Predicting the in vitro dissolution rate constant of mineral wool fibers from fiber composition

Abstract

Objective

We developed predictive formulae for the in vitro dissolution rate constant k_dis of acid-soluble synthetic vitreous fibers (SVF), paralleling our earlier work with glass wools, which are typically more soluble at neutral pH. Developing simple models for predicting the k_dis of a fiber can allow prediction of in vivo behavior, aid fiber developers, and potentially reduce in vivo testing.

Methods

The k_dis of several acid-soluble SVF were determined using high simulant fluid flow/fiber surface area (F/A) conditions via a single-fiber measurement system. Four fluids were employed, varying in base composition and citrate levels. Equations predicting the k_dis were derived from fiber chemistry and dissolution measurements for two of the fluids.

Results

Testing of several fibers showed a ∼10× increase in the k_dis when citrate was included in the simulant solution. Data from tests with Stefaniak’s citrate-free Phagoloysosmal Simulant Fluid (PSF) yielded k_dis values aligned with expectations from in vivo results, unlike results from citrate-containing modified Gamble’s solution. Predictive equations relating fiber chemistry to k_dis showed reasonable agreement between the measured and predicted values.

Conclusions

Citrate inclusion in the solution under high F/A conditions significantly increased the measured k_dis. This resulted in more biorelevant data being obtained using the PSF fluid with the high F/A method used. The developed predictive equations, sufficient for fiber development work, require refinement before a recommending their use in place of in vivo biopersistence testing. Significant fit improvements are possible through additional measurements under these experimental conditions.

Keywords:

• Glass wool
• mineral wool
• Gamble’s solution
• dissolution
• simulant fluid
• synthetic vitreous fiber (SVF)

Introduction

Motivation

Predicting the dissolution rate constant, k_dis, of synthetic vitreous fibers (SVF) from the oxide chemistry of the fiber is a valuable tool for fiber developers. Measured in vitro k_dis values have been shown to correlate with fiber biopersistence, a key factor in understanding the pathogenicity of SVF (Donaldson and Lang Tran 2004; Maxim et al. 2006). Predicting k_dis from fiber chemistry allows the prediction of the in vivo long fiber clearance time, i.e. the time expected for a fiber to clear the lungs through dissolution-mediated processes. This allows optimization of fiber composition to meet regulatory requirements (EC 1997), which are based on fiber clearance times, while minimizing the need for in vivo testing during the development process. These methods also offer a path forward to the elimination of in vivo tests from fiber regulatory processes, if the in vitro results are sufficiently predictive of the in vivo results.

We recently provided an updated equation to calculate the in vitro k_dis of SVF’s from chemical composition (Potter et al. 2017). That work was performed at near-neutral pH, as is the typical environment for evaluation of glass wools. In this work we extend this earlier work by providing similar equations for mineral wool¹ fibers k_dis measured using in vitro simulant solutions at acidic pH (Eastes et al. 2000a).

In the late 1990s, a new type of SVF was introduced which surprisingly exhibited rapid clearance from the lung, defying predictions made using contemporary in vitro measurements at near-neutral pH (Knudsen et al. 1996). These fibers, such as MMVF34, when compared to the traditional mineral wool, MMVF21,^. dissolved quickly during in vitro tests at pH 4.5 (Steenberg et al. 2001). As these less biopersistent fibers have generally replaced traditional mineral wools in the marketplace, we carried out this research to add a predictive capability like that developed previously for glass wool fibers.

The rapid clearance of mineral wool fibers such as MMVF34 is now understood to be related to the actions of alveolar macrophages upon fibers, which offer two acidic environments relevant to fiber dissolution. These acidic environments include the phagolysosome within the macrophage, and the sealed area between the macrophage and fiber associated with frustrated phagocytosis (Eastes et al. 2007). Of these two environments, the latter is more relevant to the health effects of fibers as it relates directly to the clearance of long fibers, i.e. those fibers of sufficient length to lead to frustrated phagocytosis.

Measuring k_dis

The k_dis measured by current in vitro procedures depends primarily on three factors, which in turn align with the classic factors correlated to silicate glass dissolution in aqueous media: (Paul 1982; Mattson 1994a)

• fluid chemical composition (pH, ionic strength, complexing agents)

• ratio of fluid flow rate to fiber surface area (F/A),

• oxide chemistry of the measured fiber

The solution temperature is an additional technical factor. In practice, it has been fixed at 37 °C in existing in vitro studies, consistent with the in vivo environment. Thus, it is typically not addressed in context of in vitro results, but would also be expected to effect the dissolution rate if significantly varied (Paul 1982)

In this study, we considered the use of two existing methods for measuring k_dis. The guideline of Sebastien et al. for measuring the k_dis of SVF in vitro was developed concurrent to the development of HT-type mineral wool fibers (Sebastian 2002; Guldberg et al. 2003). Following this guideline, a researcher flows a modified Gamble’s solution acidified to pH 4.5 (designated as EU 4.5 in this work) and an F/A target of 0.03 µm/s over the target SVF. Periodic analysis of the simulant fluid composition over time for glass constituents, particularly silicon, allows calculation of the k_dis.

Another methodology was described by Potter (2000). While less commonly practiced, this guideline allows direct determination of fiber diameter change at high F/A (> 1000 µm/s) conditions that offer less sensitivity of the k_dis value to flow rate variation and rapid fluid replenishment near the fiber (Mattson 1994a). Although more complex in practice, the direct microscopic determination of diameter also offers the advantage of direct observation of diameter change, vs. inference from measurement of dissolution products.

Simulant fluids

While several simulant fluids have been applied to the study of fiber and particle dissolution (Innes et al. 2021), a modified version of Gamble’s solution has historically been used to simulate extracellular fluids when testing in vitro solubility of fibers and particles. Common practical modifications include reducing calcium to minimize the probability of precipitates, and the addition of formaldehyde or other agents to control bacterial action (Kanapilly et al. 1973). Using the guideline described by Sebastian et al. the choice of a modified Gamble’s solution also produced stable and reproducible measuring conditions at the low F/A conditions selected. Within this measurement envelope, the relative ranking of fibers was also found to be independent of the fluid composition or the range of F/A tested (Guldberg et al. 2003).

Although the Sebastian et al. guideline allows the estimation of the lung clearance time of a fiber based on its dissolution rate relative to that of other fibers that have been tested in vivo, it would be preferable to have an in vitro k_dis measurement guideline that operates at higher F/A conditions. Studies on glass wool have shown that at the F/A ratios achieved in single-fiber tests, minimal variation in k_dis is observed relative to small changes in flow rate, unlike low F/A conditions, potentially improving test robustness (Mattson 1994b). To facilitate this goal, we chose to examine in further detail the simulant fluid composition as a facilitating factor for dissolution rate control at the higher F/A per the Potter guideline.

In an in vitro test, elevating F/A promotes rapid replenishment of fluid at the fiber-fluid interface, preventing local buildup of dissolved glass constituents from impacting the dissolution rate, a known challenge in glass dissolution studies (Paul 1982). Since the publication of the guideline described by Sebastien et al. it is not clear that we have learned significantly more about the chemical composition of the in vivo environments of alveolar macrophages than was known during the guideline development (Stefaniak et al. 2005; Innes et al. 2021). There is evidence that an F/A higher than the standard 0.03 µm/s in the guideline is appropriate to mimic the in vivo environment. Christiansen found that the calcium content of murine lysosomes was restored on a time scale of minutes, which is consistent with rapid fluid turnover (Christensen et al. 2002). Silver found that activated rat macrophages reduced the pH in the area between itself and the culture dish to which it had attached by 0.5 to 2 pH units per hour, suggesting rapid fluid exchange between macrophage and medium (Silver et al. 1988). Stefaniak, using a novel fluid (Phagoloysosmal Simulant Fluid (designated as PSF in this work), found that the dissolution rate of BeO particles in acellular in vitro testing agreed well with that in murine macrophage cellular in vitro dissolution testing under conditions in which the final dissolved Be concentration was low (0.05 ppm), and the pH remained constant throughout the testing (Stefaniak et al. 2005).

Considering possible simulant fluids, the most significant difference noted between the EU 4.5 and PSF is that EU 4.5 contains 0.52 mmol/L citrate, while PSF is citrate-free. While being a part of the buffer in Gamble’s solution, it is well known that citrate at lower pH increases the dissolution rate of SVFs, such as that of MMVF 34 and HT type fibers, as well as many current mineral wool formulations. Earlier work suggests a significant, perhaps 10×, reduction in the dissolution of mineral wools when using citrate-free solutions vs. citrate containing fluids (Mogensen 1984; Barly et al. 2019; Sauer et al. 2021). Increasing the citrate concentration in Gamble’s solution with HT fibers was specifically shown to significantly increase dissolution rates (Guldberg et al., 1998; Steenberg et al. 2001)

Materials and methods

Experiment design

We report on three experiments in this work:

• A screening study using a single fiber type

  in a 2 × 2 factorial format to elucidate the impact of simulant fluid and the effect of citrate upon the measured k_dis

• A screening study using four fibers that were previously tested in vivo to evaluate the hypothesis that a higher F/A method with PSF would produce results consistent with expectations from in vivo biopersistence testing.

• Measurements of the k_dis using both PSF and EU 4.5 and many fibers to develop a dataset sufficient to develop new predictive relationships for k_dis as a function of fiber composition.

Synthetic vitreous fibers

The compositions of the fibers tested in the initial experiments are listed in Table 1. These fibers were of interest for the screening experiment, as all of them have been subjected to in vivo tests in previously reported studies (Eastes et al. 2000b). The oxide chemistry for the full set of fibers used in the coefficient development is provided in the Supplemental Data Table. All tested fibers were produced without organic binders or other additives applied during fabrication, so no treatment to remove organics was applied prior to dissolution testing. The fiber diameter was directly measured during testing, with individual fibers typically in the 2–12 micron diameter range before testing.

Table 1. Oxide chemistry of several synthetic vitreous fibers (SVF) used in this study.

Oxide	MMVF34 (wt %)	QFHA19 (wt %)	QFHA22 (wt %)	QFHA25 (wt %)
SiO2	38.9	40.1	41.6	38.3
Al2O3	23.2	18.9	21.6	22.7
CaO	15.0	19.1	15.6	25.2
MgO	9.6	11.6	9.8	6.6
Na2O	1.9	1.8	1.7	0.7
K2O	0.8	0.5	0.4	0.4
Fe2O3	7.4	5.9	7.3	4.4
TiO2	2.1	1.6	1.7	1.2
P2O5	0.4	0.0	0.0	0.0

Notes: MMVF34 data as per reference (Barly et al. 2019). QFHA19, QFHA22, QFHA25 data as per reference (Sauer et al. 2021).

Simulant fluids

We chose to investigate Stefaniak’s PSF as a possible alternative to EU 4.5 as we performed the measurements to develop a predictive coefficient. This was based on the good agreement between the acellular and cellular in vitro dissolutions of the BeO particles (Stefaniak et al. 2005) under dilute conditions. We also slightly modified Stefaniak’s PSF pH to 4.8, based on the pH data in the literature related to phagolysosomes (Silver et al. 1988; Nyberg et al. 1989a, 1989b, 1992)

The fluid compositions are listed in Table 2. Simulant fluids differed in the buffers used and in several other components. The details of the fluid recipes are provided in the Supplemental Data. All were prepared from deionized water and reagent-grade chemicals. Inorganic components are typically added as stock solutions of known concentrations per the formulary as listed in the Supplemental Materials. The modified EU 4.5 formulation followed that of Sebastian et al. The PSF formulation followed that of Stefaniak et al. with the pH adjusted to 4.8 as above.

Table 2. Solution chemistry of the simulant fluids used within this work.

Oxide	EU 4.5 With citrate (ppm)	EU 4.5 Without citrate (ppm)	PSF Without citrate (ppm)	PSF With citrate (ppm)
Na^+	3552	3552	2685	2685
K^+	–	–	782	782
Ca^+2	7.9	7.9	7.91	7.91
Mg^+2	25	25	–	–
Cl^−	4500	4500	4034	4034
HCO^3−	1416	1416	–	–
${\text{PO}}_4^{3-}$	100	100	95	95
${\text{SO}}_4^{-2}$	53	53	48	48
Citrate^−1	98	–	–	98
Tartrate^−1	116	116	–	–
Lactate^−1	139	139	–	–
Pyruvate^−1	136	136	–	–
glycine^−1	118	118	444	444
H2CO	799	799	1998	1998
CH3OH	270	270	675	675
Pthalate^−1	–	–	3303	3303

Note: PSF: Stefaniak’s Phagoloysosmal Simulant Fluid.

Dissolution rate constant measurement

We chose to use the optical method of Potter for this work which provided a higher F/A (>1000 µm/s) than the Sebastian et al. guideline (Potter 2000). The flow cell used, as illustrated in Figure 1 and Figure 2, was an updated design from our previous studies. The updated cell maintained the same flow geometry while being substantially designed to minimize the potential for cell leakage. These updates did not change the geometry in the region of fluid–fiber interaction, but did improve reliability, and enabled the use of the available digital optical microscopes. Further details related to the cell design are included in the Supplementary Material.

Figure 1. Schematic of the updated optical method cell. The fibers were mounted across the slot located mid-cell. Design schematics are provided in the Supplemental Materials.

Figure 2. Typical overview of the flow cell with several fibers mounted in the fluid path (SVF: QFHA22, Fluid: PSF) with the arrows annotating single fibers.

Dissolution rate constants were determined using direct measurement of the fiber diameter change over time when exposed to flowing fluid at 37 °C. The flow rates were fixed at ∼120 mL/day to provide an F/A of >1000 µm/s. The pH of the fluid input was held within 0.1 pH units of the target. Diameter measurements on ∼10 fibers per sample were made by directly observing the fibers in the test cells using a Keyence VHX-7000 digital optical microscope, calibrated via an appropriate stage micrometer. The diameter of the fiber over a short section was then determined using the NIH ImageJ software.

Figure 3 presents an image of a typical fiber as tested during a diameter measurement. From the measured change in fiber diameter over time, k_dis was calculated, following Potter, Equations (1)(1) $${\mathrm{t}}_{\text{dis}}=\mathrm{\hspace{0.17em}}\mathrm{t}/{\mathrm{d}}_{\mathrm{o}}-\mathrm{d}$$(1) and (2)(2) $$.{\mathrm{k}}_{\text{dis}}=\mathrm{\hspace{0.17em}}10^5*\mathrm{r}/24*2*{\mathrm{t}}_{\text{dis}}$$(2) , as per Potter (2000): (1) $${\mathrm{t}}_{\text{dis}}=\mathrm{\hspace{0.17em}}\mathrm{t}/{\mathrm{d}}_{\mathrm{o}}-\mathrm{d}$$(1) where:

Figure 3. Typical micrographs of an individual fiber during testing. The red box represents the length over which the diameter was measured at each time step.

• t_dis is the time in days to dissolve a 1-micron diameter fiber

• t is the dissolution time in days

• d_o-d is the change in diameter over the dissolution time t in microns (2) $$.{\mathrm{k}}_{\text{dis}}=\mathrm{\hspace{0.17em}}10^5*\mathrm{r}/24*2*{\mathrm{t}}_{\text{dis}}$$(2)

where:

• k_dis is the dissolution rate constant in ng/cm²/hr

• ρ is the fiber density in gr/mL

Data reduction

The k_dis for each fiber was calculated following the method outlined by Potter (2000). To develop a predictive ability based on fiber chemical composition, we fitted the fiber composition vs. k_dis data to the form of Equation (3)(3) $${\hspace{0.17em}\log \hspace{0.17em}\mathrm{\hspace{0.17em}}\mathrm{k}}_{\text{dis}}=\mathrm{\hspace{0.17em}}100·{\sum }_{\mathrm{i}}{\mathrm{P}}_{\mathrm{i}}·{\mathrm{W}}_{\mathrm{i}}/{\sum }_{\mathrm{i}}{\mathrm{W}}_{\mathrm{i}}$$(3) , following earlier work. (EC 1997) (3) $${\hspace{0.17em}\log \hspace{0.17em}\mathrm{\hspace{0.17em}}\mathrm{k}}_{\text{dis}}=\mathrm{\hspace{0.17em}}100·{\sum }_{\mathrm{i}}{\mathrm{P}}_{\mathrm{i}}·{\mathrm{W}}_{\mathrm{i}}/{\sum }_{\mathrm{i}}{\mathrm{W}}_{\mathrm{i}}$$(3) where:

• k_dis is the dissolution rate constant in ng/cm²/hr

• P_i is the coefficient for oxide i

• W_i is the mass percent of oxide i in the glass

• Σ_iW_i is the sum of all oxides for which P_i was determined.

The coefficients P_i and the 80% confidence interval were determined by applying standard linear regression techniques to a set of measured k_dis for fibers of known chemical composition. Calculations were performed using the Excel software.

Results

Screening study

Given the known impact of citrate level on the dissolution of the target fibers in this study, we first performed a brief screening study to quantify the impact on a model fiber. Our hypotheses were:

• The EU 4.5 fluid would produce significantly different dissolution rates than the PSF under the studied F/A and other conditions.

• The presence of citrate in either fluid would increase the k_dis over citrate-free solutions

The screening results are presented in Table 3. We found that the data clearly supported the hypothesis that the citrate-containing EU 4.5 fluid would produce significantly different dissolution rates than the PSF within the other method parameters. The ∼10× higher rate measured in fluid EU 4.5 fluid exceeds the uncertainty of the measurements. The data also showed that the citrate-free versions of each fluid, when compared with the citrate-containing versions, yielded a significant difference. A ∼10× higher k_dis was observed in the citrate-containing version of either simulant fluid, compared to a citrate-free version.

Table 3. Dissolution rate constant data from the screening experiment of QFHA19 fiber, showing significant increase in k_dis with citrate addition.

Test fluid	Citrate (mmol/L)	Kdis (ng/cm^2/hr)
EU 4.5 (no citrate)	−0-	1860 ± 1195
EU 4.5	0.52	22 288 ± 844
PSF	−0-	2165 ± 194
PSF (citrate added)	0.5	32 615 ± 3396

Note: PSF: Stefaniak’s Phagoloysosmal Simulant Fluid; K_dis: dissolution rate constant.

Subjectively, during routine in-test monitoring, it was noted that the PSF pH stability was a significant advantage. While occasional adjustments (acidification) of the EU 4.5 fluid in the feed reservoir were necessary, this was rarely if ever needed with the PSF reservoir.

Comparison to in vivo results

Expanding on the initial screening results, we tested the four fibers listed in Table 1 in both EU 4.5 and PSF. The results are summarized in Table 4. As expected, following the screening test, differences in the measured k_dis fluid vs. fluid were typically 10× or greater, with the rates in the citrate-containing EU 4.5 consistently greater than those measured in the citrate-free PSF. Table 4 also includes a dissolution rate constant (K_vv) calculated for each fiber. The K_vv value is based on a model developed by measuring the long fiber clearance rate during in vivo biopersistence testing and calculating the in vivo dissolution rate necessary to produce the observed rate of long fiber clearance (Eastes et al. 2000b). This provided a basis for comparing the various in vitro k_dis values to what would be expected during regulatory in vivo tests. The results obtained with PSF in three of four cases were higher (2x to 4x) than the calculated K_vv for the same fiber. This clearly contrasted the greater observed difference in the data measured with EU 4.5, where k_dis was higher in all four cases by 10× to 100× than was measured from in vivo biopersistence testing data.

Table 4. Comparing measured dissolution rate constants obtained for several mineral fibers to those calculated from the Eastes model from in vivo data.

Fiber	Kdis in EU4.5 (ng/cm^2/hr)	Kdis in PSF (ng/cm^2/hr)	${\mathrm{K}}_{\text{vv}}^{\mathrm{a}}$ (ng/cm^2/hr)
MMVF34	42 474	157	698
QFHA19	22 288	2165	974
QFHA22	35 734	1008	406
QFHA25	48 481	1462	337

Note: PSF: Stefaniak’s Phagoloysosmal Simulant Fluid; K_dis: dissolution rate constant; K_vv: dissolution rate constant from in vivo biopersistence testing.

^aIn this work, we use the ‘K_vv’ nomenclature to designate a dissolution rate constant calculated from in vivo data per Eastes et al. This was done to differentiate vs. dissolution rate constants ‘K_dis’ derived from in vitro dissolution measurements.

Coefficient development

After completing the screening test, we measured the k_dis of several mineral wool fibers in the citrate-containing EU 4.5, and the citrate-free PSF s. The full dataset is included in the Supplemental File. The coefficients fitted to Equation (3)(3) $${\hspace{0.17em}\log \hspace{0.17em}\mathrm{\hspace{0.17em}}\mathrm{k}}_{\text{dis}}=\mathrm{\hspace{0.17em}}100·{\sum }_{\mathrm{i}}{\mathrm{P}}_{\mathrm{i}}·{\mathrm{W}}_{\mathrm{i}}/{\sum }_{\mathrm{i}}{\mathrm{W}}_{\mathrm{i}}$$(3) for both fluids are listed in Table 5. Figure 4 and Figure 5 compare the predicted values of k_dis with the measured values.

Figure 4. The dissolution rate constant, k_dis, calculated by Equation (3)(3) $${\hspace{0.17em}\log \hspace{0.17em}\mathrm{\hspace{0.17em}}\mathrm{k}}_{\text{dis}}=\mathrm{\hspace{0.17em}}100·{\sum }_{\mathrm{i}}{\mathrm{P}}_{\mathrm{i}}·{\mathrm{W}}_{\mathrm{i}}/{\sum }_{\mathrm{i}}{\mathrm{W}}_{\mathrm{i}}$$(3) using the EU 4.5 coefficients compared to the measured value. The solid line is the ideal fit; the 80% confidence interval is an interval of x/3.9.

Figure 5. The dissolution rate constant, k_dis, calculated by Equation (3)(3) $${\hspace{0.17em}\log \hspace{0.17em}\mathrm{\hspace{0.17em}}\mathrm{k}}_{\text{dis}}=\mathrm{\hspace{0.17em}}100·{\sum }_{\mathrm{i}}{\mathrm{P}}_{\mathrm{i}}·{\mathrm{W}}_{\mathrm{i}}/{\sum }_{\mathrm{i}}{\mathrm{W}}_{\mathrm{i}}$$(3) using the PSF coefficients compared to the measured value. The solid line is the ideal fit; the 80% confidence interval is an interval of x/6.9. PSF, Stefaniak’s Phagoloysosmal Simulant Fluid.

Table 5. Coefficients for the calculation of the fiber dissolution rate constant, k_dis, in EU 4.5 and PSF.

Oxide^a	EU4.5 fluid			PSF
	Coefficient, Pi	Standard error	Fiber range (wt %)	Coefficient, Pi	Standard error	Fiber range (wt %)
SiO2	−0.0989	0.0145	37.0–60.7	−0.0705	0.0399	37.0–56.5
Al2O3	0.1412	0.0137	0.5–37.0	0.0735	0.0262	0.5–25.0
CaO	0.1355	0.0091	0.0–37.5	0.1210	0.0417	15.0–34.8
MgO	0.1698	0.0330	5.0–21.0	0.0918	0.0592	5.0–20.4
Na2O	0.3861	0.1214	0.0–5.4	−0.0080	0.2601	0.0–4.0
K2O	0.2559	0.2804	0.0–1.4	−0.4677	0.2543	0.0–2.1
Fe2O3	0.0911	0.0624	0.0–6.9	0.1483	0.0751	0.4–9.3
FeO + MnO	0.2000	0.0452	0.0–7.2	0.1131	0.1093	0.3–7.5
TiO2	−0.3665	0.1568	0.0–3.0	−0.0237	0.4131	0.0–2.1
P2O5	0.3301	0.1234	0.0–5.4	–	–	–

Note: PSF: Stefaniak’s Phagoloysosmal Simulant Fluid.

^aAll oxide concentrations are as weight %.

Discussion

Our intent at the outset of this work was to simply measure many fibers with single-fiber method and the citrate-containing EU 4.5 fluid, to develop a k_dis to fiber chemistry correlation. Our initial measurements of several mineral wools per the methodology employed in this study, using the citrate-containing EU 4.5 fluid produced the results that led to the expansion of this work. From previous study, we expected a minimal change in rate vs. expectations from studies using the Sebastien et al. method. We were surprised to measure k_dis in the EU 4.5 fluid in the range 20,000–1,000,000 ng/cm²/hr, several orders of magnitude higher than expectations set from our current understanding of in vivo behavior of these fibers. These high k_dis values, when used to predict fiber clearance times, would for example predict the complete dissolution of a 1 µm diameter fiber in 1 to 7 h, vs. the ∼10 days expected from corresponding in vivo studies. This discrepancy substantially motivated our further study and this report.

Previous work suggested that the inclusion of citrate in simulant fluids significantly increased the in vitro dissolution rate of HT fibers and similar SVF which provided a plausible hypothesis for these elevated rates (Mogensen 1984). In turn, this motivated our decision to explore the PSF fluid and verify that the citrate content was a significant factor. The significance of the citrate content on the measure k_dis via the single fiber method was supported by the results of the screening study, which showed a > 10× increase in rate when citrate, at the level in the EU 4.5 fluid, was present in either fluid.

This finding and the work with the PSF suggested that the combination of the single-fiber method with PSF produced results more closely aligned with those observed during in vivo testing. Given the clear impact of citrate inclusion observed, the results also suggest that, if tested, a citrate-free version of EU 4.5 (modified Gamble’s solution) could work equally in the context of a high F/A method. This is encouraging, as the goal of this type of testing is to predict in vivo biopersistence measurements, which are the basis for current regulation in the SVF space.

While pH stability in practice was not considered in our experimental design, pH can have a significant impact on glass dissolution tests and k_dis measurements (Mattson 1994a). This observation may indicate an important practical advantage to PSF in this test or any k_dis test method.

Predictive equations

The equations following the Equation (3)(3) $${\hspace{0.17em}\log \hspace{0.17em}\mathrm{\hspace{0.17em}}\mathrm{k}}_{\text{dis}}=\mathrm{\hspace{0.17em}}100·{\sum }_{\mathrm{i}}{\mathrm{P}}_{\mathrm{i}}·{\mathrm{W}}_{\mathrm{i}}/{\sum }_{\mathrm{i}}{\mathrm{W}}_{\mathrm{i}}$$(3) , using the coefficients in Table 5, are the best estimates of the effects of the various oxides on the measured k_dis. Both sets of coefficients, particularly those for the PSF, should be considered as preliminary. We publish this work in part hoping to stimulate further work in this area by other researchers, and on the assumption that some predictive ability is better than none. The issues of concern with these coefficients are as follows:

1. The prediction intervals were large. A predicted value of 100 for a PSF, for example, corresponds to a measured k_dis between 14 and 690 ng/cm²/hr.

2. Many of the coefficients for PSF are smaller than their corresponding standard error.

3. The PSF coefficients for Na₂O and K₂O are certainly questionable because we would expect each of these oxides to increase the dissolution rate in an acid solution, as they do in the EU 4.5 fluid.

This last point can be visualized more clearly in Figure 6. Ideally, we expect similar magnitudes and directionalities for the oxides. As Na₂O and K₂O are expected to depolymerize the silicate network of the fiber and render it more vulnerable to hydrolytic attack, we would expect large, positive coefficients, which were clearly not observed after data reduction.

Figure 6. Visual comparison of the coefficients by fluid, showing general agreement within the major oxides.

The common factor with each of the identified issues was the limited data points on which the fits were based and the ranges of variations of the individual oxides within the dataset. Measurements of additional fiber compositions should allow improved coefficients to be determined and the predictive power to increase.

Further development of the coefficients for PSF is clearly needed. To date, our measurements at high F/A conditions via the Potter method, using the PSF fluid, yield dissolution rates that are in better agreement with the K_vv values calculated for each fiber than those in the EU 4.5 fluid. We see this as an important point, as accurately and more precisely reproducing the in vivo result via an in vitro test provides a more acceptable in vitro/in vivo correlation (IVIVC), and thus greater potential for avoiding future in vivo tests.

The results presented here do not invalidate earlier conclusions about the health effects of fibers based on in vitro k_dis measurements using EU 4.5 fluid with the Sebastian guideline. As the Sebastian differs in F/A, fluid selection, cell design, etc. we cannot simply conclude one guideline is preferred. We believe that the data show that the use of PSF in combination with high F/A conditions, for evaluating mineral wool fibers, provides a more relevant correlation to the expected K_vv values than that obtained with the EU 4.5 fluid per the test method used herein.

Conclusions

Under the high F/A conditions in this study, it was clear that the inclusion of a significant amount of citrate in the simulant fluids significantly increased the measured k_dis of the measured SVF. This confirms the observations of previous studies, while illustrating the criticality of fluid composition to the overall in vitro measurement guideline.

The combination of the high F/A and PSF in this study produced k_dis values within an order of magnitude of the values expected from in vivo biopersistence testing, following estimates based on Eastes’ method (Eastes et al. 2000b). This suggest that the use of a low- or citrate-free fluid, such as Stefaniak’s PSF, with following the Potter guideline for k_dis determination has potential to produce results similar to standard in vivo tests. These results, coupled with the excellent pH stability of the PSF fluid during testing, suggest that further development of the Potter guideline to include PSF coupled with high F/A is warranted. Comparative study with typical SVF, the PSF fluid, and a range of flow conditions, to understand whether PSF offers advantage in other in vitro guidelines is also suggested. Broader study of other fluid components as well as pH would make sense in the context of such further studies, given the demonstrated sensitivity of these results to fluid make-up.

While this work was sufficient to develop a preliminary ability predict k_dis from fiber chemistry within the range of relevant SVF, the uncertainty in the resulting fits was mildly disappointing. We believe that the existing relationships are sufficient for fiber development study. However, significant refinement is required before they can be used in place of in vivo biopersistence testing. While experimental uncertainty is a natural cause to consider, we believe that additional measurements per guideline with a broader set of fibers will allow significant improvement in fit.

Based on these findings, we would continue to work in this area 1) to further evaluate the validity of PSF under high F/A conditions as a mimic for the lung environment relevant to long fiber clearance in the lung, and 2) to improve our ability to predict the fiber k_dis under these conditions directly from the chemical composition of the fiber, as we move toward the goal of identifying methods sufficient to eliminate the need for in vivo biopersistence testing of SVFs.

Author contributions

RMP: Study design, experiment design, data analysis, manuscript preparation and editing; JWH: Study design, experiment design, manuscript preparation and editing; JGH: Study design, manuscript preparation and editing.

Abbreviations
SVF	=	Synthetic Vitreous Fiber Silicate fibers blown, drawn, or spun from a melt which retain an amorphous structure; typically including glass wool, rock wool, stone wool, and slag wool used for insulating purposes etc.
EU 4.5	=	Our internal designation for the modified Gamble’s solution following the Sebastian et al guideline
PSF	=	Our internal designation for the Phagolysosomal Simulant Fluid per Stefaniak’s work
F/A	=	Flow (volume flow rate) ÷ Surface Area of the tested fiber; a critical parameter in dissolution study design for glass in general, and fibers in particular
Kdis	=	Dissolution rate constant following for example; also described as ‘dissolution rate coefficient’ in the Sebastian et al. guideline
HT-type fibers	=	Higher Al2O3 mineral wool fibers as per Knudsen et. al. and similar
MMVF34	=	Example of an HT-type Higher Al2O3 mineral wool fiber with published in vivo test results (Hesterberg et al. 1998)
QFHA19, 22, 25	=	Higher Al2O3 mineral wool fibers with published in vivo test results
Kvv	=	Dissolution rate constant calculated from long fiber clearance data produced in an in vivo test per Eastes’
IVIVC	=	In vitro – in vivo correlation

Acknowledgments

We thank Andy Broderick and Heather Kessler of Owens Corning, who performed the measurements reported herein, as well as adding insights that aided in cell design. We also acknowledge Adam Davis of Owens Corning for his assistance with the detailed design and 3D printing of our updated single-fiber test cells.

Disclosure statement

All authors are current or former employees of Owens Corning, a company engaged in the manufacture of synthetic vitreous fibers.

Data availability statement

All data generated or analyzed during this study are included in this published article [and its Supplementary Information Files].

Additional information

Funding

This work was funded exclusively by Owens Corning.

Notes

1 In this report, we will take ‘mineral fibers’ or ‘mineral wool’ to refer to synthetic vitreous fibers (SVF) typically based on slag and stone melt formulas, as typical of those used for thermal insulation.