect
  • Downgrade one confidence level if one of the factors is considered indirect
  • Downgrade two confidence levels if two or more of the factors are considered indirect
• Imprecision. Evaluation of the narrowness of the effect size estimates and whether the studies have adequate statistical power. Data are considered imprecise when the ratio of the upper to lower 95% CIs for most studies is ≥10 for tests of ratio measures (e.g., odds ratios) and ≥100 for absolute measures (e.g., percent control response). Adequate statistical power is determined if the study can detect a potentially biologically meaningful difference between groups (20% change from control response for categorical data or risk ratio of 1.5 for continuous data). Below are the criteria used to determine whether the initial confidence in the body of evidence for each outcome should be downgraded for imprecision:

  • No downgrade if there are no serious imprecisions
  • Downgrade one confidence level for serious imprecisions
  • Downgrade two confidence levels for very serious imprecisions
• Publication bias. Evaluation of the concern that studies with statistically significant results are more likely to be published than studies without statistically significant results.

  • Downgrade one level of confidence for cases where there is serious concern with publication bias

Four properties of the body of evidence were considered to determine whether the confidence rating should be upgraded:

• Large magnitude of effect. Evaluation of whether the magnitude of effect is sufficiently large so that it is unlikely to have occurred as a result of bias from potential confounding factors.

  • Upgrade one confidence level if there is evidence of a large magnitude of effect in a few studies, provided that the studies have an overall low risk of bias and there is no serious unexplained inconsistency among the studies of similar dose or exposure levels; confidence can also be upgraded if there is one study examining the outcome, provided that the study has an overall low risk of bias
• Dose response. Evaluation of the dose-response relationships measured within a study and across studies. Below are the criteria used to determine whether the initial confidence in the body of evidence for each outcome should be upgraded:

  • Upgrade one confidence level for evidence of a monotonic dose-response gradient
  • Upgrade one confidence level for evidence of a non-monotonic dose-response gradient where there is prior knowledge that supports a non-monotonic dose-response and a nonmonotonic dose-response gradient is observed across studies
• Plausible confounding or other residual biases. This factor primarily applies to human studies and is an evaluation of unmeasured determinants of an outcome such as residual bias towards the null (e.g., “healthy worker” effect) or residual bias suggesting a spurious effect (e.g., recall bias). Below is the criterion used to determine whether the initial confidence in the body of evidence for each outcome should be upgraded:

  • Upgrade one confidence level for evidence that residual confounding or bias would underestimate an apparent association or treatment effect (i.e., bias toward the null) or suggest a spurious effect when results suggest no effect
• Consistency in the body of evidence. Evaluation of consistency across animal models and species, consistency across independent studies of different human populations and exposure scenarios, and consistency across human study types. Below is the criterion used to determine whether the initial confidence in the body of evidence for each outcome should be upgraded:

  • Upgrade one confidence level if there is a high degree of consistency in the database

Presumed Health Effects

• Respiratory effects following long-term inhalation exposure to molybdenum oxides

  • Inadequate evidence from studies of molybdenum oxide workers (Ott et al. 2004; Walravens et al. 1979).
  • High level of evidence from chronic studies in rats and mice exposed to molybdenum trioxide (NTP 1997). Respiratory effects were not observed following acute- or intermediate-duration inhalation exposure.
• Renal effects

  • No data in humans.
  • High level of evidence of histological alterations in kidneys, alterations in renal function, and/or increased lipid levels in the kidneys in orally exposed rats (Bandyopadhyay et al. 1981; Bompart et al. 1990; Murray et al. 2014a; Rana and Kumar 1980c, 1983; Rana et al. 1980).

Not Classifiable as a Hazard to Humans

• Hepatic effects

  • Inadequate evidence of increased risk of self-reported liver conditions from a cross-sectional study (Mendy et al. 2012).
  • High evidence of no histological alterations following intermediate or chronic inhalation exposure of rats and mice to molybdenum trioxide (NTP 1997), acute oral exposure of rabbits to ammonium heptamolybdate (Bersenyi et al. 2008), or intermediate oral exposure of rats to sodium molybdate (Murray et al. 2014a;).
  • Moderate evidence of increases in clinical chemistry parameters and/or liver lipid levels in rabbits following acute oral exposure (Bersenyi et al. 2008) or rats exposed orally exposed to high doses (Rana and Chauhan 2000; Rana and Kumar 1980b, 1980c, 1983; Rana and Prakash 1986; Rana et al. 1980, 1985).
  • The hazard identification for hepatic effects was downgraded to Not Classifiable because the toxicological significance of the alterations in serum enzyme levels and lipid levels were not known and well-designed inhalation and oral laboratory animal studies have not reported histological alterations.
• Alterations in uric acid levels

  • Low evidence of an effect in cross-sectional studies (Koval’skiy et al. 1961; Walravens et al. 1979).
  • High confidence in an animal study not finding an effect (Murray et al. 2014a).
• Reproductive effects

  • Low level of evidence of male reproductive effects in cross-sectional studies (Lewis and Meeker 2015; Meeker et al. 2008, 2010).
  • Two high-quality, intermediate-duration (Murray et al. 2014a) and 2-generation (Murray et al. 2019) studies have not reported reproductive effects.
  • There is a moderate level of evidence of male and/or female reproductive effects in orally exposed rats (Fungwe et al. 1990; Pandey and Singh 2002), mice (Zhai et al. 2013; Zhang et al. 2013), and rabbits (Bersenyi et al. 2008).
• Developmental effects

  • Low evidence of an effect in a cross-sectional study. Two cross-sectional studies reported no alterations in newborn body weight (Shirai et al. 2010; Vazquez-Salas et al. 2014); one study reported decreases in psychomotor development indices (Vazquez-Salas et al. 2014).
  • Three studies in rats did not find alterations in resorptions, post-implantation losses, or fetal body weights (Jeter and Davis 1954; Murray et al. 2014b, 2019); the initial confidence levels for two of these studies were high and the third study was very low. A fourth study (initial high confidence level) involving male-only exposure found decreases in number of live fetuses and fetal body weights (Pandey and Singh 2002). The animal studies had different study designs (male only, female only, male and female exposure) making a comparison across studies difficult. Additionally, none of the animal studies evaluated potential neurodevelopmental effects, which were observed in an epidemiology study. Thus, the available data were not considered adequate for drawing a conclusion on the potential developmental toxicity of molybdenum in humans.